Update README.md

README.md

@@ -1,3 +1,106 @@
 ---
 license: cc-by-4.0
+language:
+  - en
+pretty_name: Bookmaker Margin Panel
+size_categories:
+  - 1K<n<10K
+task_categories:
+  - tabular-regression
+  - tabular-classification
+tags:
+  - sports-betting
+  - bookmaker-margin
+  - overround
+  - gambling-economics
+  - panel-data
+  - tabular
+configs:
+  - config_name: default
+    data_files:
+      - betbetter_margin_panel.csv
+      - betbetter_prop_margin_panel.csv
 ---
+
+# Bookmaker Margin Panel — daily measured betting margins across 22 leagues and 3 market types
+
+A daily panel of **measured bookmaker margins (overround/vigorish)** across head-to-head, spread
+and totals markets: one row per day × league × region × market type, aggregated from over
+4 million individually recorded odds observations (2.7 million cleaned markets).
+
+To our knowledge this is the only openly published bookmaker margin panel. It quantifies the
+*price* of sports betting — the amount by which a market's implied probabilities exceed 100% —
+across operators, sports and time.
+
+| | |
+|---|---|
+| **Files** | Main panel: 3,494 cells (day × league × region × market type × time-to-start) · Props panel: 1,148 cells |
+| **Underlying observations** | 2,002,681 cleaned main markets + 699,733 two-sided player-prop markets |
+| **Bookmakers aggregated** | 52 (not identified individually — see Limitations; exchanges excluded) |
+| **Leagues** | 22 competition keys across 15 sports, including 9 soccer leagues |
+| **Market types** | head-to-head (2- and 3-way), spread, total |
+| **Period** | 29 June – 2 August 2026, 35 days (collection continues; quarterly updates planned) |
+| **Licence** | CC BY 4.0 |
+
+## Columns
+
+| Column | Meaning |
+|---|---|
+| `snap_date` | Observation date (UTC) |
+| `sport` | League/competition key, e.g. `baseball_mlb`, `soccer_epl` |
+| `region` | `au`, `us` or `offshore_intl` — the licensing region of the operators aggregated in the cell |
+| `market_type` | `h2h`, `spread` or `total` (main panel); specific prop market key in the props file |
+| `time_to_start` | Hours from observation to scheduled start, bucketed (under 6h to over 7 days) — main panel |
+| `outcomes` | Number of priced outcomes (2, or 3 for soccer-style h2h with the draw) |
+| `mean_margin_pct` | Mean margin that day: sum of the outcomes' implied probabilities minus 100, in percent |
+| `n_markets` | Number of distinct market observations in the cell (cells with fewer than 10 are excluded) |
+| `n_bookmakers` | Number of distinct operators aggregated in the cell |
+
+## Method
+
+For every pre-match market observation, all priced outcomes were converted to implied
+probabilities (1 ÷ decimal odds) and summed; the excess over 100% is the margin. Two-way and
+three-way markets are kept in separate cells because their margins are not comparable. Excluded:
+in-play observations, implausible margins (below 0%, or above 20% two-way / 35% three-way —
+suspended or near-settled prices), and betting exchanges.
+Full methodology: https://betbetter.world/studies/bookmaker-margins
+
+## Headline findings from this panel
+
+- **Player props average a 7.93% margin — 66% more than main head-to-head markets (4.77%)**.
+  To our knowledge this is the first public measurement of prop-market margins (699,733
+  two-sided prop markets, MLB/NHL/soccer/basketball prop families).
+- The market-type price ladder: h2h **4.77%** &rarr; spreads **4.90%** &rarr; totals **5.45%**
+  &rarr; three-way soccer h2h **7.67%** &rarr; player props **7.93%**.
+- **Margin has a term structure**: it peaks 1&ndash;3 days before the start (5.23%) and
+  compresses to 4.71% in the final six hours — betting early costs measurably more.
+- **Australian-licensed bookmakers average 5.25%, US-licensed 4.72%, offshore/international
+  4.55%** — Australians pay roughly 11% more margin than Americans for equivalent bets.
+- Individual operators are not identified in these files; operator-level statistics (a 3.16% to
+  5.78% spread between the cheapest and dearest bookmaker) are published in the accompanying
+  study: https://betbetter.world/studies/bookmaker-margins
+
+## Limitations
+
+One month of collection (a longer panel accumulates with each planned update). Individual
+bookmakers are deliberately not identified in the files: cells aggregate across all operators in
+a region (n_bookmakers reports how many), so no operator's prices or price patterns are
+recoverable — the files are irreversible statistics, not a data feed in any form. Head-to-head is
+the tightest market type bookmakers offer; margins on multis and player props are materially
+higher, so these figures are a floor for the cost of betting generally. No individual odds,
+prices or market observations are included — both files are aggregate statistics derived from a
+licensed feed that does not permit redistribution of the underlying prices.
+
+## Licence and citation
+
+CC BY 4.0 — free to use, including commercially, with attribution.
+
+> Bet Better (2026). *Bookmaker Margin Panel: daily measured betting margins across 22 leagues,
+> three market types and player-prop markets.* https://betbetter.world/studies/bookmaker-margins
+
+Maintained by [Bet Better](https://betbetter.world), which also publishes a
+[free sports model API](https://betbetter.world/api/) and an
+[open AFL dataset](https://doi.org/10.5281/zenodo.21612783).
+
+> This dataset contains no odds, no prices and no wagering data — aggregate statistics only. 18+.
+> Most people lose money gambling. If gambling is causing you harm: gamblinghelponline.org.au.