Inicial

requirements.txt

@@ -1,3 +1,9 @@
-altair
-pandas
-streamlit
+streamlit>=1.32.0
+pandas>=2.1.0
+numpy>=1.26.0
+plotly>=5.18.0
+scikit-learn>=1.4.0
+anthropic>=0.40.0
+duckduckgo-search>=6.0.0
+kaleido>=0.2.1
+

src/CLAUDE.md

@@ -0,0 +1,41 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Running the Application
+
+```bash
+pip install -r requirements.txt
+streamlit run streamlit_app.py
+```
+
+There are no tests, no build step, and no linting configuration.
+
+## Architecture
+
+Streamlit football analytics app (Spanish UI) for match prediction and Kelly Criterion betting recommendations. Data source: [football-data.co.uk](https://www.football-data.co.uk).
+
+### Project structure
+
+```
+sta20/
+├── streamlit_app.py      # Entry point: page config + UI orchestration (Steps 1-6)
+├── src/
+│   ├── config.py         # All constants, URLs, column mappings, colors, thresholds, RF hyperparams
+│   ├── data_loader.py    # cargar_fixtures(), cargar_historico(), ultimos_n()
+│   ├── statistics.py     # calcular_estadisticas(), construir_comparacion(), colorear_*, mostrar_*
+│   ├── charts.py         # All grafico_*() Plotly functions
+│   ├── models.py         # preparar_dataset_rf(), entrenar_rf_equipo(), mostrar_rf_equipo()
+│   └── betting.py        # calc_probs(), kelly()
+├── requirements.txt
+└── .gitignore
+```
+
+### Key design decisions
+
+- **`src/config.py` is the single source of truth** for every magic number, color, URL, and column mapping. Edit constants there, not in the modules.
+- `cargar_fixtures()` and `cargar_historico()` use `@st.cache_data`; `entrenar_rf_equipo()` also caches — avoid mutating their return values.
+- `calc_probs()` blends probabilities: **40% home team history + 40% away team history + 20% league average**. Falls back to league average when a team has fewer than `MIN_PARTIDOS_EQUIPO` (3) matches.
+- Separate `RandomForestRegressor` models are trained per team using rolling-window features (configurable 3–15 matches via `RF_N_ROLLING_*` constants).
+- The UI flow in `streamlit_app.py` uses `st.stop()` after each mandatory selection so the page never renders with missing data.
+- Local-scope style functions (`_color_kelly`, `_color_ve`) in `streamlit_app.py` are intentional — they exist only where used and depend on config constants imported at the top.

src/dataset_partidos_procesados.csv

@@ -0,0 +1,51 @@
+Div,Date,Time,HomeTeam,AwayTeam,Referee,B365H,B365D,B365A,BFDH,BFDD,BFDA,BMGMH,BMGMD,BMGMA,BVH,BVD,BVA,BWH,BWD,BWA,CLH,CLD,CLA,LBH,LBD,LBA,PSH,PSD,PSA,MaxH,MaxD,MaxA,AvgH,AvgD,AvgA,BFEH,BFED,BFEA,B365>2.5,B365<2.5,P>2.5,P<2.5,Max>2.5,Max<2.5,Avg>2.5,Avg<2.5,BFE>2.5,BFE<2.5,AHh,B365AHH,B365AHA,PAHH,PAHA,MaxAHH,MaxAHA,AvgAHH,AvgAHA,BFEAHH,BFEAHA,B365CH,B365CD,B365CA,BFDCH,BFDCD,BFDCA,BMGMCH,BMGMCD,BMGMCA,BVCH,BVCD,BVCA,BWCH,BWCD,BWCA,CLCH,CLCD,CLCA,LBCH,LBCD,LBCA,PSCH,PSCD,PSCA,MaxCH,MaxCD,MaxCA,AvgCH,AvgCD,AvgCA,BFECH,BFECD,BFECA,B365C>2.5,B365C<2.5,PC>2.5,PC<2.5,MaxC>2.5,MaxC<2.5,AvgC>2.5,AvgC<2.5,BFEC>2.5,BFEC<2.5,AHCh,B365CAHH,B365CAHA,PCAHH,PCAHA,MaxCAHH,MaxCAHA,AvgCAHH,AvgCAHA,BFECAHH,BFECAHA
+E1,17/03/2026,19:45,Watford,Wrexham,A Herczeg,2.2,3.4,3.0,2.3,3.3,3.0,2.2,3.4,3.2,2.3,3.3,3.1,2.25,3.4,3.1,,,,,,,,,,2.36,3.45,3.2,2.27,3.34,3.07,2.38,3.45,3.35,2.1,1.73,,,2.1,1.85,2.0,1.77,2.1,1.88,-0.25,2.0,1.85,,,2.0,1.93,1.92,1.85,2.04,1.93,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E1,18/03/2026,19:45,Southampton,Norwich,L Smith,1.91,3.5,3.75,1.95,3.6,3.5,1.86,3.75,3.85,1.95,3.7,3.6,1.95,3.75,3.5,,,,,,,,,,1.96,3.8,3.9,1.93,3.64,3.63,2.04,3.9,3.9,1.73,2.1,,,1.73,2.18,1.69,2.1,1.79,2.18,-0.5,1.98,1.88,,,1.98,1.89,1.89,1.84,2.04,1.95,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,AFC Wimbledon,Leyton Orient,A Dale,2.25,3.25,3.0,2.3,3.3,3.0,2.28,3.25,3.05,2.3,3.25,3.1,2.3,3.3,3.0,,,,,,,,,,2.3,3.5,3.25,2.27,3.26,3.05,2.4,3.35,3.35,2.1,1.7,,,2.1,1.75,2.05,1.7,2.2,1.79,-0.25,2.03,1.83,,,2.03,1.83,1.95,1.77,2.04,1.92,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Blackpool,Port Vale,W Finnie,1.86,3.3,4.1,1.95,3.4,3.75,1.97,3.3,3.7,1.9,3.4,4.1,1.95,3.3,3.8,,,,,,,,,,1.97,3.5,4.1,1.91,3.37,3.92,1.97,3.6,4.3,2.15,1.67,,,2.15,1.76,2.08,1.69,2.2,1.79,-0.5,1.93,1.93,,,1.93,1.93,1.87,1.83,1.97,2.0,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Bolton,Doncaster,D Drysdale,1.56,4.1,4.75,1.6,4.0,5.0,1.65,4.0,4.6,1.6,4.2,5.0,1.63,4.1,4.75,,,,,,,,,,1.65,4.4,5.1,1.61,4.1,4.83,1.65,4.4,5.5,1.6,2.3,,,1.62,2.38,1.58,2.27,1.64,2.46,-1.0,2.05,1.8,,,2.15,1.8,2.03,1.71,2.08,1.86,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Bradford,Mansfield,J Oldham,1.86,3.5,3.9,1.91,3.5,3.75,1.93,3.5,3.65,1.87,3.5,4.1,1.9,3.4,3.9,,,,,,,,,,1.93,3.6,4.2,1.89,3.47,3.87,1.94,3.7,4.3,2.03,1.83,,,2.03,1.86,1.93,1.8,2.08,1.9,-0.5,1.93,1.93,,,1.93,1.93,1.85,1.85,1.94,2.04,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Burton,Reading,E Duckworth,2.2,3.25,3.2,2.2,3.4,3.1,2.28,3.35,2.95,2.25,3.4,3.1,2.15,3.4,3.1,,,,,,,,,,2.28,3.5,3.2,2.21,3.36,3.08,2.4,3.45,3.3,1.98,1.88,,,1.98,1.91,1.89,1.84,2.02,1.95,-0.25,1.98,1.88,,,2.16,1.88,1.96,1.77,2.05,1.92,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Cardiff,Wycombe,C Brook,1.62,3.8,4.75,1.67,4.0,4.6,1.74,3.75,4.25,1.65,4.0,4.8,1.68,3.9,4.6,,,,,,,,,,1.74,4.2,5.0,1.66,3.92,4.61,1.72,4.1,5.3,1.7,2.1,,,1.73,2.1,1.69,2.08,1.75,2.26,-0.75,1.85,2.0,,,2.08,2.0,1.88,1.85,1.91,2.05,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Huddersfield,Lincoln,B Speedie,3.0,3.1,2.35,3.0,3.1,2.38,2.88,3.25,2.38,3.1,3.1,2.38,2.95,3.25,2.35,,,,,,,,,,3.1,3.3,2.42,3.0,3.15,2.36,3.25,3.35,2.5,2.35,1.57,,,2.38,1.64,2.26,1.58,2.48,1.65,0.25,1.78,2.1,,,1.8,2.2,1.7,2.07,1.86,2.12,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Luton,Exeter,A Chilowicz,1.57,3.8,5.0,1.62,4.0,5.0,1.7,3.7,4.7,1.62,3.9,5.25,1.65,3.9,4.8,,,,,,,,,,1.7,4.0,5.25,1.64,3.83,4.93,1.69,4.0,5.7,1.85,2.0,,,1.85,2.0,1.8,1.94,1.9,2.06,-0.75,1.83,2.03,,,1.93,2.03,1.83,1.89,1.87,2.08,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Peterboro,Rotherham,M Russell,2.35,3.6,2.55,2.4,3.6,2.6,2.43,3.5,2.65,2.45,3.6,2.63,2.45,3.7,2.5,,,,,,,,,,2.5,3.75,2.65,2.43,3.6,2.58,2.58,3.8,2.78,1.67,2.15,,,1.67,2.25,1.65,2.14,1.72,2.32,-0.25,2.03,1.83,,,2.1,2.1,1.93,1.82,2.24,1.77,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Plymouth,Stevenage,E Bell,1.91,3.25,3.9,1.91,3.5,3.25,1.97,3.35,3.7,1.95,3.3,4.0,2.0,3.25,3.75,,,,,,,,,,2.0,3.5,4.0,1.96,3.3,3.78,2.08,3.4,4.1,2.25,1.62,,,2.25,1.68,2.2,1.61,2.38,1.7,-0.5,1.98,1.88,,,1.98,1.88,1.92,1.79,2.08,1.89,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,19:45,Stockport,Northampton,S Oldham,1.42,4.33,6.25,1.44,4.33,6.5,1.52,4.1,5.75,1.45,4.4,7.0,1.49,4.4,5.75,,,,,,,,,,1.52,4.5,7.0,1.46,4.31,6.25,1.5,4.8,7.0,1.73,2.08,,,1.75,2.1,1.71,2.04,1.8,2.22,-1.25,2.05,1.8,,,2.06,1.8,2.03,1.71,2.12,1.81,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E2,17/03/2026,20:00,Barnsley,Wigan,R Martin,2.2,3.2,3.1,2.2,3.5,3.0,2.08,3.45,3.25,2.25,3.5,3.0,2.3,3.4,2.95,,,,,,,,,,2.3,3.7,3.25,2.21,3.42,3.05,2.36,3.55,3.25,1.85,1.95,,,1.85,2.0,1.8,1.93,1.9,2.06,-0.25,1.98,1.88,,,1.98,1.91,1.87,1.84,2.03,1.92,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:30,Newport County,Bromley,P Howard,3.8,3.5,1.86,3.75,3.5,1.9,3.85,3.55,1.85,3.9,3.5,1.91,3.9,3.4,1.9,,,,,,,,,,4.25,3.7,1.91,3.89,3.48,1.88,4.3,3.6,1.99,1.95,1.85,,,1.95,1.92,1.88,1.83,2.08,1.87,0.5,1.93,1.93,,,1.93,1.93,1.86,1.85,2.0,1.99,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Accrington,Notts County,O Yates,4.2,3.25,1.81,4.0,3.3,1.91,3.95,3.25,1.91,4.1,3.4,1.91,3.9,3.3,1.95,,,,,,,,,,4.2,3.5,1.95,4.04,3.31,1.9,4.5,3.55,1.98,2.15,1.67,,,2.2,1.67,2.13,1.64,2.3,1.72,0.5,1.93,1.93,,,1.93,1.93,1.84,1.87,2.0,1.98,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Bristol Rvs,Shrewsbury,D Rock,2.01,3.2,3.6,2.05,3.3,3.5,2.07,3.15,3.5,2.05,3.3,3.6,2.0,3.25,3.7,,,,,,,,,,2.1,3.5,3.8,2.04,3.24,3.58,2.14,3.4,3.9,2.2,1.65,,,2.2,1.73,2.12,1.66,2.24,1.79,-0.5,2.08,1.73,,,2.18,1.8,2.04,1.7,2.14,1.83,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Cheltenham,Crewe,S Copeland,2.65,3.1,2.55,2.75,3.3,2.5,2.55,3.25,2.6,2.7,3.25,2.6,2.7,3.2,2.5,,,,,,,,,,2.8,3.5,2.6,2.69,3.22,2.54,2.86,3.35,2.72,2.1,1.7,,,2.1,1.8,2.02,1.72,2.18,1.81,0.0,1.98,1.88,,,1.98,1.88,1.92,1.81,2.01,1.91,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Chesterfield,Oldham,R Joyce,1.96,3.4,3.6,2.05,3.4,3.4,1.98,3.4,3.45,2.0,3.5,3.6,2.05,3.4,3.4,,,,,,,,,,2.05,3.7,3.6,2.01,3.46,3.45,2.08,3.65,3.85,1.9,1.9,,,1.91,1.91,1.85,1.86,1.95,1.99,-0.5,2.03,1.83,,,2.03,1.83,1.97,1.76,2.1,1.87,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Crawley Town,Barnet,H Wager,3.25,3.8,1.96,3.4,3.6,2.0,3.65,3.45,1.89,3.5,3.6,2.0,3.5,3.6,1.98,,,,,,,,,,3.65,3.8,2.0,3.43,3.61,1.97,3.6,3.85,2.1,1.75,2.05,,,1.8,2.1,1.74,2.0,1.86,2.12,0.5,1.83,2.03,,,1.83,2.03,1.78,1.94,1.87,2.1,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Gillingham,Swindon,E Heaslip,3.0,3.3,2.25,2.88,3.4,2.3,2.88,3.25,2.33,3.0,3.4,2.3,2.95,3.4,2.25,,,,,,,,,,3.0,3.6,2.33,2.91,3.4,2.28,3.15,3.55,2.42,1.98,1.88,,,1.98,1.91,1.88,1.84,1.98,1.98,0.25,1.83,2.03,,,1.83,2.2,1.72,2.03,1.88,2.08,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Grimsby,Fleetwood Town,W Cartmel,1.91,3.5,3.6,1.95,3.5,3.6,1.89,3.4,3.7,1.95,3.5,3.7,1.98,3.4,3.6,,,,,,,,,,1.98,3.7,3.8,1.94,3.45,3.67,2.04,3.65,4.0,1.95,1.85,,,1.95,1.91,1.88,1.84,2.04,1.9,-0.5,2.0,1.85,,,2.0,1.85,1.91,1.8,2.04,1.9,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Salford,Barrow,I Searle,1.45,4.1,7.0,1.44,4.2,6.5,1.54,3.9,5.5,1.45,4.2,7.0,1.49,4.2,6.0,,,,,,,,,,1.54,4.33,7.0,1.47,4.12,6.45,1.53,4.5,7.2,1.9,1.9,,,1.91,1.93,1.83,1.88,1.94,2.02,-1.0,1.83,2.03,,,1.83,2.03,1.77,1.98,1.88,2.02,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Tranmere,Harrogate,K Dowle,2.15,3.25,3.2,2.2,3.3,3.2,2.0,3.35,3.45,2.2,3.4,3.2,2.25,3.25,3.1,,,,,,,,,,2.25,3.4,3.45,2.17,3.3,3.22,2.28,3.4,3.6,2.03,1.83,,,2.03,1.86,1.94,1.79,2.04,1.94,-0.25,1.93,1.93,,,1.93,1.93,1.87,1.85,1.96,2.0,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,17/03/2026,19:45,Walsall,Cambridge,S Mather,3.8,3.1,2.05,3.6,3.1,2.1,3.25,3.0,2.18,3.8,3.1,2.1,3.7,3.0,2.1,,,,,,,,,,3.8,3.2,2.18,3.6,3.08,2.11,4.0,3.25,2.18,2.6,1.48,,,2.62,1.53,2.51,1.48,2.76,1.53,0.25,2.05,1.8,,,2.05,1.8,1.98,1.76,2.11,1.86,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+E3,18/03/2026,20:00,Milton Keynes Dons,Colchester,S Mulhall,1.66,3.6,4.75,1.73,3.5,5.0,1.77,3.5,4.25,1.7,3.6,5.0,1.71,3.6,4.8,,,,,,,,,,1.77,3.75,5.0,1.7,3.57,4.79,1.74,3.9,5.7,2.08,1.73,,,2.12,1.78,2.04,1.7,2.22,1.72,-0.75,1.93,1.93,,,1.93,1.93,1.87,1.86,1.98,1.96,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+EC,17/03/2026,19:45,Brackley Town,Solihull,C Walchester,2.55,3.4,2.45,2.5,3.4,2.4,2.6,3.2,2.4,2.55,3.4,2.45,2.55,3.3,2.3,,,,,,,,,,2.66,3.4,2.5,2.55,3.34,2.43,2.7,3.6,2.56,1.75,2.05,,,1.75,2.1,1.73,1.98,1.85,2.12,0.0,1.95,1.85,,,1.95,1.85,1.89,1.8,2.0,1.9,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+EC,17/03/2026,19:45,Gateshead,Wealdstone,A Jackson,2.45,3.6,2.45,2.5,3.75,2.3,2.43,3.5,2.4,2.45,3.7,2.4,2.4,3.6,2.3,,,,,,,,,,2.5,3.75,2.5,2.44,3.62,2.4,2.6,3.9,2.54,1.5,2.5,,,1.5,2.75,1.46,2.54,1.53,2.72,0.0,1.9,1.9,,,1.9,1.9,1.85,1.84,1.98,1.93,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+EC,17/03/2026,19:45,Southend,Rochdale,C Breakspear,2.6,3.1,2.6,2.6,3.1,2.5,2.7,3.1,2.35,2.6,3.1,2.6,2.5,3.0,2.5,,,,,,,,,,2.7,3.1,2.71,2.6,3.08,2.54,2.82,3.2,2.76,2.15,1.67,,,2.15,1.68,2.09,1.65,2.2,1.79,0.0,1.9,1.9,,,1.9,1.9,1.85,1.85,1.95,1.93,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+EC,17/03/2026,19:45,Truro,Scunthorpe,A Quelch,3.6,3.5,1.9,3.4,3.5,1.91,3.95,3.55,1.73,3.5,3.5,1.93,3.4,3.4,1.87,,,,,,,,,,3.95,3.55,1.96,3.55,3.48,1.88,3.7,3.65,2.04,1.73,2.08,,,1.77,2.1,1.7,2.01,1.8,2.12,0.5,1.85,1.95,,,1.85,1.95,1.79,1.89,1.88,2.04,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+EC,17/03/2026,19:45,Woking,Yeovil,R Watkins,1.91,3.4,3.5,1.91,3.25,3.6,1.96,3.2,3.5,1.93,3.3,3.7,1.93,3.2,3.5,,,,,,,,,,2.0,3.4,3.8,1.93,3.27,3.6,2.04,3.35,4.0,2.05,1.75,,,2.05,1.75,1.99,1.72,2.16,1.79,-0.5,1.95,1.85,,,1.95,1.85,1.89,1.79,2.04,1.89,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,18:00,Palermo,Juve Stabia,,1.62,3.3,6.0,1.67,3.5,5.0,1.65,3.6,5.0,1.65,3.4,5.25,1.68,3.6,5.0,,,,,,,,,,1.68,3.6,6.0,1.64,3.48,5.25,1.74,3.6,6.2,2.2,1.65,,,2.2,1.72,2.13,1.65,2.34,1.73,-0.75,1.85,1.95,,,1.9,1.95,1.84,1.87,1.97,1.99,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,19:00,Catanzaro,Modena,,2.65,3.25,2.55,2.75,3.1,2.5,2.88,3.15,2.38,2.63,3.2,2.5,2.8,3.1,2.5,,,,,,,,,,2.88,3.25,2.6,2.74,3.11,2.5,2.98,3.25,2.7,2.08,1.73,,,2.1,1.75,2.04,1.7,2.12,1.84,0.0,1.98,1.83,,,1.98,1.83,1.93,1.8,2.08,1.87,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,19:00,Mantova,Cesena,,2.35,3.4,2.82,2.3,3.4,2.88,2.3,3.3,2.85,2.3,3.25,2.8,2.35,3.3,2.9,,,,,,,,,,2.38,3.4,3.0,2.31,3.27,2.87,2.52,3.4,3.1,1.93,1.88,,,1.93,1.91,1.86,1.84,2.02,1.93,-0.25,2.08,1.73,,,2.14,1.8,2.04,1.7,2.15,1.82,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,19:00,Reggiana,Monza,,6.0,3.3,1.66,5.5,3.4,1.67,5.2,3.5,1.65,5.25,3.4,1.65,5.25,3.5,1.68,,,,,,,,,,6.0,3.5,1.68,5.33,3.39,1.66,6.2,3.6,1.75,2.3,1.6,,,2.3,1.64,2.22,1.59,2.4,1.69,0.75,1.9,1.9,,,1.9,2.06,1.8,1.91,1.98,1.99,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,19:00,Spezia,Empoli,,2.65,2.65,3.1,2.6,2.8,3.0,2.5,2.8,2.95,2.45,2.9,2.9,2.55,2.87,2.95,,,,,,,,,,2.65,2.9,3.2,2.54,2.78,2.98,2.7,3.05,3.2,2.25,1.62,,,2.25,1.62,2.2,1.6,2.36,1.72,0.0,1.75,2.05,,,1.8,2.05,1.72,2.02,1.81,2.18,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,17/03/2026,19:00,Venezia,Padova,,1.22,6.25,9.0,1.25,5.5,10.0,1.3,4.9,8.5,1.22,5.5,10.5,1.3,5.5,8.75,,,,,,,,,,1.3,6.25,11.0,1.25,5.42,9.79,1.29,6.2,13.0,1.57,2.35,,,1.6,2.4,1.57,2.28,1.63,2.5,-1.75,1.98,1.83,,,2.1,1.83,1.98,1.75,2.03,1.88,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,18/03/2026,18:00,Frosinone,Bari,,1.42,4.2,7.0,1.44,4.33,6.0,1.48,4.25,5.75,1.44,4.33,6.0,1.5,4.5,5.5,,,,,,,,,,1.5,4.5,7.0,1.45,4.3,5.94,1.49,4.7,7.0,1.57,2.35,,,1.62,2.35,1.58,2.25,1.64,2.38,-1.25,2.0,1.8,,,2.18,1.8,2.02,1.71,2.04,1.85,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,18/03/2026,19:00,Avellino,Sudtirol,,2.82,2.82,2.7,2.8,2.88,2.63,2.65,2.9,2.7,2.7,3.0,2.55,2.8,2.95,2.6,,,,,,,,,,2.9,3.0,2.7,2.76,2.9,2.63,3.0,2.72,2.58,2.3,1.6,,,2.3,1.62,2.24,1.58,2.2,1.58,0.0,1.95,1.85,,,1.95,1.85,1.91,1.8,2.05,1.76,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,18/03/2026,19:00,Carrarese,Sampdoria,,2.35,2.82,3.2,2.5,2.88,3.0,2.5,2.65,3.2,2.45,3.0,2.8,2.5,2.9,3.0,,,,,,,,,,2.5,3.0,3.6,2.46,2.79,3.08,2.52,2.96,3.3,2.25,1.62,,,2.4,1.62,2.28,1.56,2.34,1.65,-0.25,2.05,1.75,,,2.1,1.75,2.04,1.69,2.12,1.8,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+I2,18/03/2026,19:00,Pescara,Virtus Entella,,2.1,3.4,3.3,2.15,3.4,3.1,2.17,3.25,3.15,2.1,3.3,3.2,2.15,3.4,3.1,,,,,,,,,,2.2,3.4,3.33,2.14,3.31,3.16,2.2,3.5,3.7,1.95,1.85,,,1.95,1.85,1.87,1.83,1.96,1.92,-0.25,1.85,1.95,,,1.98,1.95,1.87,1.84,1.88,2.07,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+SC2,17/03/2026,19:45,Inverness C,Peterhead,G Calder,1.3,4.5,8.0,1.3,4.6,8.0,1.3,4.5,8.0,1.29,4.8,9.0,1.3,4.8,7.5,,,,,,,,,,1.33,5.0,9.0,1.29,4.62,7.93,1.35,5.2,10.0,1.67,2.15,,,1.67,2.3,1.6,2.17,1.71,2.26,-1.5,1.95,1.85,,,1.95,1.85,1.87,1.78,1.97,1.83,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+SC3,17/03/2026,19:45,Spartans,Forfar,A Hendry,1.83,3.2,3.8,1.9,3.3,3.6,1.76,3.4,3.8,1.87,3.3,3.8,1.85,3.25,3.6,,,,,,,,,,1.91,3.4,4.0,1.83,3.29,3.72,1.95,3.45,4.2,2.05,1.75,,,2.05,1.93,1.9,1.77,2.12,1.82,-0.5,1.9,1.9,,,1.9,1.9,1.81,1.83,1.95,1.96,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,17/03/2026,17:00,Fenerbahce,Gaziantep,,1.22,6.25,8.0,1.22,5.5,10.0,1.24,6.0,10.0,1.22,5.75,10.0,1.27,5.75,9.75,,,,,,,,,,1.27,6.25,11.0,1.23,5.75,9.67,1.28,6.6,12.5,1.44,2.7,,,1.5,2.7,1.45,2.58,1.51,2.88,-1.75,1.9,1.9,,,1.9,1.9,1.85,1.84,2.02,1.94,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,18/03/2026,13:00,Alanyaspor,Kocaelispor,,1.96,3.25,3.9,2.0,3.1,3.6,2.08,3.2,3.45,1.95,3.2,3.75,2.1,3.2,3.5,,,,,,,,,,2.1,3.25,3.9,1.99,3.18,3.65,2.02,3.2,4.5,2.25,1.62,,,2.25,1.68,2.2,1.61,2.28,1.66,-0.5,2.0,1.8,,,2.0,1.8,1.96,1.74,2.03,1.87,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,18/03/2026,17:00,Buyuksehyr,Antalyaspor,,1.49,4.0,6.0,1.5,4.0,5.5,1.42,4.5,6.75,1.5,4.1,5.5,1.51,4.33,5.5,,,,,,,,,,1.51,4.5,6.75,1.47,4.21,5.74,1.52,4.5,6.2,1.67,2.15,,,1.67,2.25,1.62,2.17,1.72,2.18,-1.0,1.85,1.95,,,1.85,2.12,1.75,1.97,1.87,1.97,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,18/03/2026,17:00,Eyupspor,Trabzonspor,,4.75,3.9,1.61,4.5,4.0,1.6,5.3,4.2,1.54,4.6,3.9,1.62,4.6,4.1,1.63,,,,,,,,,,5.3,4.2,1.66,4.67,3.95,1.61,5.2,4.5,1.67,1.67,2.15,,,1.67,2.2,1.65,2.13,1.71,2.22,0.75,1.98,1.83,,,2.0,1.83,1.94,1.77,2.09,1.84,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,19/03/2026,13:00,Kayserispor,Karagumruk,,1.86,3.6,3.8,1.95,3.4,3.4,1.93,3.5,3.65,1.9,3.4,3.6,1.95,3.5,3.6,,,,,,,,,,1.97,3.6,4.0,1.9,3.42,3.65,1.94,3.5,4.5,1.85,1.95,,,1.85,1.95,1.83,1.87,1.88,1.93,-0.5,1.93,1.88,,,1.95,1.88,1.87,1.81,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,19/03/2026,17:00,Besiktas,Kasimpasa,,1.35,4.75,8.0,1.33,4.6,7.5,1.36,4.8,7.5,1.33,4.75,7.5,1.37,5.0,7.25,,,,,,,,,,1.37,5.0,9.0,1.33,4.79,7.56,1.36,5.9,10.5,1.57,2.35,,,1.57,2.38,1.55,2.31,1.61,2.38,-1.5,2.0,1.8,,,2.0,1.83,1.93,1.77,1.93,1.81,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
+T1,19/03/2026,17:00,Konyaspor,Genclerbirligi,,1.76,3.5,4.5,1.8,3.4,4.0,1.83,3.6,4.0,1.75,3.4,4.33,1.82,3.5,4.2,,,,,,,,,,1.83,3.6,4.5,1.78,3.45,4.16,1.79,3.55,5.5,2.05,1.75,,,2.05,1.89,1.95,1.77,2.04,1.78,-0.5,1.78,2.03,,,1.8,2.03,1.76,1.93,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

src/pages/analisis_ia.py

@@ -0,0 +1,235 @@
+import streamlit as st
+
+from src.utils import get_pred
+
+st.session_state["_current_page"] = "analisis_ia"
+
+ctx = st.session_state.get("_ctx", {})
+if not ctx:
+    st.warning("Selecciona un partido en la página de Inicio.")
+    st.stop()
+
+eq_l      = ctx["eq_l"]
+eq_v      = ctx["eq_v"]
+liga      = ctx["liga"]
+kelly_ok  = ctx["kelly_ok"]
+probs     = ctx["probs"]
+cH, cD, cA = ctx["cH"], ctx["cD"], ctx["cA"]
+veH, veD, veA = ctx["veH"], ctx["veD"], ctx["veA"]
+pH, pD, pA    = ctx["pH"], ctx["pD"], ctx["pA"]
+rf_l, rf_v    = ctx["rf_l"], ctx["rf_v"]
+l_anota, v_recibe = ctx["l_anota"], ctx["v_recibe"]
+v_anota, l_recibe = ctx["v_anota"], ctx["l_recibe"]
+ventaja_pct    = ctx["ventaja_pct"]
+ventaja_equipo = ctx["ventaja_equipo"]
+
+# ── Panel IA ──────────────────────────────────────────────────────────────
+try:
+    import anthropic as _ant
+    _ant_ok = True
+except ImportError:
+    _ant_ok = False
+
+if not _ant_ok:
+    st.warning("Instala `anthropic>=0.40.0` en requirements.txt para activar el análisis IA.")
+    st.stop()
+
+import os as _os
+_api_key = ""
+try:
+    _api_key = st.secrets["ANTHROPIC_API_KEY"]
+except Exception:
+    _api_key = _os.environ.get("ANTHROPIC_API_KEY", "")
+
+if not _api_key:
+    _api_key = st.text_input(
+        "API Key de Anthropic:",
+        type="password",
+        placeholder="sk-ant-...",
+        help="La clave no se almacena. También puedes definirla en `.streamlit/secrets.toml` o como variable de entorno `ANTHROPIC_API_KEY`.",
+    )
+
+if not _api_key:
+    st.caption("Introduce tu API key para activar el análisis IA.")
+    st.stop()
+
+st.markdown("""<div class="fs-ai">
+  <div class="ai-hdr">
+    <div class="ai-ico">&#x1F916;</div>
+    <div>
+      <div class="ai-title">Claude Analysis</div>
+      <div class="ai-sub">claude-opus-4-6 &middot; Análisis cuantitativo de valor de apuesta</div>
+    </div>
+  </div>
+</div>""", unsafe_allow_html=True)
+
+if not st.button("🤖 Analizar con IA", type="primary", key="btn_ia"):
+    st.stop()
+
+# ── Búsqueda de noticias externas ─────────────────────────────────────────
+_noticias = []
+try:
+    from duckduckgo_search import DDGS
+    _queries = [
+        f"{eq_l} {eq_v} lesiones bajas alineación",
+        f"{eq_l} {eq_v} preview predicción",
+        f"{eq_l} últimas noticias forma reciente",
+        f"{eq_v} últimas noticias forma reciente",
+    ]
+    with st.spinner("🔍 Buscando información actualizada..."):
+        with DDGS() as ddgs:
+            for _q in _queries:
+                for _r in ddgs.text(_q, max_results=3, region="es-es"):
+                    _noticias.append(f"[{_r['title']}] {_r['body']}")
+except ImportError:
+    st.caption("⚠️ Instala `duckduckgo-search>=6.0.0` para activar la búsqueda de noticias.")
+except Exception as _e:
+    st.caption(f"⚠️ Búsqueda de noticias no disponible: {_e}")
+
+# ── Datos estadísticos ────────────────────────────────────────────────────
+_rf_disponible = rf_l is not None and rf_v is not None
+_lines = [f"Partido: {eq_l} (Local) vs {eq_v} (Visitante) — Liga: {liga}"]
+
+if kelly_ok:
+    _lines += [
+        f"Cuotas B365: Local={cH:.2f}, Empate={cD:.2f}, Visitante={cA:.2f}",
+        f"Prob. implícitas: Local={1/cH:.1%}, Empate={1/cD:.1%}, Visitante={1/cA:.1%}",
+        f"VE vs cuota: Local={veH:+.2f}%, Empate={veD:+.2f}%, Visitante={veA:+.2f}%",
+    ]
+if probs is not None:
+    _lines.append(
+        f"Prob. estimadas (histórico 40/40/20): Local={probs['H']:.1%}, Empate={probs['D']:.1%}, Visitante={probs['A']:.1%}"
+    )
+
+_lines += [
+    f"Goles esperados — {eq_l} ataca / {eq_v} defiende: {eq_l} anota {l_anota}, {eq_v} recibe {v_recibe} (Δ={v_recibe-l_anota:+.2f})",
+    f"Goles esperados — {eq_v} ataca / {eq_l} defiende: {eq_v} anota {v_anota}, {eq_l} recibe {l_recibe} (Δ={l_recibe-v_anota:+.2f})",
+]
+
+# ── xG si disponible ──────────────────────────────────────────────────────
+import pandas as _pd_ia
+
+def _xg_mean(df, col):
+    if df is None or df.empty or col not in df.columns:
+        return None
+    s = _pd_ia.to_numeric(df[col], errors="coerce").dropna()
+    return round(s.mean(), 2) if not s.empty else None
+
+_dl_ia = ctx.get("dl")
+_dv_ia = ctx.get("dv")
+_xgs_l_ia  = _xg_mean(_dl_ia, "HxG")
+_xgc_l_ia  = _xg_mean(_dl_ia, "AxG")
+_xgs_v_ia  = _xg_mean(_dv_ia, "AxG")
+_xgc_v_ia  = _xg_mean(_dv_ia, "HxG")
+
+if _xgs_l_ia is not None and _xgs_v_ia is not None:
+    _xg_match_l = round((_xgs_l_ia + (_xgc_v_ia or _xgs_l_ia)) / 2, 2)
+    _xg_match_v = round((_xgs_v_ia + (_xgc_l_ia or _xgs_v_ia)) / 2, 2)
+    _xg_total   = round(_xg_match_l + _xg_match_v, 2)
+    _lines += [
+        f"xG por partido — {eq_l}: anota {_xgs_l_ia} xG, recibe {_xgc_l_ia} xGA (rendimiento real vs xG: {round(l_anota - _xgs_l_ia, 2):+.2f})",
+        f"xG por partido — {eq_v}: anota {_xgs_v_ia} xG, recibe {_xgc_v_ia} xGA (rendimiento real vs xG: {round(v_anota - _xgs_v_ia, 2):+.2f})",
+        f"Predicción xG del partido: {eq_l}={_xg_match_l}, {eq_v}={_xg_match_v}, Total={_xg_total} ({'Sobre' if _xg_total>2.5 else 'Bajo'} 2.5 / {'Sobre' if _xg_total>3.5 else 'Bajo'} 3.5)",
+    ]
+
+if rf_l is not None and rf_v is not None:
+    _el = (get_pred(rf_l, "Goles Anotados") + get_pred(rf_v, "Goles Recibidos")) / 2
+    _ev = (get_pred(rf_v, "Goles Anotados") + get_pred(rf_l, "Goles Recibidos")) / 2
+    _et = _el + _ev
+    _el_ht = (get_pred(rf_l, "Goles Anotados (HT)") + get_pred(rf_v, "Goles Recibidos (HT)")) / 2
+    _ev_ht = (get_pred(rf_v, "Goles Anotados (HT)") + get_pred(rf_l, "Goles Recibidos (HT)")) / 2
+    _lines += [
+        f"RF FT: {eq_l}={_el:.2f} goles, {eq_v}={_ev:.2f} goles, Total={_et:.2f} ({'Sobre' if _et>2.5 else 'Bajo'} 2.5 / {'Sobre' if _et>3.5 else 'Bajo'} 3.5)",
+        f"RF HT: {eq_l}={_el_ht:.2f}, {eq_v}={_ev_ht:.2f}, Total 1T={_el_ht+_ev_ht:.2f} goles",
+    ]
+
+_equipo_ventaja = None
+if ventaja_pct > 0 and ventaja_equipo:
+    _equipo_ventaja = eq_l if eq_l in ventaja_equipo else eq_v
+    _lines.append(
+        f"VENTAJA INFORMACIONAL DECLARADA: +{ventaja_pct}% a favor de {_equipo_ventaja}. "
+        "Ajusta su probabilidad estimada al alza en ese porcentaje y recalcula VE para todos los mercados."
+    )
+
+_datos_str   = "\n".join(f"  - {l}" for l in _lines)
+_noticias_str = (
+    "\n".join(f"  · {n}" for n in _noticias[:12])
+    if _noticias
+    else "  (No se encontraron noticias externas en este momento)"
+)
+
+_prompt = f"""Actúa como un apostador profesional de fútbol y analista cuantitativo senior.
+Tu objetivo es identificar mercados con valor esperado positivo usando EXCLUSIVAMENTE los datos proporcionados.
+No inventes, no asumas y no menciones factores que no aparezcan en los datos.
+{"Si hay ventaja informacional declarada, aplícala ajustando la probabilidad del equipo beneficiado ANTES de cualquier cálculo." if _equipo_ventaja else "No hay ventaja informacional declarada; usa las probabilidades estadísticas sin modificar."}
+
+---
+DATOS ESTADÍSTICOS:
+{_datos_str}
+
+NOTICIAS Y CONTEXTO EXTERNO (fuentes públicas, valida su relevancia):
+{_noticias_str}
+---
+
+ANÁLISIS REQUERIDO — responde en Markdown estructurado:
+
+**1. Evaluación del contexto externo**
+Revisa las noticias encontradas. Extrae solo los datos que puedan cambiar el resultado:
+lesiones confirmadas, sanciones, rotaciones declaradas, rachas, condición de campo, viajes.
+Descarta titulares sin sustancia. Si una noticia contradice otra, señálalo.
+Indica cómo cada factor relevante afecta las probabilidades estimadas y en qué dirección.
+
+**2. Goles — Over/Under**
+Calcula λ local y λ visitante usando goles esperados{"y RF" if _rf_disponible else ""}{" y xG" if _xgs_l_ia is not None else ""}.
+{"Si hay datos xG, compara con los goles reales: si un equipo supera consistentemente su xG es clínico; si está por debajo, puede ser que mejore. Usa xG como ancla de predicción y goles reales como indicador de tendencia." if _xgs_l_ia is not None else ""}
+Evalúa líneas 1.5 / 2.5 / 3.5. Para cada línea: ¿la probabilidad estimada supera la implícita de las cuotas?
+Señala la línea con mayor valor y justifica con los números.
+
+**3. Resultado 1X2 y Asian Handicap**
+Compara prob. estimadas vs implícitas.
+{"Ajusta las probabilidades con la ventaja informacional declarada antes de comparar." if _equipo_ventaja else ""}
+Identifica si hay valor en Local, Empate o Visitante.
+Evalúa Draw No Bet si el empate diluye una ventaja clara.
+Indica hándicap asiático recomendado si aplica.
+
+**4. Resultado al Medio Tiempo (HT)**
+{"Usa RF HT para inferir el resultado más probable en primer tiempo." if _rf_disponible else "Infiere del volumen de goles esperados; descarta si los datos son insuficientes."}
+¿Hay sesgo de goles en 1T o 2T que genere valor en el mercado HT?
+
+**5. Córners**
+Usa medianas de córners si están disponibles en los datos.
+Evalúa Over/Under total de córners y hándicap asiático de córners.
+Si los datos son insuficientes, descarta explícitamente esta categoría.
+
+**6. Ambos Marcan (BTTS)**
+Infiere de goles anotados y recibidos por cada equipo.
+¿Ambos equipos tienen tendencia a anotar y recibir? ¿El mercado lo refleja?
+
+**7. Resumen ejecutivo**
+
+| Campo | Detalle |
+|---|---|
+| Apuesta principal | mercado · resultado · tamaño relativo (pequeño/medio/alto) |
+| Apuesta secundaria | mercado · resultado · tamaño relativo |
+| Mercados descartados | categoría — razón breve |
+| Nivel de confianza | bajo / medio / alto — justificación en una línea |
+
+No uses Kelly ni porcentajes de bankroll. Usa tamaños relativos: pequeño (<1u), medio (1–2u), alto (2–3u).
+Sé directo. Cada conclusión debe respaldarse con al menos un número de los datos."""
+
+# ── Stream de respuesta ────────────────────────────────────────────────────
+def _stream_ia():
+    _client = _ant.Anthropic(api_key=_api_key)
+    with _client.messages.stream(
+        model="claude-opus-4-6",
+        max_tokens=6000,
+        messages=[{"role": "user", "content": _prompt}],
+    ) as _s:
+        for _txt in _s.text_stream:
+            yield _txt
+
+if _noticias:
+    st.caption(f"🔍 {len(_noticias)} fragmentos de noticias incorporados al análisis")
+
+with st.container():
+    st.write_stream(_stream_ia())

src/pages/comparacion.py

@@ -0,0 +1,37 @@
+import streamlit as st
+
+from src.config import PARES_ATAQUE_LOCAL, PARES_ATAQUE_VISITANTE
+from src.statistics import construir_comparacion, mostrar_comparacion
+from src.charts import grafico_radar
+from src.utils import sec
+
+st.session_state["_current_page"] = "comparacion"
+
+ctx = st.session_state.get("_ctx", {})
+if not ctx:
+    st.warning("Selecciona un partido en la página de Inicio.")
+    st.stop()
+
+eq_l = ctx["eq_l"]
+eq_v = ctx["eq_v"]
+dl   = ctx["dl"]
+dv   = ctx["dv"]
+
+if dl.empty or dv.empty:
+    st.markdown('<div class="fs-empty"><div class="em-icon">📊</div>Datos insuficientes para la comparación.</div>', unsafe_allow_html=True)
+    st.stop()
+
+# ── Radar ─────────────────────────────────────────────────────────────────
+sec("📡", "Radar Estadístico")
+rc1, rc2, rc3 = st.columns(3)
+for col, agg in zip([rc1, rc2, rc3], ["mean", "median", "mode"]):
+    fig_r = grafico_radar(dl, dv, eq_l, eq_v, agg=agg)
+    if fig_r:
+        col.plotly_chart(fig_r, use_container_width=True)
+
+# ── Comparaciones ─────────────────────────────────────────────────────────
+sec("🏠🗡️", f"Ataque {eq_l} vs Defensa {eq_v}")
+mostrar_comparacion(construir_comparacion(eq_l, eq_v, dl, dv, PARES_ATAQUE_LOCAL, "ataque_local"))
+
+sec("✈️🗡️", f"Ataque {eq_v} vs Defensa {eq_l}")
+mostrar_comparacion(construir_comparacion(eq_l, eq_v, dl, dv, PARES_ATAQUE_VISITANTE, "ataque_visitante"))

src/pages/cuotas_kelly.py

@@ -0,0 +1,226 @@
+import streamlit as st
+import pandas as pd
+import plotly.graph_objects as go
+
+from src.config import (
+    COLOR_LOCAL, COLOR_VISITANTE,
+    KELLY_MIN_RECOMENDADO, BANK_MIN, BANK_DEFAULT, BANK_STEP,
+)
+from src.charts import _THEME as _CHART_THEME
+from src.utils import sec
+
+st.session_state["_current_page"] = "cuotas_kelly"
+
+ctx = st.session_state.get("_ctx", {})
+if not ctx:
+    st.warning("Selecciona un partido en la página de Inicio.")
+    st.stop()
+
+eq_l      = ctx["eq_l"]
+eq_v      = ctx["eq_v"]
+kelly_ok  = ctx["kelly_ok"]
+probs     = ctx["probs"]
+cH, cD, cA = ctx["cH"], ctx["cD"], ctx["cA"]
+kH, kD, kA = ctx["kH"], ctx["kD"], ctx["kA"]
+veH, veD, veA = ctx["veH"], ctx["veD"], ctx["veA"]
+pH, pD, pA    = ctx["pH"], ctx["pD"], ctx["pA"]
+
+if not kelly_ok:
+    st.markdown('<div class="fs-empty"><div class="em-icon">💰</div>Datos insuficientes para calcular Kelly (cuotas o probabilidades no disponibles).</div>', unsafe_allow_html=True)
+    st.stop()
+
+# ── Barra de probabilidades ───────────────────────────────────────────────
+sec("📊", "Probabilidades Estimadas vs Implícitas")
+_inv = 1/cH + 1/cD + 1/cA
+piH_n = 1/cH / _inv * 100
+piD_n = 1/cD / _inv * 100
+piA_n = 1/cA / _inv * 100
+st.markdown(f"""<div class="fs-prob">
+  <div class="labels">
+    <span>&#x1F3E0; {eq_l}</span>
+    <span>Empate</span>
+    <span>&#x2708;&#xFE0F; {eq_v}</span>
+  </div>
+  <div class="bar">
+    <div class="seg h" style="width:{piH_n:.1f}%"></div>
+    <div class="seg d" style="width:{piD_n:.1f}%"></div>
+    <div class="seg a" style="width:{piA_n:.1f}%"></div>
+  </div>
+  <div class="vals">
+    <span class="vh">{piH_n:.1f}%</span>
+    <span class="vd">{piD_n:.1f}%</span>
+    <span class="va">{piA_n:.1f}%</span>
+  </div>
+</div>""", unsafe_allow_html=True)
+
+# ── Tarjetas de cuotas ────────────────────────────────────────────────────
+def _odds_card(label, odd, prob_est, ve, best):
+    cls = "best" if best else ""
+    ve_cls = "pos" if ve > 0 else "neg"
+    ve_sign = "+" if ve > 0 else ""
+    return f"""<div class="oc {cls}">
+  <div class="on">{label}</div>
+  <div class="ov">{odd:.2f}</div>
+  <div class="op">{prob_est:.1f}% estimado · {1/odd*100:.1f}% implícito</div>
+  <div class="ove {ve_cls}">VE: {ve_sign}{ve:.1f}%</div>
+</div>"""
+
+st.markdown(f"""<div class="fs-odds">
+  {_odds_card(f"1 &mdash; {eq_l}", cH, pH, veH, probs["H"] > 1/cH)}
+  {_odds_card("X &mdash; Empate", cD, pD, veD, probs["D"] > 1/cD)}
+  {_odds_card(f"2 &mdash; {eq_v}", cA, pA, veA, probs["A"] > 1/cA)}
+</div>""", unsafe_allow_html=True)
+
+# ── Gráfico prob estimada vs implícita ────────────────────────────────────
+piH, piD, piA = 1/cH*100, 1/cD*100, 1/cA*100
+fig_p = go.Figure()
+fig_p.add_trace(go.Bar(
+    x=["Local", "Empate", "Visitante"],
+    y=[pH, pD, pA],
+    marker_color=[COLOR_LOCAL, "#9BA3AE", COLOR_VISITANTE],
+    text=[f"{v:.1f}%" for v in [pH, pD, pA]],
+    textposition="outside", textfont=dict(size=14), name="Estimada",
+))
+fig_p.add_trace(go.Scatter(
+    x=["Local", "Empate", "Visitante"], y=[piH, piD, piA],
+    mode="markers+text", name="Implícita (cuota)",
+    marker=dict(size=14, color="#B07D0E", symbol="diamond"),
+    text=[f"{v:.1f}%" for v in [piH, piD, piA]],
+    textposition="top center",
+))
+fig_p.update_layout(
+    height=340, title=dict(text="Prob. Estimada vs Implícita", x=0.5),
+    yaxis_title="%",
+    yaxis=dict(range=[0, max(pH, pD, pA) * 1.45]),
+    margin=dict(l=20, r=20, t=55, b=40),
+    legend=dict(orientation="h", yanchor="bottom", y=-0.25, xanchor="center", x=0.5),
+    **_CHART_THEME,
+)
+st.plotly_chart(fig_p, use_container_width=True)
+
+# ── Kelly Criterion ───────────────────────────────────────────────────────
+sec("🎯", "Kelly Criterion")
+
+def _kelly_cls(k):
+    if k > KELLY_MIN_RECOMENDADO:
+        return "positive"
+    if 0 < k <= KELLY_MIN_RECOMENDADO:
+        return "yellow"
+    return "negative"
+
+def _kelly_card_html(result, team, odd, k, prob, ve):
+    cls = _kelly_cls(k)
+    k_sign = "+" if k > 0 else ""
+    ve_sign = "+" if ve > 0 else ""
+    return f"""<div class="fs-kelly-card {cls}">
+  <div class="kc-result">{result}</div>
+  <div class="kc-team">{team}</div>
+  <div class="kc-odd">{odd:.2f}</div>
+  <div class="kc-sep"></div>
+  <div class="kc-pct">{k_sign}{k:.1f}%</div>
+  <div class="kc-label">Kelly recomendado</div>
+  <div class="kc-meta">{prob:.1f}% prob &nbsp;·&nbsp; VE: {ve_sign}{ve:.1f}%</div>
+</div>"""
+
+st.markdown(f"""<div class="fs-kelly-grid">
+  {_kelly_card_html("1 — Local", eq_l, cH, kH, pH, veH)}
+  {_kelly_card_html("X — Empate", "Empate", cD, kD, pD, veD)}
+  {_kelly_card_html("2 — Visitante", eq_v, cA, kA, pA, veA)}
+</div>""", unsafe_allow_html=True)
+
+# ── Recomendación ─────────────────────────────────────────────────────────
+df_k = pd.DataFrame([
+    {"Resultado": "Local",     "Prob. Estimada": f"{probs['H']:.1%}", "Prob. Implícita": f"{1/cH:.1%}",
+     "Cuota B365": f"{cH:.2f}", "Kelly (%)": round(kH, 2), "Valor Esperado": round(veH, 2)},
+    {"Resultado": "Empate",    "Prob. Estimada": f"{probs['D']:.1%}", "Prob. Implícita": f"{1/cD:.1%}",
+     "Cuota B365": f"{cD:.2f}", "Kelly (%)": round(kD, 2), "Valor Esperado": round(veD, 2)},
+    {"Resultado": "Visitante", "Prob. Estimada": f"{probs['A']:.1%}", "Prob. Implícita": f"{1/cA:.1%}",
+     "Cuota B365": f"{cA:.2f}", "Kelly (%)": round(kA, 2), "Valor Esperado": round(veA, 2)},
+])
+mejor = df_k.loc[df_k["Kelly (%)"].idxmax()]
+if mejor["Kelly (%)"] > 0:
+    st.markdown(f"""<div class="fs-recommend">
+  <div class="rec-icon">&#x2705;</div>
+  <div>
+    <div class="rec-title">Mejor apuesta: {mejor['Resultado']}</div>
+    <div class="rec-detail">
+      {mejor['Kelly (%)']:.2f}% del bankroll &nbsp;·&nbsp; Cuota {mejor['Cuota B365']} &nbsp;·&nbsp; VE: {mejor['Valor Esperado']:+.2f}%
+    </div>
+  </div>
+</div>""", unsafe_allow_html=True)
+else:
+    st.markdown("""<div class="fs-no-bet">
+  <div class="nb-icon">&#x26D4;</div>
+  <div class="nb-text">Kelly no recomienda apostar en ningún resultado.</div>
+</div>""", unsafe_allow_html=True)
+
+# ── Tabla detallada ───────────────────────────────────────────────────────
+sec("📋", "Tabla Detallada")
+
+def _color_kelly(v) -> str:
+    try:
+        n = float(v)
+        if n > KELLY_MIN_RECOMENDADO:
+            return "background-color:#0D9E6E;color:white;font-weight:bold"
+        elif n > 0:
+            return "background-color:#4CAF7D;color:white"
+        else:
+            return "background-color:#D93025;color:white"
+    except Exception:
+        return ""
+
+def _color_ve(v) -> str:
+    try:
+        return (
+            "background-color:#0D6E4E;color:white;font-weight:bold"
+            if float(v) > 0
+            else "background-color:#B71C1C;color:white"
+        )
+    except Exception:
+        return ""
+
+st.dataframe(
+    df_k.style
+    .map(_color_kelly, subset=["Kelly (%)"])
+    .map(_color_ve, subset=["Valor Esperado"])
+    .format({"Kelly (%)": "{:.2f}%", "Valor Esperado": "{:+.2f}%"}),
+    use_container_width=True, hide_index=True,
+)
+
+with st.expander("Ver desglose de probabilidades (40/40/20)"):
+    cp1, cp2, cp3 = st.columns(3)
+    with cp1:
+        st.markdown("**Fuente**")
+        st.markdown(f"🏠 {eq_l} en casa ({probs['nh']} p.)")
+        st.markdown(f"✈️ {eq_v} fuera ({probs['na']} p.)")
+        st.markdown(f"📊 Liga ({probs['nl']} p.)")
+        st.markdown("**🎯 Final**")
+    with cp2:
+        st.markdown("**L / E / V**")
+        st.markdown(f"{probs['ph']['H']:.1%} / {probs['ph']['D']:.1%} / {probs['ph']['A']:.1%}")
+        st.markdown(f"{probs['pa']['H']:.1%} / {probs['pa']['D']:.1%} / {probs['pa']['A']:.1%}")
+        st.markdown(f"{probs['pl']['H']:.1%} / {probs['pl']['D']:.1%} / {probs['pl']['A']:.1%}")
+        st.markdown(f"**{probs['H']:.1%} / {probs['D']:.1%} / {probs['A']:.1%}**")
+    with cp3:
+        st.markdown("**Peso**")
+        st.markdown("40%")
+        st.markdown("40%")
+        st.markdown("20%")
+        st.markdown("**100%**")
+
+# ── Simulador de bankroll ─────────────────────────────────────────────────
+sec("🏦", "Simulador de Bankroll")
+bank = st.number_input("Bankroll ($):", min_value=BANK_MIN, value=BANK_DEFAULT, step=BANK_STEP)
+sim = []
+for _, r in df_k.iterrows():
+    k_pct = max(r["Kelly (%)"], 0)
+    apuesta = bank * k_pct / 100
+    sim.append({
+        "Resultado": r["Resultado"],
+        "Kelly %": f"{k_pct:.2f}%",
+        "Apuesta ($)": f"${apuesta:.2f}",
+        "Ganancia potencial ($)": f"${apuesta * (float(r['Cuota B365']) - 1):.2f}",
+        "Cuota": r["Cuota B365"],
+    })
+st.dataframe(pd.DataFrame(sim), use_container_width=True, hide_index=True)
+st.caption("⚠️ Herramienta matemática. No garantiza ganancias. Apuesta responsablemente.")

src/pages/estadisticas.py

@@ -0,0 +1,444 @@
+import streamlit as st
+import pandas as pd
+
+from src.config import KELLY_MIN_RECOMENDADO
+from src.data_loader import ultimos_n
+from src.statistics import mostrar_tabla_partidos
+from src.utils import sec, med
+
+st.session_state["_current_page"] = "estadisticas"
+
+ctx = st.session_state.get("_ctx", {})
+if not ctx:
+    st.warning("Selecciona un partido en la página de Inicio.")
+    st.stop()
+
+eq_l     = ctx["eq_l"]
+eq_v     = ctx["eq_v"]
+liga     = ctx["liga"]
+df_hist  = ctx["df_hist"]
+kelly_ok = ctx["kelly_ok"]
+probs    = ctx["probs"]
+cH, cD, cA    = ctx["cH"],  ctx["cD"],  ctx["cA"]
+kH, kD, kA    = ctx["kH"],  ctx["kD"],  ctx["kA"]
+veH, veD, veA = ctx["veH"], ctx["veD"], ctx["veA"]
+pH, pD, pA    = ctx["pH"],  ctx["pD"],  ctx["pA"]
+
+# ── Filtro local de partidos (recomputa dl/dv al instante) ───────────────────
+_dl_all = df_hist[df_hist["HomeTeam"] == eq_l]
+_dv_all = df_hist[df_hist["AwayTeam"] == eq_v]
+
+_n_max  = min(len(_dl_all), len(_dv_all), 20)
+_opts   = [0] + list(range(3, _n_max + 1))
+_stored = st.session_state.get("n_both")
+_n_idx  = (
+    _opts.index(_stored) if _stored in _opts else
+    (_opts.index(5)      if 5      in _opts else
+     (1 if len(_opts) > 1 else 0))
+)
+
+_fc, _ = st.columns([3, 7])
+with _fc:
+    _n = st.selectbox(
+        f"📊 Partidos a analizar  ·  🏠 {len(_dl_all)} · ✈️ {len(_dv_all)} disponibles",
+        _opts,
+        index=_n_idx,
+        format_func=lambda x: "Todos los disponibles" if x == 0 else f"Últimos {x}",
+        key="n_both",
+    )
+
+dl = ultimos_n(_dl_all, _n)
+dv = ultimos_n(_dv_all, _n)
+el = ev = "Todos" if _n == 0 else f"Últimos {_n}"
+
+l_anota  = med(dl, "FTHG") if not dl.empty else 0.0
+v_recibe = med(dv, "FTHG") if not dv.empty else 0.0
+v_anota  = med(dv, "FTAG") if not dv.empty else 0.0
+l_recibe = med(dl, "FTAG") if not dl.empty else 0.0
+
+# ════════════════════════════════════════════════════════════════════════════
+# RESUMEN RÁPIDO — barra de probs + recomendación
+# ════════════════════════════════════════════════════════════════════════════
+if kelly_ok:
+    _inv = 1/cH + 1/cD + 1/cA
+    piH_n = 1/cH / _inv * 100
+    piD_n = 1/cD / _inv * 100
+    piA_n = 1/cA / _inv * 100
+    st.markdown(f"""<div class="fs-prob" style="margin-bottom:8px">
+  <div class="labels">
+    <span>&#x1F3E0; {eq_l} &nbsp;<strong>{piH_n:.1f}%</strong> &nbsp;·&nbsp; Cuota {cH:.2f} &nbsp;<small style="opacity:.65">(hist. {pH:.1f}%)</small></span>
+    <span>Empate &nbsp;<strong>{piD_n:.1f}%</strong> &nbsp;·&nbsp; Cuota {cD:.2f} &nbsp;<small style="opacity:.65">(hist. {pD:.1f}%)</small></span>
+    <span>&#x2708;&#xFE0F; {eq_v} &nbsp;<strong>{piA_n:.1f}%</strong> &nbsp;·&nbsp; Cuota {cA:.2f} &nbsp;<small style="opacity:.65">(hist. {pA:.1f}%)</small></span>
+  </div>
+  <div class="bar">
+    <div class="seg h" style="width:{piH_n:.1f}%"></div>
+    <div class="seg d" style="width:{piD_n:.1f}%"></div>
+    <div class="seg a" style="width:{piA_n:.1f}%"></div>
+  </div>
+</div>""", unsafe_allow_html=True)
+
+    _best = max(
+        [("Local",      eq_l,    cH, kH, veH),
+         ("Empate",     "Empate", cD, kD, veD),
+         ("Visitante",  eq_v,    cA, kA, veA)],
+        key=lambda x: x[3],
+    )
+    if _best[3] > 0:
+        st.markdown(f"""<div class="fs-recommend" style="margin-bottom:16px">
+  <div class="rec-icon">&#x2705;</div>
+  <div>
+    <div class="rec-title">Mejor apuesta: {_best[0]} — {_best[1]}</div>
+    <div class="rec-detail">
+      Kelly {_best[3]:.2f}% del bankroll &nbsp;·&nbsp; Cuota {_best[2]:.2f} &nbsp;·&nbsp; VE: {_best[4]:+.2f}%
+    </div>
+  </div>
+</div>""", unsafe_allow_html=True)
+    else:
+        st.markdown("""<div class="fs-no-bet" style="margin-bottom:16px">
+  <div class="nb-icon">&#x26D4;</div>
+  <div class="nb-text">Kelly no recomienda apostar en ningún resultado.</div>
+</div>""", unsafe_allow_html=True)
+
+# ════════════════════════════════════════════════════════════════════════════
+# HELPERS
+# ════════════════════════════════════════════════════════════════════════════
+
+# (col_l = column in dl, col_v = column in dv, invert = True when less is better)
+_CATS = [
+    ("⚽ Goles FT",         "FTHG", "FTAG",  False),
+    ("🕐 Goles HT",         "HTHG", "HTAG",  False),
+    ("🛡️ Goles Recibidos",  "FTAG", "FTHG",  True),
+    ("🎯 Tiros",            "HS",   "AS",    False),
+    ("🎯 T. a puerta",      "HST",  "AST",   False),
+    ("🚩 Córners",          "HC",   "AC",    False),
+    ("📋 Faltas",           "HF",   "AF",    True),
+]
+
+
+def _pill_color(val: float, median: float, invert: bool) -> str:
+    diff = val - median
+    if invert:
+        diff = -diff
+    if diff > 0.5:  return "#0D9E6E"
+    if diff > -0.5: return "#9BA3AE"
+    return "#D93025"
+
+
+def _result_pills_html(series: pd.Series, invert: bool) -> str:
+    if series.empty:
+        return ""
+    series_median = series.median()
+    pills = "".join(
+        f'<div class="rp" style="background:{_pill_color(v, series_median, invert)}" title="{v}">'
+        f'{int(v) if float(v) == int(v) else f"{v:.1f}"}</div>'
+        for v in reversed(series.tolist())
+    )
+    return f'<div class="fs-result-pills">{pills}</div>'
+
+
+def _stats_panel(df: pd.DataFrame, col: str, team: str, period: str,
+                 invert: bool, is_home: bool) -> None:
+    icon      = "🏠" if is_home else "✈️"
+    panel_cls = "fs-panel-home" if is_home else "fs-panel-away"
+    title_cls = "home"          if is_home else "away"
+
+    st.markdown(
+        f'<div class="{panel_cls}">'
+        f'<div class="fs-panel-title {title_cls}">'
+        f'{icon} {team} — {period}</div></div>',
+        unsafe_allow_html=True,
+    )
+
+    if df.empty or col not in df.columns:
+        st.caption("Sin datos suficientes.")
+        return
+
+    s = pd.to_numeric(df[col], errors="coerce").dropna()
+    if s.empty:
+        st.caption("Columna no disponible en este período.")
+        return
+
+    media   = round(s.mean(), 2)
+    mediana = round(s.median(), 2)
+    moda_s  = s.mode()
+    moda    = round(float(moda_s.iloc[0]), 2) if not moda_s.empty else "—"
+
+    _c1, _c2, _c3 = st.columns(3)
+    _c1.metric("📊 Media",   media)
+    _c2.metric("📍 Mediana", mediana)
+    _c3.metric("🎯 Moda",    moda)
+
+    st.caption(f"Últimos {len(s)} partidos (más reciente primero):")
+    st.markdown(_result_pills_html(s, invert), unsafe_allow_html=True)
+
+
+def _center_compare(df_l: pd.DataFrame, col_l: str,
+                    df_v: pd.DataFrame, col_v: str,
+                    invert: bool) -> None:
+    """Compact comparison block shown between the two panels."""
+    if df_l.empty or col_l not in df_l.columns:
+        return
+    if df_v.empty or col_v not in df_v.columns:
+        return
+
+    s_l = pd.to_numeric(df_l[col_l], errors="coerce").dropna()
+    s_v = pd.to_numeric(df_v[col_v], errors="coerce").dropna()
+    if s_l.empty or s_v.empty:
+        return
+
+    med_l = round(s_l.median(), 2)
+    med_v = round(s_v.median(), 2)
+    diff  = round(med_l - med_v, 2)
+
+    # Who has the advantage?
+    if diff == 0:
+        adv_label, adv_cls = "Igualados", "neu"
+    elif (diff > 0 and not invert) or (diff < 0 and invert):
+        adv_label, adv_cls = f"↑ {eq_l}", "pos"
+    else:
+        adv_label, adv_cls = f"↑ {eq_v}", "neg"
+
+    sign = "+" if diff >= 0 else ""
+
+    st.markdown(
+        f'<div class="fs-compare-cell">'
+        f'<div class="cc-lbl">Mediana</div>'
+        f'<div class="cc-val">{med_l:.2f} · {med_v:.2f}</div>'
+        f'<div class="cc-delta {adv_cls}">{sign}{diff} {adv_label}</div>'
+        f'</div>',
+        unsafe_allow_html=True,
+    )
+
+    # Bar comparison
+    max_v = max(med_l, med_v, 0.01)
+    pct_l = med_l / max_v * 100
+    pct_v = med_v / max_v * 100
+    st.markdown(
+        f'<div style="margin-bottom:8px">'
+        f'<div style="font-size:10px;color:var(--text-muted);margin-bottom:3px">🏠 {eq_l}</div>'
+        f'<div style="height:8px;border-radius:4px;background:var(--bg-hover);overflow:hidden">'
+        f'<div style="width:{pct_l:.0f}%;height:100%;background:#2570D4;border-radius:4px"></div></div>'
+        f'<div style="font-size:10px;color:var(--text-muted);margin:5px 0 3px">✈️ {eq_v}</div>'
+        f'<div style="height:8px;border-radius:4px;background:var(--bg-hover);overflow:hidden">'
+        f'<div style="width:{pct_v:.0f}%;height:100%;background:#D93025;border-radius:4px"></div></div>'
+        f'</div>',
+        unsafe_allow_html=True,
+    )
+
+    # Mean comparison
+    mean_l = round(s_l.mean(), 2)
+    mean_v = round(s_v.mean(), 2)
+    st.markdown(
+        f'<div style="font-size:11px;color:var(--text-muted)">'
+        f'Media: <strong>{mean_l}</strong> vs <strong>{mean_v}</strong></div>',
+        unsafe_allow_html=True,
+    )
+
+
+# ════════════════════════════════════════════════════════════════════════════
+# TABS DE CATEGORÍA
+# ════════════════════════════════════════════════════════════════════════════
+tabs = st.tabs([c[0] for c in _CATS])
+
+for tab, (cat_label, col_l, col_v, invert) in zip(tabs, _CATS):
+    with tab:
+        _lc, _cc, _rc = st.columns([5, 3, 5], gap="medium")
+
+        with _lc:
+            _stats_panel(dl, col_l, eq_l, el, invert, is_home=True)
+
+        with _cc:
+            _center_compare(dl, col_l, dv, col_v, invert)
+
+        with _rc:
+            _stats_panel(dv, col_v, eq_v, ev, invert, is_home=False)
+
+        # ── Ataque vs Defensa — solo en Goles FT ──────────────────────────
+        if col_l == "FTHG" and not dl.empty and not dv.empty:
+            sec("⚔️", f"Ataque vs Defensa — Goles esperados en el partido")
+
+            _av1, _av2 = st.columns(2, gap="medium")
+
+            with _av1:
+                st.markdown(f"##### 🏠 {eq_l} ataca · ✈️ {eq_v} defiende")
+                diff_a = round(v_recibe - l_anota, 2)
+                _a1, _a2, _a3 = st.columns(3)
+                _a1.metric(f"⚽ {eq_l} anota",  l_anota)
+                _a2.metric(f"🛡️ {eq_v} recibe", v_recibe)
+                _a3.metric(
+                    "Δ Def vs Ata", f"{diff_a:+.2f}",
+                    delta="Ventaja defensa" if diff_a < 0 else
+                          "Ventaja ataque"  if diff_a > 0 else "Equilibrio",
+                    delta_color="normal",
+                )
+
+            with _av2:
+                st.markdown(f"##### ✈️ {eq_v} ataca · 🏠 {eq_l} defiende")
+                diff_b = round(l_recibe - v_anota, 2)
+                _b1, _b2, _b3 = st.columns(3)
+                _b1.metric(f"⚽ {eq_v} anota",  v_anota)
+                _b2.metric(f"🛡️ {eq_l} recibe", l_recibe)
+                _b3.metric(
+                    "Δ Def vs Ata", f"{diff_b:+.2f}",
+                    delta="Ventaja defensa" if diff_b < 0 else
+                          "Ventaja ataque"  if diff_b > 0 else "Equilibrio",
+                    delta_color="normal",
+                )
+
+# ════════════════════════════════════════════════════════════════════════════
+# xG — GOLES ESPERADOS
+# ════════════════════════════════════════════════════════════════════════════
+def _xg_s(df, col):
+    if df.empty or col not in df.columns:
+        return pd.Series(dtype=float)
+    return pd.to_numeric(df[col], errors="coerce").dropna()
+
+_xgs_l  = _xg_s(dl,  "HxG")   # xG anotados  — local como local
+_xgc_l  = _xg_s(dl,  "AxG")   # xG recibidos — local como local
+_xgs_v  = _xg_s(dv,  "AxG")   # xG anotados  — visitante como visitante
+_xgc_v  = _xg_s(dv,  "HxG")   # xG recibidos — visitante como visitante
+_goals_l = _xg_s(dl, "FTHG")
+_goals_v = _xg_s(dv, "FTAG")
+
+if not (_xgs_l.empty and _xgs_v.empty):
+    sec("📐", "xG — Goles Esperados")
+
+    _xc1, _xc2 = st.columns(2, gap="medium")
+
+    for _col, _xgs, _xgc, _goals, _team, _is_home in [
+        (_xc1, _xgs_l, _xgc_l, _goals_l, eq_l, True),
+        (_xc2, _xgs_v, _xgc_v, _goals_v, eq_v, False),
+    ]:
+        icon = "🏠" if _is_home else "✈️"
+        with _col:
+            st.markdown(f"**{icon} {_team}**")
+            if _xgs.empty:
+                st.caption("Sin datos xG.")
+            else:
+                _xg_avg  = round(_xgs.mean(), 2)
+                _xgc_avg = round(_xgc.mean(), 2) if not _xgc.empty else None
+                _real    = round(_goals.mean(), 2) if not _goals.empty else None
+                _m1, _m2, _m3 = st.columns(3)
+                _m1.metric("xG anotados", _xg_avg)
+                if _xgc_avg is not None:
+                    _m2.metric("xGA recibidos", _xgc_avg)
+                if _real is not None:
+                    _delta = round(_real - _xg_avg, 2)
+                    _lbl = "sobre xG" if _delta > 0 else ("bajo xG" if _delta < 0 else "=xG")
+                    _m3.metric("Goles reales", _real, delta=f"{_delta:+.2f} {_lbl}")
+
+            # Historial pill de xG por partido
+            if not _xgs.empty:
+                st.caption(f"xG por partido (más reciente primero):")
+                st.markdown(_result_pills_html(_xgs, invert=False), unsafe_allow_html=True)
+
+    # ── Predicción xG del partido ─────────────────────────────────────────
+    if not _xgs_l.empty and not _xgs_v.empty and not _xgc_l.empty and not _xgc_v.empty:
+        st.markdown("---")
+        st.markdown("##### Predicción del partido (xG)")
+        _xg_match_l = round((_xgs_l.mean() + _xgc_v.mean()) / 2, 2)
+        _xg_match_v = round((_xgs_v.mean() + _xgc_l.mean()) / 2, 2)
+        _xg_total   = round(_xg_match_l + _xg_match_v, 2)
+
+        _px1, _px2, _px3, _px4, _px5 = st.columns(5)
+        _px1.metric(f"xG {eq_l[:14]}", _xg_match_l)
+        _px2.metric(f"xG {eq_v[:14]}", _xg_match_v)
+        _px3.metric("Total xG", _xg_total)
+        _px4.metric("O/U 2.5", "Sobre" if _xg_total > 2.5 else "Bajo",
+                    delta=f"{_xg_total - 2.5:+.2f}")
+        _px5.metric("O/U 3.5", "Sobre" if _xg_total > 3.5 else "Bajo",
+                    delta=f"{_xg_total - 3.5:+.2f}")
+
+        # Comparar con predicción de goles reales
+        _est_l_real = round((l_anota + v_recibe) / 2, 2)
+        _est_v_real = round((v_anota + l_recibe) / 2, 2)
+        st.caption(
+            f"Predicción por goles reales: {eq_l} {_est_l_real} — {eq_v} {_est_v_real} "
+            f"(total {round(_est_l_real + _est_v_real, 2)}) · "
+            f"Diferencia xG vs reales: {round(_xg_total - (_est_l_real + _est_v_real), 2):+.2f}"
+        )
+
+# ════════════════════════════════════════════════════════════════════════════
+# DATOS DETALLADOS
+# ════════════════════════════════════════════════════════════════════════════
+sec("📋", "Datos Detallados")
+
+_dt1, _dt2 = st.columns(2, gap="medium")
+with _dt1:
+    with st.expander(f"Partidos de {eq_l} como local ({el})"):
+        mostrar_tabla_partidos(
+            dl,
+            cols_extra=["FTHG", "FTAG", "FTR", "HxG", "AxG",
+                        "HTHG", "HTAG", "HS", "HST", "HC", "HF", "HY", "HR",
+                        "B365H", "B365D", "B365A"],
+        )
+with _dt2:
+    with st.expander(f"Partidos de {eq_v} como visitante ({ev})"):
+        mostrar_tabla_partidos(
+            dv,
+            cols_extra=["FTHG", "FTAG", "FTR", "HxG", "AxG",
+                        "HTAG", "HTHG", "AS", "AST", "AC", "AF", "AY", "AR",
+                        "B365H", "B365D", "B365A"],
+        )
+
+with st.expander("🗂️ Histórico completo de la liga"):
+    mostrar_tabla_partidos(
+        df_hist,
+        cols_extra=["FTHG", "FTAG", "FTR", "HS", "AS", "HST", "AST",
+                    "HC", "AC", "HF", "AF", "HY", "AY", "HR", "AR",
+                    "B365H", "B365D", "B365A"],
+    )
+
+# ════════════════════════════════════════════════════════════════════════════
+# DESCARGA — imagen para redes sociales
+# ════════════════════════════════════════════════════════════════════════════
+sec("📤", "Compartir en redes sociales")
+
+st.caption("Genera una imagen 1080×1080 lista para Instagram, Twitter/X o WhatsApp.")
+
+if st.button("🖼️ Generar imagen del resumen", key="btn_gen_img"):
+    import pandas as _pd
+    from src.social_image import generar_post_imagen
+
+    def _mean(df, col):
+        if df.empty or col not in df.columns:
+            return None
+        s = _pd.to_numeric(df[col], errors="coerce").dropna()
+        return round(s.mean(), 2) if not s.empty else None
+
+    def _mean0(df, col):
+        v = _mean(df, col)
+        return v if v is not None else 0.0
+
+    # xG match prediction (None si la liga no tiene datos xG)
+    _xgs_l = _mean(dl, "HxG")
+    _xgc_l = _mean(dl, "AxG")
+    _xgs_v = _mean(dv, "AxG")
+    _xgc_v = _mean(dv, "HxG")
+    if all(v is not None for v in [_xgs_l, _xgc_l, _xgs_v, _xgc_v]):
+        _xg_img_l = round((_xgs_l + _xgc_v) / 2, 2)
+        _xg_img_v = round((_xgs_v + _xgc_l) / 2, 2)
+    else:
+        _xg_img_l = _xg_img_v = None
+
+    with st.spinner("Generando imagen..."):
+        _img_bytes = generar_post_imagen(
+            eq_l=eq_l, eq_v=eq_v, liga=liga,
+            pH=pH, pD=pD, pA=pA,
+            cH=cH, cD=cD, cA=cA,
+            l_anota=_mean0(dl, "FTHG"), v_recibe=_mean0(dv, "FTHG"),
+            v_anota=_mean0(dv, "FTAG"), l_recibe=_mean0(dl, "FTAG"),
+            kelly_ok=kelly_ok,
+            periodo=el,
+            xg_l=_xg_img_l,
+            xg_v=_xg_img_v,
+        )
+    st.image(_img_bytes, use_container_width=False, width=540)
+    _fname = f"{eq_l}_vs_{eq_v}.png".replace(" ", "_")
+    st.download_button(
+        "⬇️ Descargar imagen",
+        data=_img_bytes,
+        file_name=_fname,
+        mime="image/png",
+        type="primary",
+    )

src/pages/inicio.py

@@ -0,0 +1,185 @@
+from datetime import date
+
+import streamlit as st
+
+from src.config import LIGA_INFO
+
+st.session_state["_current_page"] = "inicio"
+
+df_proximos = st.session_state.get("_df_proximos")
+if df_proximos is None:
+    st.error("No se pudo cargar los fixtures.")
+    st.stop()
+
+today = date.today()
+df_up = df_proximos[df_proximos["Date"].dt.date >= today].copy()
+
+sort_cols = ["Date", "Time_local"] if "Time_local" in df_up.columns else ["Date"]
+df_up = df_up.sort_values(sort_cols, na_position="last").reset_index(drop=True)
+
+_sel = st.session_state.get("_selected_match", {}) or {}
+
+
+def _liga_info(div: str) -> dict:
+    return LIGA_INFO.get(div, {"pais": div, "bandera": "🏟️", "liga": div})
+
+
+def _is_selected(row) -> bool:
+    return (
+        bool(_sel)
+        and _sel.get("HomeTeam") == row["HomeTeam"]
+        and _sel.get("AwayTeam") == row["AwayTeam"]
+        and _sel.get("Div") == row["Div"]
+    )
+
+
+def _card(row, key: str) -> None:
+    hora  = row.get("Time_local", "")
+    info  = _liga_info(row["Div"])
+    meta  = f"{hora}" if hora else row["Date"].strftime("%d %b")
+    sel   = _is_selected(row)
+    cls   = "selected" if sel else ""
+    st.markdown(f"""<div class="fs-mc {cls}">
+  <div class="mc-league">{info['liga']}</div>
+  <div class="mc-home">{row['HomeTeam']}</div>
+  <div class="mc-vs">vs</div>
+  <div class="mc-away">{row['AwayTeam']}</div>
+  <div class="mc-meta">{meta}</div>
+</div>""", unsafe_allow_html=True)
+    label    = "✓ Seleccionado" if sel else "Analizar →"
+    btn_type = "primary" if sel else "secondary"
+    if st.button(label, key=key, use_container_width=True, type=btn_type):
+        st.session_state["_selected_match"] = row.to_dict()
+        pg_dest = st.session_state.get("_p_estadisticas", "pages/estadisticas.py")
+        st.switch_page(pg_dest)
+
+
+# ── Fechas disponibles ───────────────────────────────────────────────────────
+_fechas_disp  = sorted(df_up["Date"].dt.date.unique())
+_fecha_map    = {f.strftime("%a %d %b %Y"): f for f in _fechas_disp}
+_fecha_opts   = list(_fecha_map.keys())
+_default_fecha = today if today in _fechas_disp else (_fechas_disp[0] if _fechas_disp else today)
+_def_str      = _default_fecha.strftime("%a %d %b %Y")
+_def_idx      = _fecha_opts.index(_def_str) if _def_str in _fecha_opts else 0
+
+# ── Cabecera: título + búsqueda + fecha ─────────────────────────────────────
+total_ligas    = df_up["Div"].nunique() if "Div" in df_up.columns else "-"
+total_partidos = len(df_up)
+
+_h1, _h2, _h3 = st.columns([4, 3, 2])
+
+with _h1:
+    st.markdown(f"""<div class="fs-header">
+  <div>
+    <h1>Partidos</h1>
+    <div class="sub">&#x1F3C6; {total_ligas} ligas &middot; &#x26BD; {total_partidos} disponibles</div>
+  </div>
+</div>""", unsafe_allow_html=True)
+
+with _h2:
+    _search = st.text_input(
+        "Buscar",
+        placeholder="🔍 Buscar equipo...",
+        key="_search",
+        label_visibility="collapsed",
+    )
+
+with _h3:
+    if _fecha_opts:
+        _fecha_sel_str = st.selectbox(
+            "Fecha",
+            _fecha_opts,
+            index=_def_idx,
+            key="_fecha_filtro",
+            label_visibility="collapsed",
+        )
+        _fecha_sel = _fecha_map.get(_fecha_sel_str, _default_fecha)
+    else:
+        _fecha_sel = _default_fecha
+
+if df_up.empty:
+    st.markdown('<div class="fs-empty"><div class="em-icon">📭</div>No hay partidos próximos disponibles.</div>', unsafe_allow_html=True)
+    st.stop()
+
+# ── Próximos 5 (independiente del filtro de fecha) ───────────────────────────
+_df_next5 = df_up
+if _search:
+    _q = _search.lower().strip()
+    _df_next5 = _df_next5[
+        _df_next5["HomeTeam"].str.lower().str.contains(_q, na=False) |
+        _df_next5["AwayTeam"].str.lower().str.contains(_q, na=False)
+    ]
+
+df_next5   = _df_next5.head(5)
+next5_keys = set(zip(df_next5["HomeTeam"], df_next5["AwayTeam"], df_next5["Div"]))
+
+if not df_next5.empty:
+    st.markdown('<div class="fs-sec"><span>⚡</span> Próximos partidos</div>', unsafe_allow_html=True)
+    cols5 = st.columns(len(df_next5), gap="small")
+    for i, (_, row) in enumerate(df_next5.iterrows()):
+        with cols5[i]:
+            _card(row, f"n5_{i}")
+
+# ── Partidos del día seleccionado ────────────────────────────────────────────
+st.markdown("---")
+_is_today   = _fecha_sel == today
+_fecha_label = "Hoy" if _is_today else _fecha_sel.strftime("%a %d de %B")
+
+_sl, _sr = st.columns([6, 2])
+with _sl:
+    st.markdown(
+        f'<div class="fs-sec"><span>🗓️</span> Partidos · {_fecha_label}</div>',
+        unsafe_allow_html=True,
+    )
+
+df_day = df_up[df_up["Date"].dt.date == _fecha_sel].copy()
+
+if _search:
+    _q = _search.lower().strip()
+    df_day = df_day[
+        df_day["HomeTeam"].str.lower().str.contains(_q, na=False) |
+        df_day["AwayTeam"].str.lower().str.contains(_q, na=False)
+    ]
+
+if df_day.empty:
+    if _search:
+        st.markdown(
+            f'<div class="fs-empty"><div class="em-icon">🔍</div>'
+            f'Sin resultados para "<strong>{_search}</strong>" el {_fecha_label}.</div>',
+            unsafe_allow_html=True,
+        )
+    else:
+        st.markdown(
+            f'<div class="fs-empty"><div class="em-icon">📭</div>'
+            f'No hay partidos el {_fecha_label}. Elige otra fecha.</div>',
+            unsafe_allow_html=True,
+        )
+    st.stop()
+
+# Añadir columnas de país
+df_day["_pais"]    = df_day["Div"].apply(lambda d: _liga_info(d)["pais"])
+df_day["_bandera"] = df_day["Div"].apply(lambda d: _liga_info(d)["bandera"])
+df_day["_liga"]    = df_day["Div"].apply(lambda d: _liga_info(d)["liga"])
+
+sort_time = "Time_local" if "Time_local" in df_day.columns else "Date"
+df_day = df_day.sort_values(["_pais", sort_time], na_position="last")
+
+idx = 100
+for pais, grupo_pais in df_day.groupby("_pais", sort=True):
+    bandera = grupo_pais["_bandera"].iloc[0]
+    ligas   = " · ".join(sorted(grupo_pais["_liga"].unique()))
+    n       = len(grupo_pais)
+
+    st.markdown(
+        f'<div class="fs-date-label">{bandera} {pais} &nbsp;·&nbsp; {ligas}'
+        f' &nbsp;·&nbsp; {n} partido{"s" if n != 1 else ""}</div>',
+        unsafe_allow_html=True,
+    )
+
+    chunks = [grupo_pais.iloc[i:i + 4] for i in range(0, len(grupo_pais), 4)]
+    for chunk in chunks:
+        row_cols = st.columns(4, gap="small")
+        for j, (_, row) in enumerate(chunk.iterrows()):
+            with row_cols[j]:
+                _card(row, f"list_{idx}")
+                idx += 1

src/pages/prediccion_rf.py

@@ -0,0 +1,108 @@
+import streamlit as st
+import plotly.graph_objects as go
+
+from src.config import (
+    COLOR_LOCAL, COLOR_VISITANTE,
+    COLOR_LOCAL_HT, COLOR_VISITANTE_HT,
+    COLOR_MEDIA,
+    RF_N_ROLLING_MIN, RF_N_ROLLING_MAX, RF_N_ROLLING_DEFAULT,
+)
+from src.charts import _THEME as _CHART_THEME
+from src.models import mostrar_rf_equipo
+from src.utils import sec, get_pred
+
+st.session_state["_current_page"] = "prediccion_rf"
+
+ctx = st.session_state.get("_ctx", {})
+if not ctx:
+    st.warning("Selecciona un partido en la página de Inicio.")
+    st.stop()
+
+eq_l = ctx["eq_l"]
+eq_v = ctx["eq_v"]
+rf_l = ctx["rf_l"]
+rf_v = ctx["rf_v"]
+
+# ── Modelos RF por equipo ─────────────────────────────────────────────────
+col_rf1, col_rf2 = st.columns(2, gap="medium")
+
+with col_rf1:
+    st.markdown(f'<div class="fs-panel-home"><div class="fs-panel-title home">&#x1F3E0; {eq_l} — Local</div></div>', unsafe_allow_html=True)
+    if rf_l is not None and not rf_l.empty:
+        mostrar_rf_equipo(rf_l, eq_l, COLOR_LOCAL)
+    else:
+        st.markdown(f'<div class="fs-empty"><div class="em-icon">&#x1F916;</div>Datos insuficientes para {eq_l}.</div>', unsafe_allow_html=True)
+
+with col_rf2:
+    st.markdown(f'<div class="fs-panel-away"><div class="fs-panel-title away">&#x2708;&#xFE0F; {eq_v} — Visitante</div></div>', unsafe_allow_html=True)
+    if rf_v is not None and not rf_v.empty:
+        mostrar_rf_equipo(rf_v, eq_v, COLOR_VISITANTE)
+    else:
+        st.markdown(f'<div class="fs-empty"><div class="em-icon">&#x1F916;</div>Datos insuficientes para {eq_v}.</div>', unsafe_allow_html=True)
+
+# ── Estimación final ──────────────────────────────────────────────────────
+if rf_l is None or rf_v is None:
+    st.stop()
+
+sec("📋", "Estimación Final de Goles")
+
+gl_anota     = get_pred(rf_l, "Goles Anotados")
+gl_recibe    = get_pred(rf_l, "Goles Recibidos")
+gv_anota     = get_pred(rf_v, "Goles Anotados")
+gv_recibe    = get_pred(rf_v, "Goles Recibidos")
+gl_anota_ht  = get_pred(rf_l, "Goles Anotados (HT)")
+gl_recibe_ht = get_pred(rf_l, "Goles Recibidos (HT)")
+gv_anota_ht  = get_pred(rf_v, "Goles Anotados (HT)")
+gv_recibe_ht = get_pred(rf_v, "Goles Recibidos (HT)")
+
+est_local    = (gl_anota + gv_recibe) / 2
+est_visit    = (gv_anota + gl_recibe) / 2
+est_total    = est_local + est_visit
+est_local_ht = (gl_anota_ht + gv_recibe_ht) / 2
+est_visit_ht = (gv_anota_ht + gl_recibe_ht) / 2
+est_total_ht = est_local_ht + est_visit_ht
+
+ce1, ce2, ce3, ce4, ce5, ce6 = st.columns(6)
+ce1.metric(f"🏠 {eq_l} FT", f"{est_local:.2f}")
+ce2.metric(f"✈️ {eq_v} FT", f"{est_visit:.2f}")
+ce3.metric("⚽ Total FT", f"{est_total:.2f}",
+           delta="Más de 2.5" if est_total > 2.5 else "Menos de 2.5",
+           delta_color="normal" if est_total > 2.5 else "inverse")
+ce4.metric(f"🏠 {eq_l} HT", f"{est_local_ht:.2f}")
+ce5.metric(f"✈️ {eq_v} HT", f"{est_visit_ht:.2f}")
+ce6.metric("⚽ Total HT", f"{est_total_ht:.2f}",
+           delta="Más de 0.5" if est_total_ht > 0.5 else "Menos de 0.5",
+           delta_color="normal" if est_total_ht > 0.5 else "inverse")
+
+fig_est = go.Figure()
+fig_est.add_trace(go.Bar(
+    x=[f"{eq_l} FT", f"{eq_l} HT", f"{eq_v} FT", f"{eq_v} HT"],
+    y=[est_local, est_local_ht, est_visit, est_visit_ht],
+    marker_color=[COLOR_LOCAL, COLOR_LOCAL_HT, COLOR_VISITANTE, COLOR_VISITANTE_HT],
+    text=[f"{v:.2f}" for v in [est_local, est_local_ht, est_visit, est_visit_ht]],
+    textposition="outside", textfont=dict(size=14),
+))
+fig_est.add_hline(
+    y=est_total / 2, line_dash="dash", line_color=COLOR_MEDIA,
+    annotation_text=f"Total FT: {est_total:.2f} goles",
+)
+fig_est.update_layout(
+    height=360, title=dict(text="Estimación de Goles (FT y HT)", x=0.5),
+    yaxis_title="Goles",
+    yaxis=dict(range=[0, max(est_local, est_visit, 0.1) * 1.7]),
+    margin=dict(l=20, r=20, t=55, b=40), showlegend=False,
+    **_CHART_THEME,
+)
+st.plotly_chart(fig_est, use_container_width=True)
+
+# ── Ajuste de ventana RF (después de mostrar resultados) ──────────────────
+st.markdown("---")
+_sc, _ = st.columns([3, 7])
+with _sc:
+    st.slider(
+        "🤖 Ventana RF — partidos usados para el modelo:",
+        RF_N_ROLLING_MIN, RF_N_ROLLING_MAX,
+        st.session_state.get("nrf", RF_N_ROLLING_DEFAULT), 1,
+        key="nrf",
+        help="Cambia el número de partidos recientes usados para entrenar el modelo. El análisis se recalculará.",
+    )

src/requirements.txt

@@ -0,0 +1,9 @@
+streamlit>=1.32.0
+pandas>=2.1.0
+numpy>=1.26.0
+plotly>=5.18.0
+scikit-learn>=1.4.0
+anthropic>=0.40.0
+duckduckgo-search>=6.0.0
+kaleido>=0.2.1
+

src/src/betting.py

@@ -0,0 +1,60 @@
+import pandas as pd
+
+from src.config import MIN_PARTIDOS_EQUIPO
+
+
+def calc_probs(df_hist: pd.DataFrame, eq_local: str, eq_visit: str) -> dict | None:
+    """
+    Mezcla ponderada de probabilidades:
+      40% forma local del equipo de casa
+      40% forma visitante del equipo fuera
+      20% media de liga
+    """
+    if "FTR" not in df_hist.columns or len(df_hist) == 0:
+        return None
+
+    t = len(df_hist)
+    pl = {
+        "H": (df_hist["FTR"] == "H").sum() / t,
+        "D": (df_hist["FTR"] == "D").sum() / t,
+        "A": (df_hist["FTR"] == "A").sum() / t,
+    }
+
+    dh = df_hist[df_hist["HomeTeam"] == eq_local]
+    if len(dh) >= MIN_PARTIDOS_EQUIPO:
+        nh = len(dh)
+        ph = {
+            "H": (dh["FTR"] == "H").sum() / nh,
+            "D": (dh["FTR"] == "D").sum() / nh,
+            "A": (dh["FTR"] == "A").sum() / nh,
+        }
+    else:
+        ph = pl.copy()
+
+    da = df_hist[df_hist["AwayTeam"] == eq_visit]
+    if len(da) >= MIN_PARTIDOS_EQUIPO:
+        na = len(da)
+        pa = {
+            "H": (da["FTR"] == "H").sum() / na,
+            "D": (da["FTR"] == "D").sum() / na,
+            "A": (da["FTR"] == "A").sum() / na,
+        }
+    else:
+        pa = pl.copy()
+
+    r = {k: 0.4 * ph[k] + 0.4 * pa[k] + 0.2 * pl[k] for k in ["H", "D", "A"]}
+    s = sum(r.values())
+    if s > 0:
+        r = {k: v / s for k, v in r.items()}
+
+    return {
+        **r,
+        "ph": ph, "pa": pa, "pl": pl,
+        "nh": len(dh), "na": len(da), "nl": t,
+    }
+
+
+def kelly(prob: float, odds: float) -> float:
+    """Kelly criterion: fracción óptima del bankroll a apostar."""
+    b = odds - 1
+    return (b * prob - (1 - prob)) / b * 100 if b > 0 else 0

src/src/charts.py

@@ -0,0 +1,210 @@
+import numpy as np
+import pandas as pd
+import plotly.graph_objects as go
+
+from src.config import (
+    COLOR_LOCAL,
+    COLOR_VISITANTE,
+    COLOR_NEGATIVO,
+    COLOR_EMPATE,
+    COLOR_MEDIA,
+    COLOR_LOCAL_ALPHA,
+    COLOR_VISITANTE_ALPHA,
+    KELLY_GAUGE_MAX,
+    RADAR_CATS,
+    RADAR_COLS_LOCAL,
+    RADAR_COLS_VISIT,
+)
+
+_THEME = dict(
+    paper_bgcolor="rgba(0,0,0,0)",
+    plot_bgcolor="rgba(0,0,0,0)",
+    font=dict(
+        family="-apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif",
+        color="#1A1F2E",
+        size=12,
+    ),
+)
+
+
+def _agg_col(df, col: str, agg: str) -> float:
+    if col not in df.columns:
+        return np.nan
+    s = pd.to_numeric(df[col], errors="coerce").dropna()
+    if s.empty:
+        return np.nan
+    if agg == "median":
+        return s.median()
+    if agg == "mode":
+        m = s.mode()
+        return float(m.iloc[0]) if not m.empty else np.nan
+    return s.mean()  # default: mean
+
+
+def grafico_radar(
+    df_local, df_visit, eq_local: str, eq_visit: str, agg: str = "mean"
+) -> go.Figure | None:
+    titulos = {"mean": "Promedio", "median": "Mediana", "mode": "Moda"}
+    titulo = titulos.get(agg, agg.capitalize())
+
+    vl = [0 if np.isnan(v) else v for v in [_agg_col(df_local, c, agg) for c in RADAR_COLS_LOCAL]]
+    vv = [0 if np.isnan(v) else v for v in [_agg_col(df_visit, c, agg) for c in RADAR_COLS_VISIT]]
+    if all(v == 0 for v in vl + vv):
+        return None
+    mx = [max(a, b, 0.01) for a, b in zip(vl, vv)]
+    nl = [v / m * 100 for v, m in zip(vl, mx)]
+    nv = [v / m * 100 for v, m in zip(vv, mx)]
+    fig = go.Figure()
+    fig.add_trace(go.Scatterpolar(
+        r=nl + [nl[0]], theta=RADAR_CATS + [RADAR_CATS[0]], fill="toself",
+        name=f"{eq_local} (Local)", line=dict(color=COLOR_LOCAL, width=2),
+        fillcolor="rgba(30,136,229,0.15)",
+        customdata=[f"{v:.2f}" for v in vl] + [f"{vl[0]:.2f}"],
+        hovertemplate="%{theta}: %{customdata}<extra></extra>",
+    ))
+    fig.add_trace(go.Scatterpolar(
+        r=nv + [nv[0]], theta=RADAR_CATS + [RADAR_CATS[0]], fill="toself",
+        name=f"{eq_visit} (Visitante)", line=dict(color=COLOR_NEGATIVO, width=2),
+        fillcolor="rgba(217,48,37,0.15)",
+        customdata=[f"{v:.2f}" for v in vv] + [f"{vv[0]:.2f}"],
+        hovertemplate="%{theta}: %{customdata}<extra></extra>",
+    ))
+    fig.update_layout(
+        polar=dict(radialaxis=dict(visible=True, range=[0, 110], showticklabels=False)),
+        showlegend=True, height=380,
+        title=dict(text=titulo, x=0.5, font=dict(size=14)),
+        margin=dict(l=50, r=50, t=55, b=35),
+        legend=dict(orientation="h", yanchor="bottom", y=-0.18, xanchor="center", x=0.5),
+        **_THEME,
+    )
+    return fig
+
+
+
+def grafico_tendencia_goles(df, col_anot: str, col_rec: str, titulo: str) -> go.Figure | None:
+    dc = df.copy().reset_index(drop=True)
+    for c in [col_anot, col_rec]:
+        if c in dc.columns:
+            dc[c] = pd.to_numeric(dc[c], errors="coerce")
+
+    if col_anot not in dc.columns or dc.empty:
+        return None
+    dc = dc.dropna(subset=[col_anot])
+    dc["J"] = range(1, len(dc) + 1)
+    dc["PM_A"] = dc[col_anot].rolling(3, min_periods=1).mean()
+    fig = go.Figure()
+    fig.add_trace(go.Bar(x=dc["J"], y=dc[col_anot], name="Anotados", marker_color=COLOR_LOCAL_ALPHA))
+    if col_rec in dc.columns:
+        dc["PM_R"] = dc[col_rec].rolling(3, min_periods=1).mean()
+        fig.add_trace(go.Bar(x=dc["J"], y=dc[col_rec], name="Recibidos", marker_color=COLOR_VISITANTE_ALPHA))
+    fig.add_trace(go.Scatter(
+        x=dc["J"], y=dc["PM_A"], name="Prom. Anot. (3)",
+        line=dict(color="#1565C0", width=3),
+    ))
+    if "PM_R" in dc.columns:
+        fig.add_trace(go.Scatter(
+            x=dc["J"], y=dc["PM_R"], name="Prom. Rec. (3)",
+            line=dict(color="#C62828", width=3, dash="dash"),
+        ))
+    fig.update_layout(
+        barmode="group", height=320, title=dict(text=titulo, x=0.5),
+        xaxis_title="Jornada", yaxis_title="Goles",
+        margin=dict(l=20, r=20, t=50, b=40),
+        legend=dict(orientation="h", yanchor="bottom", y=-0.3, xanchor="center", x=0.5),
+        **_THEME,
+    )
+    return fig
+
+
+def grafico_dist_goles(df, col: str, titulo: str, color: str) -> go.Figure | None:
+    if col not in df.columns or df.empty:
+        return None
+    s = pd.to_numeric(df[col], errors="coerce").dropna()
+    if s.empty:
+        return None
+    cnt = s.value_counts().sort_index()
+    fig = go.Figure()
+    fig.add_trace(go.Bar(
+        x=cnt.index.astype(int), y=cnt.values,
+        marker_color=color, text=cnt.values, textposition="outside",
+    ))
+    fig.add_vline(
+        x=s.mean(), line_dash="dash", line_color=COLOR_MEDIA, line_width=2,
+        annotation_text=f"Prom: {s.mean():.2f}", annotation_position="top right",
+    )
+    fig.update_layout(
+        height=280, title=dict(text=titulo, x=0.5),
+        xaxis_title="Goles", yaxis_title="Frecuencia",
+        xaxis=dict(dtick=1), margin=dict(l=20, r=20, t=50, b=40),
+        **_THEME,
+    )
+    return fig
+
+
+def grafico_pred_equipo(df_rf, equipo: str, color: str) -> go.Figure | None:
+    if df_rf is None or df_rf.empty:
+        return None
+    fig = go.Figure()
+    fig.add_trace(go.Bar(
+        x=df_rf["Estadística"], y=df_rf["Predicción RF"], name="Predicción RF",
+        marker_color=color,
+        text=[f"{v:.2f}" for v in df_rf["Predicción RF"]], textposition="outside",
+    ))
+    fig.add_trace(go.Bar(
+        x=df_rf["Estadística"], y=df_rf["Promedio Histórico"], name="Promedio",
+        marker_color=COLOR_EMPATE,
+        text=[f"{v:.2f}" for v in df_rf["Promedio Histórico"]], textposition="outside",
+    ))
+    my = max(df_rf["Predicción RF"].max(), df_rf["Promedio Histórico"].max()) * 1.4
+    fig.update_layout(
+        barmode="group", height=400, title=dict(text=f"Predicción RF — {equipo}", x=0.5),
+        xaxis_tickangle=-35, yaxis=dict(range=[0, my]),
+        margin=dict(l=20, r=20, t=60, b=80),
+        legend=dict(orientation="h", yanchor="bottom", y=-0.35, xanchor="center", x=0.5),
+        **_THEME,
+    )
+    return fig
+
+
+def grafico_gauge_kelly(val: float, titulo: str, color: str) -> go.Figure:
+    v = max(0, min(val, KELLY_GAUGE_MAX))
+    fig = go.Figure(go.Indicator(
+        mode="gauge+number", value=v,
+        number={"suffix": "%", "font": {"size": 36}},
+        title={"text": titulo, "font": {"size": 14}},
+        gauge={
+            "axis": {"range": [0, KELLY_GAUGE_MAX]},
+            "bar": {"color": color},
+            "bgcolor": "white",
+            "borderwidth": 1,
+            "steps": [
+                {"range": [0, 5], "color": "#E8F5E9"},
+                {"range": [5, 15], "color": "#C8E6C9"},
+                {"range": [15, 30], "color": "#FFF9C4"},
+                {"range": [30, 50], "color": "#FFCDD2"},
+            ],
+            "threshold": {"line": {"color": "red", "width": 3}, "thickness": 0.8, "value": 25},
+        },
+    ))
+    fig.update_layout(height=250, margin=dict(l=20, r=20, t=50, b=20), **_THEME)
+    return fig
+
+
+def grafico_kelly_barras(df_kelly) -> go.Figure:
+    cols = {"Local": COLOR_LOCAL, "Empate": COLOR_EMPATE, "Visitante": COLOR_VISITANTE}
+    fig = go.Figure()
+    for _, r in df_kelly.iterrows():
+        k = max(r["Kelly (%)"], 0)
+        fig.add_trace(go.Bar(
+            y=[r["Resultado"]], x=[k], orientation="h",
+            marker_color=cols.get(r["Resultado"], "#666"),
+            text=f"{k:.2f}%", textposition="outside", showlegend=False,
+        ))
+    fig.update_layout(
+        height=200, title=dict(text="Kelly Criterion — % Óptimo", x=0.5),
+        xaxis_title="Kelly %",
+        xaxis=dict(range=[0, max(df_kelly["Kelly (%)"].max() * 1.5, 10)]),
+        margin=dict(l=20, r=20, t=50, b=30),
+        **_THEME,
+    )
+    return fig

src/src/config.py

@@ -0,0 +1,202 @@
+# --- URLs de datos ---
+URL_FIXTURES = "https://www.football-data.co.uk/fixtures.csv"
+URL_HISTORICO_BASE = "https://www.football-data.co.uk/mmz4281/2526/{liga}.csv"
+ARCHIVO_DATASET = "dataset_partidos_procesados.csv"
+
+# --- Configuración de la página ---
+PAGE_TITLE = "Análisis de Ligas"
+PAGE_LAYOUT = "wide"
+
+# --- Colores ---
+COLOR_LOCAL = "#2570D4"
+COLOR_VISITANTE = "#7C4DCC"
+COLOR_EMPATE = "#9BA3AE"
+COLOR_POSITIVO = "#0D9E6E"
+COLOR_NEGATIVO = "#D93025"
+COLOR_NEUTRO = "#5A6270"
+COLOR_MEDIA = "#B07D0E"
+COLOR_ADVERTENCIA = "#f9a825"
+COLOR_LOCAL_ALPHA = "rgba(37,112,212,0.3)"
+COLOR_VISITANTE_ALPHA = "rgba(124,77,204,0.3)"
+COLOR_LOCAL_HT = "#7DAAED"
+COLOR_VISITANTE_HT = "#B599E0"
+
+# --- Hiperparámetros del modelo Random Forest ---
+RF_N_ESTIMATORS = 100
+RF_MAX_DEPTH = 6
+RF_MIN_SAMPLES_SPLIT = 4
+RF_RANDOM_STATE = 42
+RF_N_ROLLING_DEFAULT = 10
+RF_N_ROLLING_MIN = 3
+RF_N_ROLLING_MAX = 15
+
+# --- Umbrales ---
+CV_UMBRAL_ESTABLE = 10.0    # CV < 10% → dato estable
+R2_UMBRAL_BUENO = 0.5       # R² ≥ 0.5 → buen ajuste
+R2_UMBRAL_ACEPTABLE = 0.3   # R² ≥ 0.3 → ajuste aceptable
+KELLY_MIN_RECOMENDADO = 5.0  # Kelly > 5% → apuesta recomendada
+KELLY_GAUGE_MAX = 50        # Rango máximo del gauge Kelly
+
+# --- Bankroll ---
+BANK_MIN = 100
+BANK_DEFAULT = 1_000
+BANK_STEP = 100
+
+# --- Tipos de apuesta ---
+TIPOS_APUESTA = ["⚽ Goles", "🎯 Tiros", "🎯 T. a puerta", "🚩 Córners", "📋 Otros"]
+
+COLS_APUESTA = {
+    "⚽ Goles":       {"local": ["FTHG", "HTHG"], "visit": ["FTAG", "HTAG"],
+                       "label": "Goles anotados por partido"},
+    "🎯 Tiros":       {"local": ["HS"],  "visit": ["AS"],
+                       "label": "Tiros totales por partido"},
+    "🎯 T. a puerta": {"local": ["HST"], "visit": ["AST"],
+                       "label": "Tiros a puerta por partido"},
+    "🚩 Córners":     {"local": ["HC"],  "visit": ["AC"],
+                       "label": "Córners por partido"},
+    "📋 Otros":       {"local": ["HF", "HY", "HR"], "visit": ["AF", "AY", "AR"],
+                       "label": "Faltas, amarillas y rojas por partido"},
+}
+
+# --- Información de ligas (código → país, bandera, nombre) ---
+LIGA_INFO = {
+    "E0":  {"pais": "Inglaterra",   "bandera": "🏴󠁧󠁢󠁥󠁮󠁧󠁿", "liga": "Premier League"},
+    "E1":  {"pais": "Inglaterra",   "bandera": "🏴󠁧󠁢󠁥󠁮󠁧󠁿", "liga": "Championship"},
+    "E2":  {"pais": "Inglaterra",   "bandera": "🏴󠁧󠁢󠁥󠁮󠁧󠁿", "liga": "League One"},
+    "E3":  {"pais": "Inglaterra",   "bandera": "🏴󠁧󠁢󠁥󠁮󠁧󠁿", "liga": "League Two"},
+    "EC":  {"pais": "Inglaterra",   "bandera": "🏴󠁧󠁢󠁥󠁮󠁧󠁿", "liga": "Conference"},
+    "SC0": {"pais": "Escocia",      "bandera": "🏴󠁧󠁢󠁳󠁣󠁴󠁿", "liga": "Premiership"},
+    "SC1": {"pais": "Escocia",      "bandera": "🏴󠁧󠁢󠁳󠁣󠁴󠁿", "liga": "Championship"},
+    "SC2": {"pais": "Escocia",      "bandera": "🏴󠁧󠁢󠁳󠁣󠁴󠁿", "liga": "League One"},
+    "SC3": {"pais": "Escocia",      "bandera": "🏴󠁧󠁢󠁳󠁣󠁴󠁿", "liga": "League Two"},
+    "D1":  {"pais": "Alemania",     "bandera": "🇩🇪", "liga": "Bundesliga"},
+    "D2":  {"pais": "Alemania",     "bandera": "🇩🇪", "liga": "2. Bundesliga"},
+    "I1":  {"pais": "Italia",       "bandera": "🇮🇹", "liga": "Serie A"},
+    "I2":  {"pais": "Italia",       "bandera": "🇮🇹", "liga": "Serie B"},
+    "SP1": {"pais": "España",       "bandera": "🇪🇸", "liga": "La Liga"},
+    "SP2": {"pais": "España",       "bandera": "🇪🇸", "liga": "Segunda División"},
+    "F1":  {"pais": "Francia",      "bandera": "🇫🇷", "liga": "Ligue 1"},
+    "F2":  {"pais": "Francia",      "bandera": "🇫🇷", "liga": "Ligue 2"},
+    "N1":  {"pais": "Países Bajos", "bandera": "🇳🇱", "liga": "Eredivisie"},
+    "B1":  {"pais": "Bélgica",      "bandera": "🇧🇪", "liga": "First Division A"},
+    "P1":  {"pais": "Portugal",     "bandera": "🇵🇹", "liga": "Primeira Liga"},
+    "T1":  {"pais": "Turquía",      "bandera": "🇹🇷", "liga": "Süper Lig"},
+    "G1":  {"pais": "Grecia",       "bandera": "🇬🇷", "liga": "Super League"},
+    "ARG": {"pais": "Argentina",    "bandera": "🇦🇷", "liga": "Primera División"},
+    "BRA": {"pais": "Brasil",       "bandera": "🇧🇷", "liga": "Série A"},
+    "MEX": {"pais": "México",       "bandera": "🇲🇽", "liga": "Liga MX"},
+}
+
+# --- Mínimo de partidos para análisis ---
+MIN_PARTIDOS_EQUIPO = 3
+MIN_PARTIDOS_RF = 10
+MIN_PARTIDOS_RF_TARGET = 5
+
+# --- Mapeos de columnas ---
+COLS_LOCAL = {
+    "FTHG": "Goles (FT)",
+    "HTHG": "Goles (HT)",
+    "HS": "Tiros",
+    "HST": "Tiros a Puerta",
+    "HC": "Córners",
+    "HF": "Faltas",
+    "HY": "Tarjetas Amarillas",
+    "HR": "Tarjetas Rojas",
+}
+COLS_VISITANTE = {
+    "FTAG": "Goles (FT)",
+    "HTAG": "Goles (HT)",
+    "AS": "Tiros",
+    "AST": "Tiros a Puerta",
+    "AC": "Córners",
+    "AF": "Faltas",
+    "AY": "Tarjetas Amarillas",
+    "AR": "Tarjetas Rojas",
+}
+COLS_CUOTAS = {
+    "B365H": "Bet365 Local",
+    "B365D": "Bet365 Empate",
+    "B365A": "Bet365 Visitante",
+}
+RENAME_PARTIDOS = {
+    "Date": "Fecha",
+    "Div": "Liga",
+    "HomeTeam": "Local",
+    "AwayTeam": "Visitante",
+    "FTHG": "Goles Local (FT)",
+    "FTAG": "Goles Visitante (FT)",
+    "FTR": "Resultado",
+    "HTHG": "Goles Local (HT)",
+    "HTAG": "Goles Visitante (HT)",
+    "HS": "Tiros Local",
+    "AS": "Tiros Visitante",
+    "HST": "Tiros a Puerta Local",
+    "AST": "Tiros a Puerta Visitante",
+    "HC": "Córners Local",
+    "AC": "Córners Visitante",
+    "HF": "Faltas Local",
+    "AF": "Faltas Visitante",
+    "HY": "Amarillas Local",
+    "AY": "Amarillas Visitante",
+    "HR": "Rojas Local",
+    "AR": "Rojas Visitante",
+    "B365H": "Bet365 Local",
+    "B365D": "Bet365 Empate",
+    "B365A": "Bet365 Visitante",
+    "Referee": "Árbitro",
+}
+
+# --- Pares de comparación (nombre_display, col_ataque, col_defensa) ---
+PARES_ATAQUE_LOCAL = [
+    ("Goles (FT)", "FTHG", "FTHG"),
+    ("Goles (HT)", "HTHG", "HTHG"),
+    ("Tiros", "HS", "HS"),
+    ("Tiros a Puerta", "HST", "HST"),
+    ("Córners", "HC", "HC"),
+    ("Faltas", "HF", "HF"),
+]
+PARES_ATAQUE_VISITANTE = [
+    ("Goles (FT)", "FTAG", "FTAG"),
+    ("Goles (HT)", "HTAG", "HTAG"),
+    ("Tiros", "AS", "AS"),
+    ("Tiros a Puerta", "AST", "AST"),
+    ("Córners", "AC", "AC"),
+    ("Faltas", "AF", "AF"),
+]
+
+# --- Features y targets del modelo ---
+FEATURE_COLS = [
+    "FTHG", "FTAG", "HTHG", "HTAG",
+    "HS", "AS", "HST", "AST",
+    "HC", "AC", "HF", "AF",
+    "HY", "AY", "HR", "AR",
+]
+RF_TARGETS_LOCAL = {
+    "FTHG": "Goles Anotados",
+    "FTAG": "Goles Recibidos",
+    "HTHG": "Goles Anotados (HT)",
+    "HTAG": "Goles Recibidos (HT)",
+    "HS": "Tiros",
+    "HST": "Tiros a Puerta",
+    "HC": "Córners",
+    "HF": "Faltas",
+    "HY": "Amarillas",
+    "HR": "Rojas",
+}
+RF_TARGETS_VISIT = {
+    "FTAG": "Goles Anotados",
+    "FTHG": "Goles Recibidos",
+    "HTAG": "Goles Anotados (HT)",
+    "HTHG": "Goles Recibidos (HT)",
+    "AS": "Tiros",
+    "AST": "Tiros a Puerta",
+    "AC": "Córners",
+    "AF": "Faltas",
+    "AY": "Amarillas",
+    "AR": "Rojas",
+}
+
+# --- Categorías para el radar ---
+RADAR_CATS = ["Goles", "Tiros", "Tiros a Puerta", "Córners", "Faltas", "Amarillas"]
+RADAR_COLS_LOCAL = ["FTHG", "HS", "HST", "HC", "HF", "HY"]
+RADAR_COLS_VISIT = ["FTAG", "AS", "AST", "AC", "AF", "AY"]

src/src/data_loader.py

@@ -0,0 +1,28 @@
+import streamlit as st
+import pandas as pd
+
+from src.config import URL_HISTORICO_BASE, ARCHIVO_DATASET
+
+
+@st.cache_data
+def cargar_fixtures(url: str) -> pd.DataFrame | None:
+    try:
+        df = pd.read_csv(url)
+        df.to_csv(ARCHIVO_DATASET, index=False)
+        return df
+    except Exception as e:
+        st.error(f"Error al cargar fixtures: {e}")
+        return None
+
+
+@st.cache_data
+def cargar_historico(liga: str) -> pd.DataFrame | None:
+    try:
+        return pd.read_csv(URL_HISTORICO_BASE.format(liga=liga))
+    except Exception as e:
+        st.warning(f"No se pudieron cargar datos históricos para '{liga}': {e}")
+        return None
+
+
+def ultimos_n(df: pd.DataFrame, n: int) -> pd.DataFrame:
+    return df if n == 0 else df.tail(n)

src/src/models.py

@@ -0,0 +1,169 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+from sklearn.ensemble import RandomForestRegressor
+
+from src.config import (
+    FEATURE_COLS,
+    RF_N_ESTIMATORS,
+    RF_MAX_DEPTH,
+    RF_MIN_SAMPLES_SPLIT,
+    RF_RANDOM_STATE,
+    MIN_PARTIDOS_RF,
+    MIN_PARTIDOS_RF_TARGET,
+    COLOR_POSITIVO,
+    COLOR_NEGATIVO,
+    COLOR_ADVERTENCIA,
+)
+from src.charts import grafico_pred_equipo
+
+
+def preparar_dataset_rf(df_hist: pd.DataFrame, n_rolling: int = 5) -> pd.DataFrame:
+    df = df_hist.copy().reset_index(drop=True)
+    for c in FEATURE_COLS:
+        if c in df.columns:
+            df[c] = pd.to_numeric(df[c], errors="coerce")
+    filas = []
+    for idx in range(len(df)):
+        row = df.iloc[idx]
+        prev = df.iloc[:idx]
+        if len(prev) < n_rolling:
+            continue
+        feat: dict = {}
+        for c in FEATURE_COLS:
+            if c in df.columns:
+                dh = prev[prev["HomeTeam"] == row["HomeTeam"]].tail(n_rolling)
+                feat[f"H_{c}_avg"] = dh[c].mean() if not dh.empty else 0
+                da = prev[prev["AwayTeam"] == row["AwayTeam"]].tail(n_rolling)
+                feat[f"A_{c}_avg"] = da[c].mean() if not da.empty else 0
+        feat["H_n"] = len(
+            pd.concat([
+                prev[prev["HomeTeam"] == row["HomeTeam"]],
+                prev[prev["AwayTeam"] == row["HomeTeam"]],
+            ]).tail(n_rolling)
+        )
+        feat["A_n"] = len(
+            pd.concat([
+                prev[prev["HomeTeam"] == row["AwayTeam"]],
+                prev[prev["AwayTeam"] == row["AwayTeam"]],
+            ]).tail(n_rolling)
+        )
+        for c in FEATURE_COLS:
+            if c in df.columns:
+                feat[f"target_{c}"] = row[c]
+        filas.append(feat)
+    return pd.DataFrame(filas)
+
+
+@st.cache_data
+def entrenar_rf_equipo(
+    df_hist: pd.DataFrame,
+    eq_local: str,
+    eq_visit: str,
+    n_rolling: int,
+    targets_dict: dict,
+) -> pd.DataFrame | None:
+    df_rf = preparar_dataset_rf(df_hist, n_rolling)
+    if df_rf.empty or len(df_rf) < MIN_PARTIDOS_RF:
+        return None
+
+    fnames = [c for c in df_rf.columns if c.startswith("H_") or c.startswith("A_")]
+    df_rf[fnames] = df_rf[fnames].fillna(0)
+
+    # Punto de predicción: características del próximo partido
+    fp: dict = {}
+    for c in FEATURE_COLS:
+        if c in df_hist.columns:
+            dh = df_hist[df_hist["HomeTeam"] == eq_local].tail(n_rolling)
+            fp[f"H_{c}_avg"] = pd.to_numeric(dh[c], errors="coerce").mean() if not dh.empty else 0
+            da = df_hist[df_hist["AwayTeam"] == eq_visit].tail(n_rolling)
+            fp[f"A_{c}_avg"] = pd.to_numeric(da[c], errors="coerce").mean() if not da.empty else 0
+    fp["H_n"] = len(
+        pd.concat([
+            df_hist[df_hist["HomeTeam"] == eq_local],
+            df_hist[df_hist["AwayTeam"] == eq_local],
+        ]).tail(n_rolling)
+    )
+    fp["A_n"] = len(
+        pd.concat([
+            df_hist[df_hist["HomeTeam"] == eq_visit],
+            df_hist[df_hist["AwayTeam"] == eq_visit],
+        ]).tail(n_rolling)
+    )
+    Xp = pd.DataFrame([fp])[fnames].fillna(0)
+
+    res = []
+    for col_real, nombre in targets_dict.items():
+        tc = f"target_{col_real}"
+        if tc not in df_rf.columns:
+            continue
+        y = df_rf[tc].dropna()
+        X = df_rf.loc[y.index, fnames]
+        if len(y) < MIN_PARTIDOS_RF_TARGET:
+            continue
+        try:
+            rf = RandomForestRegressor(
+                n_estimators=RF_N_ESTIMATORS,
+                max_depth=RF_MAX_DEPTH,
+                min_samples_split=RF_MIN_SAMPLES_SPLIT,
+                random_state=RF_RANDOM_STATE,
+                n_jobs=-1,
+            )
+            rf.fit(X, y)
+            pred = rf.predict(Xp)[0]
+            res.append({
+                "Estadística": nombre,
+                "Predicción RF": round(pred, 2),
+                "Promedio Histórico": round(y.mean(), 2),
+                "Diferencia": round(pred - y.mean(), 2),
+                "R² (ajuste)": round(rf.score(X, y), 3),
+            })
+        except Exception:
+            continue
+    return pd.DataFrame(res) if res else None
+
+
+def mostrar_rf_equipo(df_rf: pd.DataFrame, equipo: str, color: str) -> None:
+    fig = grafico_pred_equipo(df_rf, equipo, color)
+    if fig:
+        st.plotly_chart(fig, use_container_width=True)
+
+    def _color_r2(val) -> str:
+        try:
+            n = float(val)
+            if n >= 0.5:
+                return f"background-color: {COLOR_POSITIVO}; color: white; font-weight: bold"
+            elif n >= 0.3:
+                return f"background-color: {COLOR_ADVERTENCIA}; color: black; font-weight: bold"
+            else:
+                return f"background-color: {COLOR_NEGATIVO}; color: white; font-weight: bold"
+        except Exception:
+            return ""
+
+    def _color_diff(val) -> str:
+        try:
+            n = float(val)
+            if n > 0:
+                return f"color: {COLOR_POSITIVO}; font-weight: bold"
+            elif n < 0:
+                return f"color: {COLOR_NEGATIVO}; font-weight: bold"
+        except Exception:
+            pass
+        return ""
+
+    styled = df_rf.style.map(_color_r2, subset=["R² (ajuste)"]).map(
+        _color_diff, subset=["Diferencia"]
+    ).format({
+        "Predicción RF": "{:.2f}",
+        "Promedio Histórico": "{:.2f}",
+        "Diferencia": "{:+.2f}",
+        "R² (ajuste)": "{:.3f}",
+    })
+    st.dataframe(styled, use_container_width=True, hide_index=True)
+
+    cols_m = st.columns(min(len(df_rf), 5))
+    for i, (_, row) in enumerate(df_rf.iterrows()):
+        if i >= len(cols_m):
+            break
+        with cols_m[i]:
+            st.metric(row["Estadística"], f"{row['Predicción RF']:.1f}", f"{row['Diferencia']:+.2f}")

src/src/social_image.py

@@ -0,0 +1,259 @@
+"""
+Genera una imagen 1080×1080 minimalista con colores claros
+para compartir en redes sociales.
+"""
+import plotly.graph_objects as go
+
+from src.config import LIGA_INFO
+
+# ── Paleta clara minimalista ─────────────────────────────────────────────────
+_BG      = "#FFFFFF"
+_BG2     = "#F6F8FA"
+_BORDER  = "#D0D7DE"
+_BLUE    = "#1A6FD4"
+_RED     = "#CF2A1E"
+_DRAW    = "#8B98A8"
+_TEXT    = "#1C2128"
+_MUTED   = "#6E7781"
+_GREEN   = "#1A7F4B"
+_ORANGE  = "#9A5900"
+_W       = 1080
+_H       = 1080
+
+_FONT = "Inter,'Helvetica Neue',Arial,sans-serif"
+
+
+def _r(type_="rect", x0=0, y0=0, x1=1, y1=1,
+       fill="#FFFFFF", border="#D0D7DE", bw=1, layer="below") -> dict:
+    return dict(type=type_, xref="paper", yref="paper",
+                x0=x0, y0=y0, x1=x1, y1=y1,
+                fillcolor=fill,
+                line=dict(color=border, width=bw),
+                layer=layer)
+
+
+def _a(text, x, y, size=16, color=_TEXT, anchor="center",
+       bold=False, italic=False) -> dict:
+    fmt = f"<b>{text}</b>" if bold else (f"<i>{text}</i>" if italic else str(text))
+    return dict(text=fmt, x=x, y=y,
+                xref="paper", yref="paper",
+                showarrow=False,
+                xanchor=anchor, yanchor="middle",
+                font=dict(size=size, color=color, family=_FONT))
+
+
+def _prob_bar(pH, pD, pA, x0, x1, y, h) -> list:
+    w = x1 - x0
+    wH, wD, wA = w * pH / 100, w * pD / 100, w * pA / 100
+    r = h / 2  # border-radius approximation — use rounded edges via line
+    return [
+        dict(type="rect", xref="paper", yref="paper",
+             x0=x0,        x1=x0+wH,       y0=y, y1=y+h,
+             fillcolor=_BLUE, line_width=0),
+        dict(type="rect", xref="paper", yref="paper",
+             x0=x0+wH,     x1=x0+wH+wD,    y0=y, y1=y+h,
+             fillcolor=_DRAW, line_width=0),
+        dict(type="rect", xref="paper", yref="paper",
+             x0=x0+wH+wD,  x1=x1,           y0=y, y1=y+h,
+             fillcolor=_RED, line_width=0),
+    ]
+
+
+def generar_post_imagen(
+    eq_l: str, eq_v: str, liga: str,
+    pH: float, pD: float, pA: float,
+    cH: float, cD: float, cA: float,
+    l_anota: float, v_recibe: float,
+    v_anota: float, l_recibe: float,
+    kelly_ok: bool,
+    periodo: str = "Todos",
+    xg_l: float | None = None,
+    xg_v: float | None = None,
+) -> bytes:
+    """
+    Genera un PNG 1080×1080 minimalista con colores claros.
+    Devuelve bytes listos para st.download_button.
+    """
+    info     = LIGA_INFO.get(liga, {})
+    bandera  = info.get("bandera", "🏟️")
+    nombre_liga = info.get("liga", liga)
+
+    est_l = round((l_anota + v_recibe) / 2, 2)
+    est_v = round((v_anota + l_recibe) / 2, 2)
+    est_t = round(est_l + est_v, 2)
+    over_lbl   = "Sobre 2.5" if est_t > 2.5 else "Bajo 2.5"
+    over_color = _GREEN if est_t > 2.5 else _ORANGE
+
+    shapes, ann = [], []
+
+    # ── Fondo blanco ─────────────────────────────────────────────────────────
+    shapes.append(_r(fill=_BG, border=_BG, bw=0))
+
+    # ── Acento superior (línea fina azul) ────────────────────────────────────
+    shapes.append(_r(y0=0.964, y1=0.972, fill=_BLUE, border=_BLUE, bw=0))
+
+    # ── Liga + bandera ────────────────────────────────────────────────────────
+    ann.append(_a(f"{bandera}  {nombre_liga}",
+                  0.5, 0.948, size=18, color=_MUTED))
+
+    # ── Separador ────────────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.926, y1=0.9265, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Equipos ───────────────────────────────────────────────────────────────
+    ann += [
+        _a(eq_l, 0.24, 0.886, size=34, bold=True, color=_BLUE),
+        _a("vs",  0.50, 0.886, size=20, color=_MUTED, italic=True),
+        _a(eq_v, 0.76, 0.886, size=34, bold=True, color=_RED),
+        _a("Local",     0.24, 0.855, size=13, color=_MUTED),
+        _a(f"{periodo}", 0.50, 0.855, size=13, color=_MUTED),
+        _a("Visitante", 0.76, 0.855, size=13, color=_MUTED),
+    ]
+
+    # ── Separador ────────────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.834, y1=0.8345, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Probabilidades ─────────────────────��──────────────────────────────────
+    if kelly_ok:
+        _inv = 1/cH + 1/cD + 1/cA
+        bH = 1/cH / _inv * 100
+        bD = 1/cD / _inv * 100
+        bA = 1/cA / _inv * 100
+        prob_lbl = "Probabilidades implícitas (cuotas)"
+    else:
+        bH, bD, bA = pH, pD, pA
+        prob_lbl = "Probabilidades estimadas"
+    ann.append(_a(prob_lbl, 0.5, 0.818, size=12, color=_MUTED))
+    shapes += _prob_bar(bH, bD, bA, 0.06, 0.94, 0.775, 0.026)
+    ann += [
+        _a(f"{bH:.1f}%", 0.18, 0.752, size=26, bold=True, color=_BLUE),
+        _a(f"{bD:.1f}%", 0.50, 0.752, size=26, bold=True, color=_DRAW),
+        _a(f"{bA:.1f}%", 0.82, 0.752, size=26, bold=True, color=_RED),
+        _a("Local",     0.18, 0.724, size=12, color=_MUTED),
+        _a("Empate",    0.50, 0.724, size=12, color=_MUTED),
+        _a("Visitante", 0.82, 0.724, size=12, color=_MUTED),
+    ]
+    if kelly_ok:
+        ann += [
+            _a(f"cuota {cH:.2f}", 0.18, 0.709, size=11, color=_MUTED, italic=True),
+            _a(f"cuota {cD:.2f}", 0.50, 0.709, size=11, color=_MUTED, italic=True),
+            _a(f"cuota {cA:.2f}", 0.82, 0.709, size=11, color=_MUTED, italic=True),
+        ]
+
+    # ── Separador ────────────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.690, y1=0.6905, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Goles esperados / xG ─────────────────────────────────────────────────
+    _has_xg = xg_l is not None and xg_v is not None
+    if _has_xg:
+        show_l, show_v = xg_l, xg_v
+        show_t = round(xg_l + xg_v, 2)
+        section_lbl = "xG predicción del partido"
+    else:
+        show_l, show_v = est_l, est_v
+        show_t = est_t
+        section_lbl = "Goles esperados por partido"
+    over_lbl   = "Sobre 2.5" if show_t > 2.5 else "Bajo 2.5"
+    over_color = _GREEN if show_t > 2.5 else _ORANGE
+
+    ann.append(_a(section_lbl, 0.5, 0.673, size=12, color=_MUTED))
+    ann += [
+        _a(f"{show_l:.2f}", 0.20, 0.638, size=48, bold=True, color=_BLUE),
+        _a(f"{show_v:.2f}", 0.80, 0.638, size=48, bold=True, color=_RED),
+        _a(f"{show_t:.2f}", 0.50, 0.645, size=28, bold=True, color=_TEXT),
+        _a("total",         0.50, 0.614, size=12, color=_MUTED),
+        _a(over_lbl,        0.50, 0.598, size=13, bold=True, color=over_color),
+        _a(eq_l[:16],   0.20, 0.602, size=11, color=_MUTED),
+        _a(eq_v[:16],   0.80, 0.602, size=11, color=_MUTED),
+    ]
+    if _has_xg:
+        ann += [
+            _a(f"(avg {est_l:.2f})", 0.20, 0.584, size=10, color=_MUTED, italic=True),
+            _a(f"(avg {est_v:.2f})", 0.80, 0.584, size=10, color=_MUTED, italic=True),
+        ]
+
+    # ── Separador ────────────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.576, y1=0.5765, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Tabla de promedios — 2 columnas ──────────────────────────────────────
+    ann.append(_a("Promedios por partido", 0.5, 0.559, size=12, color=_MUTED))
+
+    # Fila encabezados
+    ann += [
+        _a("",          0.06, 0.535, size=12, color=_MUTED, anchor="left"),
+        _a(eq_l[:16],   0.45, 0.535, size=13, bold=True, color=_BLUE, anchor="right"),
+        _a(eq_v[:16],   0.94, 0.535, size=13, bold=True, color=_RED,  anchor="right"),
+    ]
+    shapes.append(_r(y0=0.521, y1=0.5215, fill=_BORDER, border=_BORDER, bw=0))
+
+    # Columna izquierda = Local, columna derecha = Visitante (siempre)
+    rows = [
+        ("⚽  Goles anotados",              l_anota,  v_anota),   # local anota / visit anota
+        ("🛡️  Goles recibidos",             l_recibe, v_recibe),  # local recibe / visit recibe
+        ("⚔️  Local ataca · Visita defiende", l_anota,  v_recibe), # local anota vs visita recibe
+        ("⚔️  Visita ataca · Local defiende", l_recibe, v_anota),  # local recibe vs visita anota
+    ]
+    row_ys = [0.499, 0.463, 0.427, 0.391]
+
+    for (label, vl, vv), ry in zip(rows, row_ys):
+        # alternating row background
+        shapes.append(_r(y0=ry-0.018, y1=ry+0.018,
+                         fill=_BG2 if rows.index((label, vl, vv)) % 2 == 0 else _BG,
+                         border=_BG2, bw=0))
+        ann += [
+            _a(label, 0.06, ry, size=13, color=_TEXT, anchor="left"),
+            _a(f"{vl:.2f}", 0.45, ry, size=15, bold=True, color=_BLUE, anchor="right"),
+            _a(f"{vv:.2f}", 0.94, ry, size=15, bold=True, color=_RED,  anchor="right"),
+        ]
+
+    shapes.append(_r(y0=0.370, y1=0.3705, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Delta ataque ─────────────────────────────────────────────────────────
+    # Positivo = la defensa rival recibe más de lo que el atacante anota → defensa débil
+    d_l = round(v_recibe - l_anota, 2)
+    d_v = round(l_recibe - v_anota, 2)
+    d_l_color = _GREEN if d_l > 0 else (_RED if d_l < 0 else _MUTED)
+    d_v_color = _GREEN if d_v > 0 else (_RED if d_v < 0 else _MUTED)
+
+    def _delta_label(d):
+        if d > 0:   return "defensa débil"
+        if d < 0:   return "defensa sólida"
+        return "equilibrio"
+
+    ann += [
+        _a("Ventaja ataque vs defensa rival", 0.5, 0.356, size=12, color=_MUTED),
+        _a(f"{d_l:+.2f}", 0.30, 0.330, size=24, bold=True, color=d_l_color),
+        _a(f"{d_v:+.2f}", 0.70, 0.330, size=24, bold=True, color=d_v_color),
+        _a(f"{eq_l[:14]} ataca", 0.30, 0.308, size=11, color=_MUTED),
+        _a(f"{eq_v[:14]} ataca", 0.70, 0.308, size=11, color=_MUTED),
+        _a(_delta_label(d_l), 0.30, 0.291, size=11, bold=True, color=d_l_color),
+        _a(_delta_label(d_v), 0.70, 0.291, size=11, bold=True, color=d_v_color),
+    ]
+
+    # ── Separador ────────────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.282, y1=0.2825, fill=_BORDER, border=_BORDER, bw=0))
+
+    # ── Branding mínimo ───────────────────────────────────────────────────────
+    ann += [
+        _a("⚽  FutStats", 0.5, 0.255, size=20, bold=True, color=_TEXT),
+        _a("Análisis estadístico de fútbol", 0.5, 0.228,
+           size=13, color=_MUTED, italic=True),
+        _a("El análisis estadístico no garantiza resultados. Apuesta con responsabilidad.",
+           0.5, 0.200, size=11, color=_MUTED),
+    ]
+
+    # ── Acento inferior ───────────────────────────────────────────────────────
+    shapes.append(_r(y0=0.178, y1=0.186, fill=_BLUE, border=_BLUE, bw=0))
+
+    # ── Construir figura ──────────────────────────────────────────────────────
+    fig = go.Figure()
+    fig.update_layout(
+        width=_W, height=_H,
+        paper_bgcolor=_BG,
+        plot_bgcolor=_BG,
+        margin=dict(l=0, r=0, t=0, b=0),
+        shapes=shapes,
+        annotations=ann,
+        xaxis=dict(visible=False, range=[0, 1]),
+        yaxis=dict(visible=False, range=[0, 1]),
+    )
+    return fig.to_image(format="png", width=_W, height=_H, scale=2)

src/src/statistics.py

@@ -0,0 +1,137 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+
+from src.config import (
+    COLOR_POSITIVO,
+    COLOR_NEGATIVO,
+    COLOR_ADVERTENCIA,
+    RENAME_PARTIDOS,
+)
+
+
+def calcular_estadisticas(df: pd.DataFrame, cols_dict: dict) -> pd.DataFrame | None:
+    cols_validas = {k: v for k, v in cols_dict.items() if k in df.columns}
+    if not cols_validas:
+        return None
+    df_num = df[list(cols_validas.keys())].apply(pd.to_numeric, errors="coerce")
+    filas = []
+    for col_orig, col_nombre in cols_validas.items():
+        serie = df_num[col_orig].dropna()
+        if serie.empty:
+            continue
+        mo = serie.mode()
+        filas.append({
+            "Estadística": col_nombre,
+            "Promedio": round(serie.mean(), 2),
+            "Mediana": round(serie.median(), 2),
+            "Moda": mo.iloc[0] if not mo.empty else np.nan,
+            "Desv. Estándar": round(serie.std(), 2),
+        })
+    return pd.DataFrame(filas)
+
+
+def promedio_col(df: pd.DataFrame, col: str) -> float:
+    if col not in df.columns:
+        return np.nan
+    s = pd.to_numeric(df[col], errors="coerce").dropna()
+    return s.mean() if not s.empty else np.nan
+
+
+def construir_comparacion(
+    nombre_local: str,
+    nombre_visitante: str,
+    df_local: pd.DataFrame,
+    df_visit: pd.DataFrame,
+    pares: list,
+    tipo: str,
+) -> pd.DataFrame:
+    filas = []
+    for nombre_stat, col_atq, col_def in pares:
+        if tipo == "ataque_local":
+            va, vd = promedio_col(df_local, col_atq), promedio_col(df_visit, col_def)
+            la, ld = f"{nombre_local} genera", f"{nombre_visitante} recibe"
+        else:
+            va, vd = promedio_col(df_visit, col_atq), promedio_col(df_local, col_def)
+            la, ld = f"{nombre_visitante} genera", f"{nombre_local} recibe"
+        diff = vd - va if not (np.isnan(va) or np.isnan(vd)) else np.nan
+        filas.append({
+            "Estadística": nombre_stat,
+            la: round(va, 2) if not np.isnan(va) else "-",
+            ld: round(vd, 2) if not np.isnan(vd) else "-",
+            "Diferencia": round(diff, 2) if not np.isnan(diff) else "-",
+        })
+    return pd.DataFrame(filas)
+
+
+# --- Funciones de estilo ---
+
+def _colorear_diferencia_serie(series: pd.Series) -> list[str]:
+    vals = pd.to_numeric(series, errors="coerce")
+    umbral = vals.abs().std() * 0.5
+    result = []
+    for v in vals:
+        if pd.isna(v):
+            result.append("")
+        elif abs(v) <= umbral:
+            result.append(f"background-color: {COLOR_ADVERTENCIA}; color: black; font-weight: bold")
+        elif v > 0:
+            result.append(f"background-color: {COLOR_POSITIVO}; color: white; font-weight: bold")
+        else:
+            result.append(f"background-color: {COLOR_NEGATIVO}; color: white; font-weight: bold")
+    return result
+
+
+# --- Funciones de display (Streamlit) ---
+
+def _colorear_desv(series: pd.Series) -> list[str]:
+    vals = pd.to_numeric(series, errors="coerce")
+    q33 = vals.quantile(0.33)
+    q66 = vals.quantile(0.66)
+    result = []
+    for v in vals:
+        if pd.isna(v):
+            result.append("")
+        elif v <= q33:
+            result.append(f"background-color: {COLOR_POSITIVO}; color: white; font-weight: bold")
+        elif v <= q66:
+            result.append(f"background-color: {COLOR_ADVERTENCIA}; color: black; font-weight: bold")
+        else:
+            result.append(f"background-color: {COLOR_NEGATIVO}; color: white; font-weight: bold")
+    return result
+
+
+def mostrar_estadisticas_coloreadas(df_stats: pd.DataFrame | None) -> None:
+    if df_stats is None or df_stats.empty:
+        st.warning("No hay datos suficientes.")
+        return
+    num_cols = ["Promedio", "Mediana", "Desv. Estándar"]
+    styled = (
+        df_stats.style
+        .apply(_colorear_desv, subset=["Desv. Estándar"], axis=0)
+        .format({c: "{:.2f}" for c in num_cols} | {"Moda": "{}"})
+    )
+    st.dataframe(styled, use_container_width=True, hide_index=True)
+
+
+def mostrar_tabla_partidos(df: pd.DataFrame, cols_extra: list | None = None) -> None:
+    base = ["Date", "HomeTeam", "AwayTeam"]
+    if cols_extra:
+        base += [c for c in cols_extra if c in df.columns]
+    cols_mostrar = [c for c in base if c in df.columns]
+    st.dataframe(
+        df[cols_mostrar].copy().rename(columns=RENAME_PARTIDOS).reset_index(drop=True),
+        use_container_width=True,
+        hide_index=True,
+    )
+
+
+def mostrar_comparacion(df_comp: pd.DataFrame | None) -> None:
+    if df_comp is None or df_comp.empty:
+        st.warning("No hay datos para la comparación.")
+        return
+    st.dataframe(
+        df_comp.style.apply(_colorear_diferencia_serie, subset=["Diferencia"], axis=0),
+        use_container_width=True,
+        hide_index=True,
+    )

src/src/styles.py

@@ -0,0 +1,505 @@
+import streamlit as st
+
+_CSS = """<style>
+:root {
+  --bg-primary:    #F6F8FA;
+  --bg-secondary:  #FFFFFF;
+  --bg-card:       #FFFFFF;
+  --bg-hover:      #F0F2F5;
+  --bg-active:     #EBF2FF;
+  --border:        #E2E6EA;
+  --border-light:  #EAECEF;
+  --text-primary:  #1A1F2E;
+  --text-secondary:#5A6270;
+  --text-muted:    #9BA3AE;
+  --green:         #0D9E6E;
+  --green-dim:     rgba(13,158,110,0.08);
+  --blue:          #2570D4;
+  --blue-dim:      rgba(37,112,212,0.08);
+  --yellow:        #B07D0E;
+  --yellow-dim:    rgba(176,125,14,0.08);
+  --red:           #D93025;
+  --red-dim:       rgba(217,48,37,0.08);
+  --purple:        #7C4DCC;
+  --purple-dim:    rgba(124,77,204,0.08);
+  --radius-sm:     6px;
+  --radius-md:     10px;
+  --radius-lg:     14px;
+  --shadow:        0 2px 12px rgba(0,0,0,0.06);
+  --shadow-md:     0 4px 20px rgba(0,0,0,0.1);
+  --font:          -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
+}
+
+/* ─── App shell ─── */
+.stApp { background:var(--bg-primary) !important; font-family:var(--font) !important; }
+.main .block-container {
+  background:var(--bg-primary) !important;
+  padding:24px 28px 60px !important;
+  max-width:1400px !important;
+}
+
+/* ─── Hide Sidebar ─── */
+[data-testid="stSidebar"],
+[data-testid="stSidebarCollapsedControl"],
+[data-testid="stSidebarNav"],
+section[data-testid="stSidebar"] {
+  display:none !important;
+}
+.main { margin-left:0 !important; }
+
+/* ─── Typography ─── */
+h1,h2,h3,h4,h5 { font-family:var(--font) !important; color:var(--text-primary) !important; }
+
+/* ─── Metrics ─── */
+[data-testid="stMetric"] {
+  background:var(--bg-card) !important;
+  border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important;
+  padding:14px 16px !important;
+  box-shadow:var(--shadow) !important;
+}
+[data-testid="stMetricValue"] {
+  color:var(--green) !important; font-size:24px !important; font-weight:700 !important;
+}
+[data-testid="stMetricLabel"] {
+  color:var(--text-muted) !important; font-size:10px !important;
+  font-weight:600 !important; text-transform:uppercase !important; letter-spacing:0.5px !important;
+}
+[data-testid="stMetricDelta"] { font-size:11px !important; }
+
+/* ─── Dataframes ─── */
+[data-testid="stDataFrame"] {
+  border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important;
+  overflow:hidden !important;
+}
+
+/* ─── Expanders ─── */
+[data-testid="stExpander"] {
+  background:var(--bg-card) !important;
+  border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important;
+  overflow:hidden !important;
+  box-shadow:none !important;
+}
+[data-testid="stExpander"] details summary {
+  background:var(--bg-card) !important;
+  color:var(--text-secondary) !important;
+  font-size:13px !important; font-weight:500 !important; padding:12px 16px !important;
+}
+[data-testid="stExpander"] details summary:hover { background:var(--bg-hover) !important; }
+
+/* ─── Select boxes ─── */
+[data-testid="stSelectbox"]>div>div {
+  background:var(--bg-card) !important;
+  border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important;
+  color:var(--text-primary) !important; font-size:13px !important;
+}
+[data-testid="stSelectbox"] label {
+  color:var(--text-secondary) !important; font-size:11px !important;
+  font-weight:600 !important; text-transform:uppercase !important; letter-spacing:0.4px !important;
+}
+
+/* ─── Buttons ─── */
+.stButton>button {
+  background:var(--green) !important; color:#fff !important;
+  border:none !important; border-radius:var(--radius-md) !important;
+  font-weight:600 !important; font-size:13px !important;
+  padding:9px 22px !important; transition:opacity 0.15s !important; box-shadow:none !important;
+}
+.stButton>button:hover { opacity:0.88 !important; border:none !important; }
+.stButton>button[kind="primary"] { background:var(--blue) !important; }
+
+/* ─── Alerts ─── */
+[data-testid="stAlert"] { border-radius:var(--radius-md) !important; font-size:13px !important; }
+
+/* ─── Number input ─── */
+[data-testid="stNumberInput"] input {
+  background:var(--bg-card) !important; border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important; color:var(--text-primary) !important;
+}
+
+/* ─── Slider ─── */
+[data-testid="stSlider"] label {
+  color:var(--text-secondary) !important; font-size:11px !important;
+  font-weight:600 !important; text-transform:uppercase !important; letter-spacing:0.4px !important;
+}
+
+/* ─── Caption ─── */
+[data-testid="stCaptionContainer"] { color:var(--text-muted) !important; font-size:11px !important; }
+
+/* ─── Tabs ─── */
+.stTabs [data-testid="stTabBar"] {
+  background:var(--bg-card) !important;
+  border:1px solid var(--border) !important;
+  border-radius:var(--radius-md) !important;
+  padding:4px !important; gap:2px !important;
+  margin-bottom:20px !important;
+}
+.stTabs [data-testid="stTabBar"] button {
+  border-radius:var(--radius-sm) !important;
+  font-size:12px !important; font-weight:500 !important;
+  color:var(--text-secondary) !important;
+  background:transparent !important; border:none !important;
+  padding:8px 16px !important; white-space:nowrap !important;
+  transition:all 0.15s !important;
+}
+.stTabs [data-testid="stTabBar"] button:hover { color:var(--text-primary) !important; }
+.stTabs [data-testid="stTabBar"] button[aria-selected="true"] {
+  background:var(--bg-secondary) !important;
+  color:var(--text-primary) !important;
+  font-weight:600 !important;
+  box-shadow:0 1px 4px rgba(0,0,0,0.1) !important;
+}
+[data-testid="stTabPanel"] { padding:0 !important; }
+
+/* ─── Hide branding ─── */
+#MainMenu { visibility:hidden; }
+footer { visibility:hidden; }
+[data-testid="stToolbar"] { display:none !important; }
+[data-testid="stDecoration"] { display:none !important; }
+
+/* ══════════════════════════════════
+   CUSTOM COMPONENTS
+══════════════════════════════════ */
+
+/* Sidebar header */
+.fs-sb-header {
+  display:flex; align-items:center; gap:10px;
+  padding:16px 0 14px;
+  border-bottom:1px solid var(--border);
+  margin-bottom:16px;
+}
+.fs-sb-logo {
+  width:34px; height:34px; background:var(--green); border-radius:9px;
+  display:flex; align-items:center; justify-content:center;
+  font-size:17px; flex-shrink:0;
+}
+.fs-sb-title { font-size:14px; font-weight:700; color:var(--text-primary); line-height:1.2; }
+.fs-sb-sub { font-size:10px; color:var(--text-muted); }
+
+/* Sidebar section label */
+.fs-sb-sec {
+  font-size:10px; text-transform:uppercase; letter-spacing:0.8px;
+  color:var(--text-muted); font-weight:600; margin:14px 0 6px;
+}
+
+/* Sidebar match mini card */
+.fs-match-mini {
+  background:var(--bg-active); border:1px solid rgba(37,112,212,0.3);
+  border-radius:var(--radius-md); padding:10px 12px; margin:10px 0;
+}
+.fs-match-mini .home { color:var(--blue); font-weight:700; font-size:12px; }
+.fs-match-mini .away { color:var(--red); font-weight:700; font-size:12px; }
+.fs-match-mini .vs   { color:var(--text-muted); font-size:11px; padding:0 5px; }
+.fs-match-mini .meta { font-size:10px; color:var(--text-muted); margin-top:5px; }
+
+/* Sidebar stat pill */
+.fs-sb-pills {
+  display:flex; gap:6px; flex-wrap:wrap; margin-top:8px;
+}
+.fs-sb-pill {
+  display:inline-flex; align-items:center; gap:3px;
+  background:var(--bg-hover); border:1px solid var(--border);
+  border-radius:20px; padding:3px 9px;
+  font-size:11px; color:var(--text-secondary);
+}
+.fs-sb-pill span { font-weight:700; color:var(--text-primary); }
+
+/* Page header */
+.fs-header {
+  display:flex; align-items:center; gap:14px;
+  padding-bottom:18px; border-bottom:1px solid var(--border); margin-bottom:20px;
+}
+.fs-header h1 {
+  font-size:22px !important; font-weight:700 !important;
+  color:var(--text-primary) !important; margin:0 !important;
+  padding:0 !important; line-height:1.2 !important;
+}
+.fs-header .sub { font-size:12px; color:var(--text-muted); margin-top:3px; }
+
+/* Section title */
+.fs-sec {
+  display:flex; align-items:center; gap:8px;
+  font-size:11px; text-transform:uppercase; letter-spacing:0.8px;
+  color:var(--text-muted); font-weight:600; margin:20px 0 12px;
+}
+.fs-sec::after { content:''; flex:1; height:1px; background:var(--border-light); }
+
+/* Stats summary bar */
+.fs-stats-bar {
+  display:flex; align-items:stretch;
+  background:var(--bg-card); border:1px solid var(--border);
+  border-radius:var(--radius-lg); overflow:hidden; margin-bottom:20px;
+}
+.fs-stats-bar .si {
+  flex:1; text-align:center; padding:14px 16px;
+  border-right:1px solid var(--border-light);
+}
+.fs-stats-bar .si:last-child { border-right:none; }
+.fs-stats-bar .sn { font-size:24px; font-weight:700; color:var(--green); line-height:1; }
+.fs-stats-bar .sn.blue   { color:var(--blue); }
+.fs-stats-bar .sn.red    { color:var(--red); }
+.fs-stats-bar .sn.yellow { color:var(--yellow); }
+.fs-stats-bar .sl {
+  font-size:10px; color:var(--text-muted);
+  text-transform:uppercase; letter-spacing:0.5px; margin-top:4px;
+}
+
+/* Team analysis panels */
+.fs-panel-home {
+  background:var(--bg-card);
+  border:1px solid var(--border);
+  border-top:3px solid var(--blue);
+  border-radius:var(--radius-lg);
+  padding:16px 16px 4px;
+  margin-bottom:12px;
+}
+.fs-panel-away {
+  background:var(--bg-card);
+  border:1px solid var(--border);
+  border-top:3px solid var(--red);
+  border-radius:var(--radius-lg);
+  padding:16px 16px 4px;
+  margin-bottom:12px;
+}
+.fs-panel-title {
+  font-size:14px; font-weight:700; margin-bottom:12px;
+  display:flex; align-items:center; gap:8px;
+}
+.fs-panel-title.home { color:var(--blue); }
+.fs-panel-title.away { color:var(--red); }
+
+/* Probability bar */
+.fs-prob {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-radius:var(--radius-md); padding:16px; margin-bottom:14px;
+}
+.fs-prob .labels {
+  display:flex; justify-content:space-between;
+  font-size:12px; color:var(--text-secondary); margin-bottom:10px; font-weight:500;
+}
+.fs-prob .bar {
+  height:10px; border-radius:5px; background:var(--bg-hover); overflow:hidden; display:flex;
+}
+.fs-prob .seg { height:100%; transition:width 0.5s; }
+.fs-prob .seg.h { background:#2570D4; }
+.fs-prob .seg.d { background:#C0C6CE; }
+.fs-prob .seg.a { background:#D93025; }
+.fs-prob .vals {
+  display:flex; justify-content:space-between;
+  margin-top:10px; font-size:14px; font-weight:700;
+}
+.fs-prob .vh { color:#2570D4; }
+.fs-prob .vd { color:var(--text-secondary); }
+.fs-prob .va { color:#D93025; }
+
+/* Odds cards (1X2) */
+.fs-odds { display:grid; grid-template-columns:1fr 1fr 1fr; gap:10px; margin-bottom:16px; }
+.fs-odds .oc {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-radius:var(--radius-md); padding:16px 12px; text-align:center;
+  transition:border-color 0.15s;
+}
+.fs-odds .oc.best { border-color:var(--green); background:var(--green-dim); }
+.fs-odds .oc .on { font-size:11px; color:var(--text-muted); margin-bottom:6px; font-weight:500; }
+.fs-odds .oc .ov { font-size:26px; font-weight:700; color:var(--text-primary); line-height:1; }
+.fs-odds .oc.best .ov { color:var(--green); }
+.fs-odds .oc .op { font-size:11px; color:var(--text-secondary); margin-top:6px; }
+.fs-odds .oc .ove { font-size:11px; margin-top:3px; font-weight:600; }
+.fs-odds .oc .ove.pos { color:var(--green); }
+.fs-odds .oc .ove.neg { color:var(--red); }
+
+/* Kelly result cards */
+.fs-kelly-grid { display:grid; grid-template-columns:repeat(3,1fr); gap:12px; margin-bottom:20px; }
+.fs-kelly-card {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-top:4px solid var(--border);
+  border-radius:var(--radius-lg); padding:20px 14px; text-align:center;
+}
+.fs-kelly-card.positive { border-top-color:var(--green); background:var(--green-dim); border-color:rgba(13,158,110,0.2); }
+.fs-kelly-card.yellow   { border-top-color:var(--yellow); background:var(--yellow-dim); border-color:rgba(176,125,14,0.2); }
+.fs-kelly-card.negative { border-top-color:var(--red); }
+.fs-kelly-card .kc-result { font-size:10px; text-transform:uppercase; letter-spacing:0.5px; color:var(--text-muted); font-weight:600; margin-bottom:4px; }
+.fs-kelly-card .kc-team { font-size:13px; font-weight:700; color:var(--text-primary); margin-bottom:14px; white-space:nowrap; overflow:hidden; text-overflow:ellipsis; }
+.fs-kelly-card .kc-odd { font-size:32px; font-weight:700; color:var(--text-primary); line-height:1; margin-bottom:10px; }
+.fs-kelly-card.positive .kc-odd { color:var(--green); }
+.fs-kelly-card .kc-sep { width:32px; height:2px; background:var(--border); margin:0 auto 10px; border-radius:2px; }
+.fs-kelly-card .kc-pct { font-size:20px; font-weight:700; line-height:1; }
+.fs-kelly-card.positive .kc-pct { color:var(--green); }
+.fs-kelly-card.yellow   .kc-pct { color:var(--yellow); }
+.fs-kelly-card.negative .kc-pct { color:var(--red); }
+.fs-kelly-card .kc-label { font-size:10px; color:var(--text-muted); margin-top:3px; }
+.fs-kelly-card .kc-meta { margin-top:10px; padding-top:10px; border-top:1px solid var(--border-light); font-size:11px; color:var(--text-secondary); }
+.fs-kelly-card .kc-meta .pos { color:var(--green); font-weight:700; }
+.fs-kelly-card .kc-meta .neg { color:var(--red);   font-weight:700; }
+
+/* Recommendation banner */
+.fs-recommend {
+  display:flex; align-items:center; gap:14px;
+  background:var(--green-dim); border:1px solid rgba(13,158,110,0.25);
+  border-radius:var(--radius-md); padding:14px 18px; margin-top:4px;
+}
+.fs-recommend .rec-icon { font-size:24px; flex-shrink:0; }
+.fs-recommend .rec-title { font-size:13px; font-weight:700; color:var(--green); }
+.fs-recommend .rec-detail { font-size:12px; color:var(--text-secondary); margin-top:2px; }
+.fs-no-bet {
+  display:flex; align-items:center; gap:14px;
+  background:var(--red-dim); border:1px solid rgba(217,48,37,0.2);
+  border-radius:var(--radius-md); padding:14px 18px; margin-top:4px;
+}
+.fs-no-bet .nb-icon { font-size:22px; flex-shrink:0; }
+.fs-no-bet .nb-text { font-size:13px; font-weight:600; color:var(--red); }
+
+/* AI panel */
+.fs-ai {
+  background:linear-gradient(135deg,rgba(37,112,212,0.05),rgba(217,48,37,0.05));
+  border:1px solid rgba(37,112,212,0.2);
+  border-radius:var(--radius-md); padding:16px;
+}
+.fs-ai .ai-hdr { display:flex; align-items:center; gap:10px; margin-bottom:4px; }
+.fs-ai .ai-ico {
+  width:32px; height:32px; background:var(--blue-dim); border:1px solid var(--blue);
+  border-radius:var(--radius-sm); display:flex; align-items:center; justify-content:center; font-size:16px;
+}
+.fs-ai .ai-title { font-size:14px; font-weight:700; color:var(--blue); }
+.fs-ai .ai-sub { font-size:10px; color:var(--text-muted); }
+
+/* Empty state */
+.fs-empty {
+  text-align:center; padding:40px 20px;
+  color:var(--text-muted); font-size:13px;
+}
+.fs-empty .em-icon { font-size:32px; margin-bottom:8px; }
+
+/* ─── Match selection cards ─── */
+.fs-mc {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-top:3px solid var(--border-light);
+  border-radius:var(--radius-lg); padding:14px 12px;
+  transition:all 0.15s; margin-bottom:4px;
+}
+.fs-mc.selected {
+  border-top-color:var(--green);
+  background:var(--green-dim); border-color:rgba(13,158,110,0.25);
+}
+.fs-mc .mc-league {
+  font-size:10px; font-weight:700; color:var(--text-muted);
+  text-transform:uppercase; letter-spacing:0.5px; margin-bottom:8px;
+}
+.fs-mc .mc-home { color:var(--blue); font-weight:700; font-size:13px; display:block; }
+.fs-mc .mc-vs   { color:var(--text-muted); font-size:11px; padding:2px 0; display:block; }
+.fs-mc .mc-away { color:var(--red); font-weight:700; font-size:13px; display:block; }
+.fs-mc .mc-meta { font-size:11px; color:var(--text-muted); margin-top:8px; }
+
+/* date group label */
+.fs-date-label {
+  font-size:11px; font-weight:700; text-transform:uppercase; letter-spacing:0.6px;
+  color:var(--text-muted); padding:14px 0 8px; border-bottom:1px solid var(--border-light);
+  margin-bottom:10px;
+}
+
+/* Sidebar no-match state */
+.fs-sb-welcome {
+  text-align:center; padding:20px 8px;
+  color:var(--text-muted); font-size:12px;
+}
+.fs-sb-welcome .sw-icon { font-size:28px; margin-bottom:8px; }
+.fs-sb-welcome .sw-text { line-height:1.5; }
+
+/* ─── Page navigation links ─── */
+[data-testid="stPageLink"] {
+  margin:2px 0 !important;
+}
+[data-testid="stPageLink"] a {
+  border-radius:var(--radius-sm) !important;
+  padding:6px 10px !important;
+  font-size:12px !important;
+  color:var(--text-secondary) !important;
+  display:flex !important; align-items:center !important; gap:6px !important;
+  transition:all 0.15s !important;
+  text-decoration:none !important;
+  font-weight:500 !important;
+  white-space:nowrap !important;
+}
+[data-testid="stPageLink"] a:hover {
+  background:var(--bg-hover) !important;
+  color:var(--text-primary) !important;
+}
+[data-testid="stPageLink"] a[aria-current="page"] {
+  background:var(--bg-active) !important;
+  color:var(--blue) !important;
+  font-weight:600 !important;
+}
+
+/* ─── Topbar ─── */
+.fs-topbar {
+  display:flex; align-items:center; gap:10px;
+  padding:8px 0 12px; border-bottom:1px solid var(--border); margin-bottom:16px;
+}
+.fs-topbar-logo {
+  display:flex; align-items:center; gap:8px; flex-shrink:0;
+  font-size:15px; font-weight:700; color:var(--text-primary);
+}
+.fs-topbar-logo .tl-icon {
+  width:30px; height:30px; background:var(--green); border-radius:8px;
+  display:flex; align-items:center; justify-content:center; font-size:15px;
+}
+
+/* ─── Match context bar (shown when match is selected) ─── */
+.fs-match-bar {
+  display:flex; align-items:center; gap:6px; flex-wrap:wrap;
+  background:var(--bg-active); border:1px solid rgba(37,112,212,0.2);
+  border-radius:var(--radius-md); padding:8px 14px; margin-bottom:14px;
+  font-size:13px;
+}
+.fs-match-bar .home { color:var(--blue); font-weight:700; }
+.fs-match-bar .away { color:var(--red);  font-weight:700; }
+.fs-match-bar .vs   { color:var(--text-muted); padding:0 4px; font-size:11px; }
+.fs-match-bar .meta { color:var(--text-muted); font-size:11px; margin-left:4px; }
+
+/* ─── Result pills (last X results) ─── */
+.fs-result-pills {
+  display:flex; flex-wrap:wrap; gap:5px; margin-top:8px;
+}
+.rp {
+  min-width:28px; height:28px; border-radius:6px;
+  display:flex; align-items:center; justify-content:center;
+  font-size:12px; font-weight:700; color:#fff;
+  padding:0 5px;
+}
+
+/* ─── Category stats header ─── */
+.fs-cat-label {
+  font-size:10px; font-weight:700; text-transform:uppercase; letter-spacing:0.6px;
+  color:var(--text-muted); margin-bottom:6px;
+}
+
+/* ─── Compact comparison cell ─── */
+.fs-compare-cell {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-radius:var(--radius-md); padding:14px 12px;
+  text-align:center; margin-bottom:10px;
+}
+.fs-compare-cell .cc-val { font-size:24px; font-weight:700; line-height:1; }
+.fs-compare-cell .cc-lbl { font-size:10px; color:var(--text-muted); margin-top:4px; text-transform:uppercase; letter-spacing:0.4px; }
+.fs-compare-cell .cc-delta { font-size:12px; font-weight:600; margin-top:6px; }
+.fs-compare-cell .cc-delta.pos { color:var(--green); }
+.fs-compare-cell .cc-delta.neg { color:var(--red); }
+.fs-compare-cell .cc-delta.neu { color:var(--text-muted); }
+
+/* ─── Date filter bar ─── */
+.fs-date-filter {
+  display:flex; align-items:center; justify-content:flex-end; gap:10px;
+  margin-bottom:8px;
+}
+.fs-date-filter label {
+  font-size:11px; font-weight:600; text-transform:uppercase;
+  letter-spacing:0.5px; color:var(--text-muted);
+}
+</style>"""
+
+
+def inject_css() -> None:
+    st.markdown(_CSS, unsafe_allow_html=True)

src/src/utils.py

@@ -0,0 +1,23 @@
+import pandas as pd
+
+
+def sec(icon: str, title: str) -> None:
+    import streamlit as st
+    st.markdown(f'<div class="fs-sec"><span>{icon}</span> {title}</div>', unsafe_allow_html=True)
+
+
+def med(df: pd.DataFrame, col: str) -> float:
+    s = pd.to_numeric(df[col], errors="coerce").dropna()
+    if s.empty:
+        return 0.0
+    mean, std = s.mean(), s.std()
+    if std == 0 or (mean != 0 and std / abs(mean) < 0.5):
+        return round(mean, 2)
+    return round(s.median(), 2)
+
+
+def get_pred(df, stat: str) -> float:
+    if df is None or df.empty:
+        return 0.0
+    r = df[df["Estadística"] == stat]["Predicción RF"].values
+    return float(r[0]) if len(r) > 0 else 0.0

src/streamlit_app.py

@@ -1,40 +1,278 @@
-import altair as alt
-import numpy as np
-import pandas as pd
 import streamlit as st
+import pandas as pd
+import numpy as np
+
+from src.config import (
+    URL_FIXTURES, PAGE_TITLE, PAGE_LAYOUT,
+    RF_N_ROLLING_DEFAULT, RF_N_ROLLING_MIN, RF_N_ROLLING_MAX,
+    RF_TARGETS_LOCAL, RF_TARGETS_VISIT,
+    TIPOS_APUESTA,
+)
+from src.data_loader import cargar_fixtures, cargar_historico, ultimos_n
+from src.models import entrenar_rf_equipo
+from src.betting import calc_probs, kelly
+from src.styles import inject_css
+from src.utils import med
+
+st.set_page_config(page_title=PAGE_TITLE, layout=PAGE_LAYOUT)
+inject_css()
+
+# ════════════════════════════════════════════════════════════════════════════
+# NAVEGACIÓN — debe definirse ANTES de renderizar cualquier st.page_link()
+# ════════════════════════════════════════════════════════════════════════════
+_p_inicio    = st.Page("pages/inicio.py",        title="Inicio",         icon="🏠", default=True)
+_p_stats     = st.Page("pages/estadisticas.py",  title="Estadísticas",   icon="📊")
+_p_comp      = st.Page("pages/comparacion.py",   title="Comparación",    icon="⚔️")
+_p_rf        = st.Page("pages/prediccion_rf.py", title="Predicción RF",  icon="🤖")
+_p_kelly     = st.Page("pages/cuotas_kelly.py",  title="Cuotas & Kelly", icon="💰")
+_p_ia        = st.Page("pages/analisis_ia.py",   title="Análisis IA",    icon="🧠")
+
+pg = st.navigation(
+    [_p_inicio, _p_stats, _p_comp, _p_rf, _p_kelly, _p_ia],
+    position="hidden",
+)
+
+
+def _to_utc6(t) -> str:
+    try:
+        h, m = map(int, str(t).split(":"))
+        total = (h * 60 + m - 360) % (24 * 60)
+        return f"{total // 60:02d}:{total % 60:02d}"
+    except Exception:
+        return ""
+
+
+# ── Carga inicial ─────────────────────────────────────────────────────────────
+df_proximos = cargar_fixtures(URL_FIXTURES)
+if df_proximos is None:
+    st.error("No se pudo cargar la información de fixtures.")
+    st.stop()
+
+df_proximos["Date"] = pd.to_datetime(df_proximos["Date"], dayfirst=True, errors="coerce")
+df_proximos = df_proximos.dropna(subset=["Date"])
+if "Time" in df_proximos.columns:
+    df_proximos["Time_local"] = df_proximos["Time"].apply(_to_utc6)
+st.session_state["_df_proximos"] = df_proximos
+st.session_state["_p_estadisticas"] = _p_stats
+
+# ════════════════════════════════════════════════════════════════════════════
+# TOPBAR
+# ════════════════════════════════════════════════════════════════════════════
+_col_logo, _col_nav = st.columns([2, 8])
+
+with _col_logo:
+    st.markdown(
+        '<div class="fs-topbar-logo"><div class="tl-icon">⚽</div> FutStats</div>',
+        unsafe_allow_html=True,
+    )
+
+with _col_nav:
+    _nc = st.columns(6)
+    with _nc[0]: st.page_link(_p_inicio, label="Inicio",         icon="🏠")
+    with _nc[1]: st.page_link(_p_stats,  label="Estadísticas",   icon="📊")
+    with _nc[2]: st.page_link(_p_comp,   label="Comparación",    icon="⚔️")
+    with _nc[3]: st.page_link(_p_rf,     label="Predicción RF",  icon="🤖")
+    with _nc[4]: st.page_link(_p_kelly,  label="Cuotas & Kelly", icon="💰")
+    with _nc[5]: st.page_link(_p_ia,     label="Análisis IA",    icon="🧠")
+
+st.markdown("<hr style='margin:4px 0 16px;border:none;border-top:1px solid var(--border)'>", unsafe_allow_html=True)
+
+# ════════════════════════════════════════════════════════════════════════════
+# MATCH SELECCIONADO
+# ════════════════════════════════════════════════════════════════════════════
+_sel = st.session_state.get("_selected_match")
+
+df_hist = dl_all = dv_all = None
+liga = eq_l = eq_v = hora_str = ""
+fecha_sel = None
+nl = nv = n_rf = 0
+ventaja_pct = 0
+ventaja_equipo = None
+el = ev = "Todos"
+_has_match = False
+
+if _sel is not None:
+    liga      = _sel.get("Div", "")
+    eq_l      = _sel.get("HomeTeam", "")
+    eq_v      = _sel.get("AwayTeam", "")
+    hora_str  = _sel.get("Time_local", _sel.get("Time", ""))
+    fecha_raw = _sel.get("Date")
+    fecha_sel = pd.to_datetime(fecha_raw).date() if fecha_raw else None
+
+    hora_main = f" · {hora_str}" if hora_str else ""
+    fecha_str = fecha_sel.strftime("%d %b %Y") if fecha_sel else ""
+
+    st.markdown(
+        f'<div class="fs-match-bar">'
+        f'<span class="home">{eq_l}</span>'
+        f'<span class="vs">vs</span>'
+        f'<span class="away">{eq_v}</span>'
+        f'<span class="meta">&#x1F3C6; {liga}'
+        f'{" &middot; " + fecha_str if fecha_str else ""}'
+        f'{" &middot; " + hora_str if hora_str else ""}'
+        f'</span></div>',
+        unsafe_allow_html=True,
+    )
+
+    df_hist = cargar_historico(liga)
+    if df_hist is None or df_hist.empty:
+        st.error(f"Sin datos históricos para **{liga}**.")
+        st.stop()
+
+    dl_all = df_hist[df_hist["HomeTeam"] == eq_l]
+    dv_all = df_hist[df_hist["AwayTeam"] == eq_v]
+
+    _match_id = f"{liga}:{eq_l}:{eq_v}"
+    if st.session_state.get("_match_id") != _match_id:
+        st.session_state.pop("n_both", None)
+        st.session_state["_match_id"] = _match_id
+
+    # ── Configuración específica por página ───────────────────────────────
+    _current_page = st.session_state.get("_current_page", "")
+
+    n_max  = min(len(dl_all), len(dv_all), 20)
+    _opts  = [0] + list(range(3, n_max + 1))
+    _def_n = 5
+    _n_stored = st.session_state.get("n_both")
+    _n_idx = (
+        _opts.index(_n_stored) if _n_stored in _opts else
+        (_opts.index(_def_n)   if _def_n   in _opts else
+         (1 if len(_opts) > 1 else 0))
+    )
+    _n_default = _opts[_n_idx]
+
+    # Defaults — each branch only overrides what it owns
+    nl = nv        = _n_default
+    n_rf           = st.session_state.get("nrf",            RF_N_ROLLING_DEFAULT)
+    ventaja_pct    = st.session_state.get("ventaja_ia",     0)
+    ventaja_equipo = st.session_state.get("ventaja_equipo", None)
+    bet_type       = st.session_state.get("_bet_type",      TIPOS_APUESTA[0])
+
+    if _current_page == "comparacion":
+        _cc, _ = st.columns([3, 7])
+        with _cc:
+            nl = nv = st.selectbox(
+                f"📊 Últimos partidos  ·  🏠 {len(dl_all)} · ✈️ {len(dv_all)} disp.",
+                _opts, index=_n_idx,
+                format_func=lambda x: "Todos" if x == 0 else f"Últimos {x}",
+                key="n_both",
+            )
+
+    elif _current_page == "analisis_ia":
+        _cv1, _cv2, _ = st.columns([2, 2, 6])
+        with _cv1:
+            _ventaja_opts = [0, 5, 10, 15, 20]
+            _ventaja_stored = st.session_state.get("ventaja_ia", 0)
+            _ventaja_idx = _ventaja_opts.index(_ventaja_stored) if _ventaja_stored in _ventaja_opts else 0
+            ventaja_pct = st.selectbox(
+                "📈 Ventaja informacional:",
+                _ventaja_opts,
+                index=_ventaja_idx,
+                format_func=lambda x: "Sin ventaja" if x == 0 else f"+{x}%",
+                key="ventaja_ia",
+                help="Porcentaje de ventaja sobre las cuotas del mercado.",
+            )
+        ventaja_equipo = None
+        if ventaja_pct > 0:
+            with _cv2:
+                ventaja_equipo = st.radio(
+                    "Ventaja para:",
+                    [f"🏠 {eq_l}", f"✈️ {eq_v}"],
+                    horizontal=True,
+                    key="ventaja_equipo",
+                )
+    # estadisticas, prediccion_rf, cuotas_kelly, inicio → all use defaults
+
+    _has_match = True
+
+# ════════════════════════════════════════════════════════════════════════════
+# DATOS COMPARTIDOS
+# ════════════════════════════════════════════════════════════════════════════
+if _has_match:
+    dl = ultimos_n(dl_all, nl)
+    dv = ultimos_n(dv_all, nv)
+    _periodo = "Todos" if nl == 0 else f"Últimos {nl}"
+    el = ev = _periodo
+
+    if _current_page == "prediccion_rf":
+        with st.spinner("Entrenando modelos RF..."):
+            rf_l = entrenar_rf_equipo(df_hist, eq_l, eq_v, n_rf, RF_TARGETS_LOCAL)
+            rf_v = entrenar_rf_equipo(df_hist, eq_l, eq_v, n_rf, RF_TARGETS_VISIT)
+    else:
+        rf_l = rf_v = None
+
+    cH = pd.to_numeric(_sel.get("B365H", np.nan), errors="coerce")
+    cD = pd.to_numeric(_sel.get("B365D", np.nan), errors="coerce")
+    cA = pd.to_numeric(_sel.get("B365A", np.nan), errors="coerce")
+    if np.isnan(cH) and "B365H" in dl.columns:
+        cH = pd.to_numeric(dl["B365H"], errors="coerce").mean()
+    if np.isnan(cD) and "B365D" in dl.columns:
+        cD = pd.to_numeric(dl["B365D"], errors="coerce").mean()
+    if np.isnan(cA) and "B365A" in dl.columns:
+        cA = pd.to_numeric(dl["B365A"], errors="coerce").mean()
+
+    probs    = calc_probs(df_hist, eq_l, eq_v)
+    kelly_ok = probs is not None and not any(np.isnan([cH, cD, cA]))
+
+    if kelly_ok:
+        kH = kelly(probs["H"], cH)
+        kD = kelly(probs["D"], cD)
+        kA = kelly(probs["A"], cA)
+        veH = (probs["H"] * cH - 1) * 100
+        veD = (probs["D"] * cD - 1) * 100
+        veA = (probs["A"] * cA - 1) * 100
+        pH, pD, pA = probs["H"] * 100, probs["D"] * 100, probs["A"] * 100
+    else:
+        kH = kD = kA = veH = veD = veA = pH = pD = pA = 0.0
+
+    l_anota  = med(dl, "FTHG") if not dl.empty else 0.0
+    v_recibe = med(dv, "FTHG") if not dv.empty else 0.0
+    v_anota  = med(dv, "FTAG") if not dv.empty else 0.0
+    l_recibe = med(dl, "FTAG") if not dl.empty else 0.0
+
+    bet_type = st.session_state.get("_bet_type", TIPOS_APUESTA[0])
+
+    st.session_state["_ctx"] = {
+        "liga": liga, "eq_l": eq_l, "eq_v": eq_v,
+        "fecha_sel": fecha_sel, "hora_str": hora_str,
+        "el": el, "ev": ev,
+        "dl": dl, "dv": dv, "df_hist": df_hist,
+        "rf_l": rf_l, "rf_v": rf_v,
+        "probs": probs, "kelly_ok": kelly_ok,
+        "cH": cH, "cD": cD, "cA": cA,
+        "kH": kH, "kD": kD, "kA": kA,
+        "veH": veH, "veD": veD, "veA": veA,
+        "pH": pH, "pD": pD, "pA": pA,
+        "l_anota": l_anota, "v_recibe": v_recibe,
+        "v_anota": v_anota, "l_recibe": l_recibe,
+        "ventaja_pct": ventaja_pct, "ventaja_equipo": ventaja_equipo,
+        "bet_type": bet_type,
+    }
+
+    hora_main      = f" &middot; {hora_str}" if hora_str else ""
+    fecha_str_main = fecha_sel.strftime("%d %b %Y") if fecha_sel else ""
+    st.markdown(f"""<div class="fs-header">
+  <div>
+    <h1>
+      <span style="color:var(--blue)">{eq_l}</span>
+      <span style="color:var(--text-muted);font-weight:400;font-size:18px;padding:0 12px">vs</span>
+      <span style="color:var(--red)">{eq_v}</span>
+    </h1>
+    <div class="sub">&#x1F3C6; {liga} &middot; &#x1F4C5; {fecha_str_main}{hora_main}</div>
+  </div>
+</div>""", unsafe_allow_html=True)
+
+    st.markdown(f"""<div class="fs-stats-bar">
+  <div class="si"><div class="sn">{len(dl)}</div><div class="sl">&#x1F3E0; {eq_l} local</div></div>
+  <div class="si"><div class="sn blue">{len(dv)}</div><div class="sl">&#x2708;&#xFE0F; {eq_v} visit.</div></div>
+  <div class="si"><div class="sn red">{len(df_hist)}</div><div class="sl">Histórico {liga}</div></div>
+  <div class="si"><div class="sn yellow">{len(df_proximos)}</div><div class="sl">Fixtures disponibles</div></div>
+</div>""", unsafe_allow_html=True)
+
+else:
+    st.session_state.pop("_ctx", None)
 
-"""
-# Welcome to Streamlit!
-
-Edit `/streamlit_app.py` to customize this app to your heart's desire :heart:.
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
-
-In the meantime, below is an example of what you can do with just a few lines of code:
-"""
-
-num_points = st.slider("Number of points in spiral", 1, 10000, 1100)
-num_turns = st.slider("Number of turns in spiral", 1, 300, 31)
-
-indices = np.linspace(0, 1, num_points)
-theta = 2 * np.pi * num_turns * indices
-radius = indices
-
-x = radius * np.cos(theta)
-y = radius * np.sin(theta)
-
-df = pd.DataFrame({
-    "x": x,
-    "y": y,
-    "idx": indices,
-    "rand": np.random.randn(num_points),
-})
-
-st.altair_chart(alt.Chart(df, height=700, width=700)
-    .mark_point(filled=True)
-    .encode(
-        x=alt.X("x", axis=None),
-        y=alt.Y("y", axis=None),
-        color=alt.Color("idx", legend=None, scale=alt.Scale()),
-        size=alt.Size("rand", legend=None, scale=alt.Scale(range=[1, 150])),
-    ))
+# ════════════════════════════════════════════════════════════════════════════
+# EJECUTAR PÁGINA ACTIVA
+# ════════════════════════════════════════════════════════════════════════════
+pg.run()