Initial commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+xgb_model.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,364 @@
+---
+license: cc-by-nc-4.0
+library_name: xgboost
+tags:
+- finance
+- stock-market
+- indian-stock-market
+- nse
+- bse
+- tabular
+- time-series
+- xgboost
+- classification
+- indian-market
+---
+
+# XGBoost — Indian Stock Market Prediction
+
+XGBoost binary classification model trained on historical Indian equity-market data to predict whether a stock reaches a specified positive-return threshold within a **5-trading-day horizon**.
+
+The current model targets a **+3.0% return threshold** and was trained across thousands of Indian equities using technical, market, cross-sectional, and macroeconomic features.
+
+> **Research status:** This repository contains a research model and its out-of-sample evaluation artifacts. The reported results should not be interpreted as a guarantee of future market performance or profitability.
+
+---
+
+## Model Overview
+
+| Property                | Value                                |
+| ----------------------- | ------------------------------------ |
+| Model                   | XGBoost                              |
+| Task                    | Binary classification                |
+| Prediction target       | ≥ +3.0% return within 5 trading days |
+| Dataset size            | ~6.92 million rows                   |
+| Symbols                 | 2,656                                |
+| Historical period       | 2008-09-02 → 2026-06-30              |
+| Features                | 221                                  |
+| Positive class          | 36.8%                                |
+| Negative class          | 63.2%                                |
+| `scale_pos_weight`      | 1.49                                 |
+| Training method         | Histogram-based XGBoost              |
+| Hardware                | CUDA GPU                             |
+| Maximum boosting rounds | 2,000                                |
+| Learning rate           | 0.02                                 |
+| Maximum tree depth      | 6                                    |
+| Early stopping          | 75 rounds                            |
+
+---
+
+# Dataset
+
+The training dataset contains approximately **6.92 million observations** covering **2,656 Indian stock symbols** from September 2008 through June 2026.
+
+The model uses 221 input features derived from historical market data and broader market conditions.
+
+The feature groups include:
+
+* Price returns
+* Momentum
+* Trend indicators
+* Volatility
+* Technical indicators
+* Relative strength
+* Market breadth
+* Cross-sectional statistics
+* Nifty relationships
+* Gold relationships
+* Brent crude relationships
+* USD/INR relationships
+* US-market indicators
+
+The target is a binary label representing whether the specified positive-return threshold is reached within the prediction horizon.
+
+---
+
+# Temporal Dataset Split
+
+The dataset was divided chronologically rather than randomly.
+
+| Split             | Period                  | Approx. Share |
+| ----------------- | ----------------------- | ------------: |
+| Training          | 2008-09-02 → 2021-02-15 |           70% |
+| Validation        | 2021-02-16 → 2024-08-30 |           20% |
+| Walk-forward test | 2024-09-01 → 2026-06-30 |           10% |
+
+This temporal separation is intended to reduce leakage from randomly mixing observations from different points in time.
+
+The final test period was not used for model fitting.
+
+---
+
+# Validation Performance
+
+The model achieved the following results on the held-out validation period:
+
+**2021-02-16 → 2024-08-30**
+
+| Metric                             |      Score |
+| ---------------------------------- | ---------: |
+| ROC-AUC                            | **0.9472** |
+| PR-AUC                             | **0.9042** |
+| Precision                          | **0.7981** |
+| Recall                             | **0.8364** |
+| Information Coefficient (Spearman) | **0.7140** |
+| Brier Score                        | **0.0835** |
+| Brier Skill Score                  | **0.6068** |
+| Mean Calibration Error             | **0.0749** |
+
+Validation observations:
+
+**1,698,058 rows**
+
+The validation results indicate strong separation between the positive and negative classes on this historical period.
+
+However, validation performance alone should not be treated as evidence of future profitability.
+
+---
+
+# Validation Decile Analysis
+
+Predictions were divided into ten probability-ranked groups.
+
+| Decile | Hit Rate | Samples |
+| -----: | -------: | ------: |
+|    D10 |    98.4% | 169,806 |
+|     D9 |    88.8% | 169,806 |
+|     D8 |    60.5% | 169,806 |
+|     D7 |    31.1% | 169,805 |
+|     D6 |    14.6% | 169,806 |
+|     D5 |     6.6% | 169,806 |
+|     D4 |     3.1% | 169,805 |
+|     D3 |     1.7% | 169,806 |
+|     D2 |     0.9% | 169,806 |
+|     D1 |     0.4% | 169,806 |
+
+The probability ranking shows strong separation across the validation sample, with substantially different observed positive rates between the lowest- and highest-ranked groups.
+
+These figures describe historical classification performance and do not account for transaction costs, slippage, liquidity constraints, portfolio construction, or market impact.
+
+---
+
+# Out-of-Sample Walk-Forward Evaluation
+
+The model was subsequently evaluated chronologically over the final portion of the dataset.
+
+The test period was divided into four chronological folds of approximately six months each.
+
+| Fold | Period                  |     PR-AUC |         IC |  Brier | Samples |
+| ---- | ----------------------- | ---------: | ---------: | -----: | ------: |
+| 1    | 2024-09-01 → 2025-02-28 | **0.9336** | **0.7619** | 0.0850 | 262,062 |
+| 2    | 2025-03-01 → 2025-08-31 | **0.8987** | **0.7044** | 0.0829 | 266,786 |
+| 3    | 2025-09-01 → 2026-02-28 | **0.8921** | **0.7011** | 0.0845 | 282,585 |
+| 4    | 2026-03-01 → 2026-06-30 | **0.8962** | **0.7157** | 0.0925 | 180,837 |
+
+### Walk-Forward Summary
+
+| Statistic        |      Value |
+| ---------------- | ---------: |
+| Mean PR-AUC      | **0.9052** |
+| Mean IC          | **0.7208** |
+| Mean Brier Score | **0.0862** |
+| Minimum IC       | **0.7011** |
+| Maximum IC       | **0.7619** |
+
+The model maintained relatively strong classification performance across all four chronological folds.
+
+Performance did, however, vary between periods. The difference between the highest and lowest observed IC is approximately **0.061**.
+
+This suggests that model performance is not completely invariant across market regimes and warrants additional robustness testing.
+
+---
+
+# Overfitting & Generalization
+
+The model was evaluated using a chronological validation set followed by a later out-of-sample test period.
+
+The results provide evidence that the model retains predictive separation outside its training period.
+
+However, the current experiment does **not** establish that the model is free from overfitting.
+
+In particular:
+
+* Validation performance is very strong.
+* Walk-forward test performance remains strong.
+* Performance varies between chronological folds.
+* The model reached the maximum configured **2,000 boosting rounds**, with the best iteration at **1,999**.
+* Early stopping therefore did not activate before the configured maximum number of rounds.
+
+The last point means that additional experiments with a larger maximum number of boosting rounds should be evaluated carefully rather than assuming that more trees will improve generalization.
+
+Future experiments should compare performance across additional temporal folds and assess whether increasing model complexity improves out-of-sample performance or simply improves the validation period.
+
+---
+
+# Calibration
+
+The validation set produced:
+
+* Brier Score: **0.0835**
+* Brier Skill Score: **0.6068**
+* Mean Calibration Error: **0.0749**
+
+The model therefore produces probability scores that contain useful information, but the raw probabilities should not automatically be interpreted as perfectly calibrated real-world probabilities.
+
+Probability calibration should be evaluated separately if the model's output is to be interpreted probabilistically.
+
+---
+
+# Feature Set
+
+The model uses 221 features.
+
+Examples of macro and cross-market features include:
+
+```text
+usd_inr_sma20_ratio
+usd_inr_sma200_ratio
+usd_inr_momentum_10
+usd_inr_momentum_20
+usd_inr_volatility_20d
+usd_inr_trend
+usd_inr_zscore
+usd_inr_vol_zscore
+usd_inr_high_vol
+usd_inr_appreciation
+
+us_ret_1d
+us_ret_5d
+us_ret_20d
+us_ret_60d
+
+us_breadth_1d
+us_breadth_5d
+us_breadth_20d
+
+us_dispersion_1d
+us_dispersion_5d
+us_avg_vol_ratio
+us_pct_above_sma20
+us_avg_rsi
+us_avg_dist_52w_high
+
+corr_nifty_20d
+corr_nifty_60d
+beta_nifty_60d
+is_high_beta
+is_low_beta
+corr_breakdown
+
+rel_strength_5d
+rel_strength_20d
+
+corr_gold_60d
+corr_gold_rising
+
+corr_brent_60d
+corr_brent_rising
+
+corr_usd_inr_60d
+corr_usd_inr_rising
+
+corr_us_60d
+```
+
+The complete feature schema is available in `model_metadata.json`.
+
+---
+
+# Model Artifacts
+
+The repository contains the following artifacts:
+
+| File                          | Description                                                                       |
+| ----------------------------- | --------------------------------------------------------------------------------- |
+| `xgb_model.json`              | Trained XGBoost model                                                             |
+| `model_metadata.json`         | Training configuration, feature schema, split information and evaluation metadata |
+| `feature_importance.csv`      | XGBoost feature-importance statistics                                             |
+| `val_predictions.parquet`     | Validation predictions                                                            |
+| `test_wf_predictions.parquet` | Chronological walk-forward test predictions                                       |
+
+The prediction files are provided to make the reported evaluation results independently inspectable and reproducible.
+
+---
+
+# Reproducibility
+
+The model metadata records:
+
+* Feature names
+* Feature count
+* Dataset information
+* Training period
+* Validation period
+* Model parameters
+* Class weighting
+* Best iteration
+* Evaluation metrics
+* Training timestamp
+
+The model uses a fixed random seed of `42`.
+
+---
+
+# Important Limitations
+
+This model should be considered a **research artifact**, not a guaranteed trading system.
+
+The reported classification metrics do not directly measure:
+
+* Portfolio returns
+* Sharpe ratio
+* Maximum drawdown
+* Transaction costs
+* Brokerage
+* Taxes
+* Slippage
+* Bid/ask spread
+* Market impact
+* Position sizing
+* Portfolio concentration
+* Liquidity constraints
+* Capacity
+* Execution latency
+
+A model can achieve strong classification metrics while producing poor investment returns after these factors are considered.
+
+The target is also a classification threshold rather than a direct optimization of portfolio returns.
+
+Additional research is therefore required before using the model for live decision-making.
+
+---
+
+# Current Research Status
+
+The current experiment establishes a useful baseline for the **+3.0% / 5-trading-day** prediction task.
+
+The next stages of research should include:
+
+1. Additional temporal walk-forward experiments.
+2. Comparison of model complexity and boosting-round counts.
+3. Testing alternative return thresholds and horizons.
+4. Feature ablation and importance stability analysis.
+5. Leakage and feature-timing audits.
+6. Performance analysis by market regime.
+7. Calibration analysis on completely unseen periods.
+8. Portfolio-level backtesting.
+9. Transaction-cost and slippage modelling.
+10. Paper-trading validation before considering live deployment.
+
+---
+
+# License
+
+This model is published under the **Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)** license.
+
+The license permits non-commercial use subject to the terms of the license.
+
+Commercial use requires appropriate licensing.
+
+---
+
+# Author
+
+[**Sayantan Basu**](https://sayantan-basu.vercel.app)

feature_importance.csv

@@ -0,0 +1,216 @@
+feature,gain,cover,weight
+f1,5325.15283203125,68459.6328125,4698.0
+f7,4115.845703125,68196.3984375,1120.0
+f31,3964.971923828125,84257.1875,2079.0
+f36,3366.510986328125,47373.79296875,1538.0
+f212,1568.135498046875,35672.390625,1186.0
+f2,1246.1517333984375,47140.26953125,4120.0
+f8,1101.158447265625,44323.390625,947.0
+f17,931.3317260742188,40531.37890625,1520.0
+f32,862.3289184570312,37222.3828125,713.0
+f0,711.5985717773438,31164.341796875,5577.0
+f39,663.1898803710938,39934.984375,1691.0
+f38,642.7109375,23749.849609375,2272.0
+f6,605.0057983398438,29368.93359375,1538.0
+f40,541.1348876953125,39109.171875,2041.0
+f54,525.8748779296875,40191.59375,2955.0
+f69,480.3856506347656,27513.529296875,5197.0
+f46,454.45880126953125,34922.32421875,1173.0
+f41,445.4704284667969,86219.9296875,1134.0
+f61,413.2144775390625,64463.95703125,245.0
+f76,397.5409240722656,63324.75390625,710.0
+f34,336.4801025390625,36437.12109375,1087.0
+f45,310.4357604980469,65585.1015625,1011.0
+f115,282.82122802734375,25716.794921875,390.0
+f44,271.93890380859375,103531.15625,2181.0
+f210,262.3323974609375,24503.73046875,227.0
+f16,261.62164306640625,137480.15625,1655.0
+f58,258.2943420410156,21659.6640625,861.0
+f33,230.97845458984375,69676.015625,1033.0
+f53,209.64468383789062,35103.75390625,2001.0
+f71,203.9341583251953,55411.87890625,253.0
+f75,200.1075897216797,34342.58984375,160.0
+f21,193.40286254882812,55620.0859375,688.0
+f193,179.60198974609375,52923.33203125,742.0
+f59,173.84181213378906,23193.08203125,1193.0
+f70,172.29080200195312,48352.68359375,386.0
+f72,171.29737854003906,34841.23046875,617.0
+f147,168.84896850585938,26952.25390625,422.0
+f208,167.88539123535156,11878.2275390625,726.0
+f51,159.6352996826172,80163.8828125,1763.0
+f18,158.70263671875,34064.16015625,389.0
+f124,152.9352264404297,46631.21875,732.0
+f171,151.54481506347656,29977.30859375,593.0
+f126,151.440185546875,48580.6015625,528.0
+f47,150.7949981689453,108855.953125,717.0
+f207,143.04129028320312,18832.865234375,889.0
+f56,140.55490112304688,73637.28125,1255.0
+f104,140.51597595214844,26177.23828125,748.0
+f144,139.98495483398438,63243.3828125,780.0
+f101,137.34207153320312,25784.041015625,585.0
+f55,134.55760192871094,23258.4375,1728.0
+f187,133.2763214111328,37959.5546875,572.0
+f121,130.02809143066406,37742.1328125,497.0
+f91,129.34507751464844,13051.0419921875,12.0
+f202,126.53907775878906,73390.328125,604.0
+f133,126.4417953491211,11295.861328125,30.0
+f117,124.9978256225586,27211.66015625,314.0
+f162,122.44957733154297,51836.9453125,457.0
+f94,119.9894027709961,41162.88671875,278.0
+f3,119.62555694580078,33655.30859375,1124.0
+f122,117.14742279052734,48230.12890625,439.0
+f160,116.41017150878906,23336.771484375,477.0
+f89,116.15287780761719,22684.0,201.0
+f118,115.3116455078125,46612.140625,366.0
+f204,114.83635711669922,30698.259765625,504.0
+f116,114.74333953857422,33482.84375,204.0
+f195,110.9507064819336,35621.69140625,310.0
+f189,110.81322479248047,27497.9765625,340.0
+f113,109.75797271728516,47348.70703125,410.0
+f62,108.93269348144531,11772.4384765625,92.0
+f77,108.42366790771484,15604.078125,93.0
+f9,107.12332153320312,34515.3203125,258.0
+f107,106.64046478271484,27238.91015625,421.0
+f190,105.9676513671875,55963.65625,512.0
+f209,105.63643646240234,18306.990234375,18.0
+f120,105.44498443603516,33585.53515625,151.0
+f100,105.1141586303711,30311.248046875,214.0
+f206,104.19023895263672,25035.544921875,838.0
+f60,104.06464385986328,52803.38671875,1001.0
+f174,103.52881622314453,50983.33203125,419.0
+f80,103.49224853515625,23656.63671875,331.0
+f114,103.38029479980469,34085.8203125,212.0
+f84,103.3167953491211,23147.216796875,189.0
+f74,103.16171264648438,14252.2353515625,274.0
+f177,102.72142028808594,54082.24609375,446.0
+f176,102.30437469482422,35788.20703125,507.0
+f99,100.70491027832031,24830.2890625,314.0
+f164,100.23468017578125,15831.5263671875,447.0
+f197,100.16085815429688,47805.38671875,528.0
+f95,99.98596954345703,70329.3125,308.0
+f105,99.74986267089844,17183.33203125,252.0
+f81,98.96772003173828,20864.2421875,174.0
+f145,98.55354309082031,34898.09765625,533.0
+f87,98.32896423339844,33892.73828125,254.0
+f88,97.87724304199219,33016.03515625,540.0
+f199,97.772216796875,36230.62890625,290.0
+f103,97.72587585449219,70409.0234375,597.0
+f37,97.67594909667969,25916.970703125,938.0
+f42,96.77252197265625,105556.59375,393.0
+f125,96.10755157470703,28639.498046875,506.0
+f93,96.08097839355469,19186.265625,195.0
+f185,95.93507385253906,42764.640625,357.0
+f73,95.1639175415039,12892.740234375,376.0
+f106,94.98686981201172,12041.0986328125,274.0
+f158,94.78166198730469,46203.0390625,332.0
+f148,94.68411254882812,17762.63671875,420.0
+f184,94.28591918945312,35923.99609375,374.0
+f200,94.15950012207031,22957.810546875,496.0
+f194,93.82621765136719,47455.30078125,369.0
+f159,92.949462890625,48539.140625,417.0
+f136,92.61298370361328,24385.689453125,130.0
+f169,92.01233673095703,26766.517578125,434.0
+f175,91.8438949584961,23843.4921875,24.0
+f112,91.71309661865234,21628.4375,332.0
+f109,91.03904724121094,42846.83203125,316.0
+f155,90.6822509765625,42451.6171875,531.0
+f85,90.68136596679688,16686.9296875,201.0
+f149,90.15928649902344,23711.46875,299.0
+f119,90.08594512939453,29264.630859375,330.0
+f150,89.9566879272461,31202.34765625,288.0
+f201,89.85118865966797,67483.9140625,443.0
+f96,89.83715057373047,34116.7421875,31.0
+f198,89.6642837524414,19943.99609375,366.0
+f178,89.26366424560547,20985.14453125,3.0
+f156,88.75489044189453,14283.2734375,479.0
+f196,88.1899185180664,34676.578125,357.0
+f181,87.97608184814453,16436.32421875,420.0
+f86,87.14678192138672,45750.26171875,344.0
+f102,86.903564453125,27558.8515625,101.0
+f137,86.66979217529297,16996.958984375,335.0
+f35,86.66337585449219,19229.359375,853.0
+f143,86.6239013671875,34069.375,430.0
+f79,86.37348175048828,14398.39453125,77.0
+f68,86.13903045654297,31897.810546875,26.0
+f50,85.54434967041016,16195.123046875,894.0
+f205,84.95002746582031,46978.07421875,315.0
+f170,84.76980590820312,22305.462890625,288.0
+f146,84.572021484375,29964.841796875,373.0
+f111,84.1866683959961,17010.32421875,172.0
+f22,83.09573364257812,38828.37890625,102.0
+f82,82.84314727783203,20371.6328125,214.0
+f168,82.41185760498047,18014.60546875,432.0
+f83,82.0250015258789,41562.01171875,254.0
+f135,81.65332794189453,22148.048828125,53.0
+f90,81.2519302368164,31081.951171875,58.0
+f180,81.17967987060547,36046.203125,576.0
+f203,80.814208984375,47314.7578125,206.0
+f172,80.73104858398438,36601.875,381.0
+f123,80.26925659179688,54063.109375,605.0
+f92,80.21176147460938,32615.755859375,91.0
+f78,79.47978210449219,15207.662109375,290.0
+f129,78.6848373413086,21271.150390625,441.0
+f4,78.42843627929688,27709.818359375,425.0
+f161,77.77777099609375,12504.84375,80.0
+f127,76.01110076904297,18625.470703125,475.0
+f182,75.88616943359375,23327.3515625,328.0
+f128,75.36719512939453,25060.37890625,522.0
+f157,75.3602523803711,17789.9921875,375.0
+f140,75.35004425048828,1243.0164794921875,7.0
+f98,75.3267822265625,57301.0234375,55.0
+f63,75.08702087402344,26990.11328125,42.0
+f183,74.89909362792969,46639.00390625,302.0
+f108,74.82830810546875,29040.123046875,351.0
+f179,74.61183166503906,2816.818603515625,11.0
+f19,74.30308532714844,43849.71484375,443.0
+f165,74.28292846679688,20069.68359375,395.0
+f132,74.1614761352539,15046.2412109375,35.0
+f186,74.01150512695312,33781.50390625,101.0
+f167,73.72079467773438,38586.2421875,632.0
+f43,71.59841918945312,106947.28125,660.0
+f25,70.65673828125,21007.630859375,497.0
+f173,69.72639465332031,15993.78125,91.0
+f110,66.60283660888672,14914.7578125,123.0
+f28,66.08335876464844,13683.634765625,100.0
+f130,65.740478515625,5876.8193359375,8.0
+f48,65.69597625732422,35550.25,338.0
+f213,65.62059783935547,16225.6953125,369.0
+f97,65.50997924804688,24648.734375,58.0
+f10,63.9237174987793,34336.09375,98.0
+f66,63.57122802734375,20479.8828125,103.0
+f49,60.27708435058594,65060.7421875,808.0
+f52,59.52656936645508,88758.9609375,575.0
+f26,59.4591064453125,16305.6259765625,350.0
+f23,58.429962158203125,61881.41796875,24.0
+f57,58.29280471801758,19692.81640625,565.0
+f217,54.119808197021484,54751.8984375,79.0
+f188,54.1160774230957,15875.8408203125,15.0
+f131,52.61231231689453,7133.201171875,9.0
+f13,52.04670333862305,9690.5986328125,342.0
+f214,50.24469757080078,10094.3896484375,203.0
+f192,49.956268310546875,2488.151123046875,5.0
+f29,48.838172912597656,69915.8828125,82.0
+f220,47.30107116699219,13698.185546875,318.0
+f64,46.715728759765625,161601.984375,43.0
+f27,46.69657897949219,139887.625,475.0
+f152,45.48809814453125,2539.578125,2.0
+f67,44.294769287109375,26379.634765625,89.0
+f14,44.26970672607422,45942.2265625,280.0
+f191,44.25852966308594,5971.04736328125,3.0
+f20,44.10636520385742,27274.34765625,372.0
+f163,42.733341217041016,2792.3251953125,15.0
+f24,41.456214904785156,9650.279296875,11.0
+f218,38.22769546508789,44201.6640625,236.0
+f5,36.669883728027344,7017.15087890625,222.0
+f216,36.60192108154297,1729.5841064453125,227.0
+f215,35.5046501159668,2211.857666015625,36.0
+f134,34.91355895996094,5789.205078125,4.0
+f15,34.72055435180664,3006.513427734375,178.0
+f11,34.5792121887207,10816.3876953125,69.0
+f30,31.092498779296875,10023.046875,47.0
+f219,30.997909545898438,600.1138916015625,47.0
+f166,28.914865493774414,3055.8466796875,3.0
+f65,22.219697952270508,37947.83984375,3.0
+f12,21.965524673461914,2735.67236328125,188.0
+f153,21.305362701416016,1729.6878662109375,1.0
+f211,19.016956329345703,984.0364379882812,5.0

latest.json

@@ -0,0 +1,3 @@
+{
+    "version": "20260809_095516"
+}

model_metadata.json

@@ -0,0 +1,346 @@
+{
+  "trained_at": "2026-08-09T12:34:20.758570",
+  "mode": "scratch",
+  "train_end": "2021-02-15 00:00:00",
+  "val_end": "2024-08-30 00:00:00",
+  "feature_cols": [
+    "log_ret_1d",
+    "log_ret_3d",
+    "log_ret_5d",
+    "log_ret_10d",
+    "log_ret_20d",
+    "log_ret_60d",
+    "ret_1d",
+    "ret_3d",
+    "ret_5d",
+    "ret_10d",
+    "ret_20d",
+    "ret_60d",
+    "ret_120d",
+    "close_sma20_ratio",
+    "close_sma50_ratio",
+    "close_sma200_ratio",
+    "close_ema20_ratio",
+    "high_20_position",
+    "low_20_position",
+    "close_to_52w_high",
+    "close_to_52w_low",
+    "position_in_20d_range",
+    "roc_10",
+    "roc_20",
+    "roc_60",
+    "momentum_10",
+    "momentum_20",
+    "ppo",
+    "dist_sma20",
+    "dist_sma50",
+    "dist_sma200",
+    "vol_10",
+    "vol_20",
+    "vol_60",
+    "vol_ratio",
+    "atr_14",
+    "atr_percent",
+    "parkinson_volatility",
+    "adx_14",
+    "di_plus",
+    "di_minus",
+    "aroon_up",
+    "aroon_down",
+    "rsi_14",
+    "rsi_7",
+    "cci_20",
+    "stochastic_k",
+    "macd",
+    "macd_signal",
+    "macd_histogram",
+    "bb_width",
+    "bb_position",
+    "bb_squeeze",
+    "intraday_range",
+    "gap",
+    "close_position",
+    "volume_ratio",
+    "mfi",
+    "body_percent",
+    "upper_shadow_percent",
+    "lower_shadow_percent",
+    "gap_up",
+    "gap_down",
+    "inside_day",
+    "outside_day",
+    "doji",
+    "distance_from_52w_high",
+    "distance_from_52w_low",
+    "rolling_drawdown",
+    "day_of_week",
+    "month",
+    "is_month_end",
+    "nifty_log_ret_1d",
+    "nifty_log_ret_5d",
+    "nifty_log_ret_20d",
+    "nifty_ret_1d",
+    "nifty_ret_3d",
+    "nifty_ret_5d",
+    "nifty_ret_10d",
+    "nifty_ret_20d",
+    "nifty_ret_60d",
+    "nifty_close_sma20_ratio",
+    "nifty_close_sma50_ratio",
+    "nifty_close_sma200_ratio",
+    "nifty_close_ema20_ratio",
+    "nifty_high_20_position",
+    "nifty_low_20_position",
+    "nifty_close_to_52w_high",
+    "nifty_close_to_52w_low",
+    "nifty_position_in_20d_range",
+    "nifty_roc_10",
+    "nifty_roc_20",
+    "nifty_roc_60",
+    "nifty_momentum_10",
+    "nifty_momentum_20",
+    "nifty_ppo",
+    "nifty_dist_sma20",
+    "nifty_dist_sma50",
+    "nifty_dist_sma200",
+    "nifty_vol_10",
+    "nifty_vol_20",
+    "nifty_vol_60",
+    "nifty_rolling_volatility",
+    "nifty_vol_ratio",
+    "nifty_atr_14",
+    "nifty_atr_percent",
+    "nifty_parkinson_volatility",
+    "nifty_adx_14",
+    "nifty_di_plus",
+    "nifty_di_minus",
+    "nifty_aroon_up",
+    "nifty_aroon_down",
+    "nifty_rsi_14",
+    "nifty_rsi_7",
+    "nifty_cci_20",
+    "nifty_stochastic_k",
+    "nifty_macd",
+    "nifty_macd_signal",
+    "nifty_macd_histogram",
+    "nifty_bb_width",
+    "nifty_bb_position",
+    "nifty_bb_squeeze",
+    "nifty_intraday_range",
+    "nifty_gap",
+    "nifty_close_position",
+    "nifty_volume_ratio",
+    "nifty_mfi",
+    "nifty_body_percent",
+    "nifty_upper_shadow_percent",
+    "nifty_lower_shadow_percent",
+    "nifty_gap_up",
+    "nifty_gap_down",
+    "nifty_inside_day",
+    "nifty_outside_day",
+    "nifty_doji",
+    "nifty_distance_from_52w_high",
+    "nifty_distance_from_52w_low",
+    "nifty_rolling_drawdown",
+    "nifty_bull_market",
+    "nifty_bear_market",
+    "nifty_strong_trend",
+    "nifty_positive_20d",
+    "nifty_positive_60d",
+    "close_1",
+    "vix_return_1d",
+    "vix_return_5d",
+    "vix_return_20d",
+    "vix_ma20",
+    "vix_volatility",
+    "vix_zscore",
+    "vix_percentile",
+    "high_vol_regime",
+    "low_vol_regime",
+    "extreme_fear",
+    "extreme_calm",
+    "gold_ret_1d",
+    "gold_ret_5d",
+    "gold_ret_20d",
+    "gold_sma20_ratio",
+    "gold_sma200_ratio",
+    "gold_momentum_10",
+    "gold_momentum_20",
+    "gold_volatility_20d",
+    "gold_trend",
+    "gold_zscore",
+    "gold_vol_zscore",
+    "gold_high_vol",
+    "brent_ret_1d",
+    "brent_ret_5d",
+    "brent_ret_20d",
+    "brent_sma20_ratio",
+    "brent_sma200_ratio",
+    "brent_momentum_10",
+    "brent_momentum_20",
+    "brent_volatility_20d",
+    "brent_trend",
+    "brent_zscore",
+    "brent_vol_zscore",
+    "brent_high_vol",
+    "brent_spike",
+    "usd_inr_ret_1d",
+    "usd_inr_ret_5d",
+    "usd_inr_ret_20d",
+    "usd_inr_sma20_ratio",
+    "usd_inr_sma200_ratio",
+    "usd_inr_momentum_10",
+    "usd_inr_momentum_20",
+    "usd_inr_volatility_20d",
+    "usd_inr_trend",
+    "usd_inr_zscore",
+    "usd_inr_vol_zscore",
+    "usd_inr_high_vol",
+    "usd_inr_appreciation",
+    "us_ret_1d",
+    "us_ret_5d",
+    "us_ret_20d",
+    "us_ret_60d",
+    "us_breadth_1d",
+    "us_breadth_5d",
+    "us_breadth_20d",
+    "us_dispersion_1d",
+    "us_dispersion_5d",
+    "us_avg_vol_ratio",
+    "us_pct_above_sma20",
+    "us_avg_rsi",
+    "us_avg_dist_52w_high",
+    "corr_nifty_20d",
+    "corr_nifty_60d",
+    "beta_nifty_60d",
+    "is_high_beta",
+    "is_low_beta",
+    "corr_breakdown",
+    "rel_strength_5d",
+    "rel_strength_20d",
+    "corr_gold_60d",
+    "corr_gold_rising",
+    "corr_brent_60d",
+    "corr_brent_rising",
+    "corr_usd_inr_60d",
+    "corr_usd_inr_rising",
+    "corr_us_60d"
+  ],
+  "total_features": 221,
+  "best_iteration": 1999,
+  "params": {
+    "objective": "binary:logistic",
+    "eval_metric": "aucpr",
+    "tree_method": "hist",
+    "device": "cuda",
+    "learning_rate": 0.02,
+    "max_depth": 6,
+    "subsample": 0.75,
+    "colsample_bytree": 0.75,
+    "min_child_weight": 20,
+    "gamma": 0.1,
+    "reg_alpha": 0.1,
+    "reg_lambda": 1.0,
+    "seed": 42,
+    "max_bin": 256,
+    "n_estimators": 2000
+  },
+  "metrics": {
+    "val": {
+      "pr_auc": 0.9042,
+      "roc_auc": 0.9472,
+      "precision": 0.7981,
+      "recall": 0.8364,
+      "ic": 0.714,
+      "n_rows": 1698058,
+      "pos_rate": 0.3062,
+      "brier_score": 0.0835,
+      "brier_skill_score": 0.6068000197410583,
+      "mean_calib_error": 0.0749,
+      "no_skill_brier": 0.21240000426769257,
+      "calibration_curve": {
+        "fraction_of_positives": [
+          0.015583287674426738,
+          0.07435920568570255,
+          0.14871492862667884,
+          0.236217017491847,
+          0.32711561210652357,
+          0.42616482259908844,
+          0.5483246365396607,
+          0.6818195209616121,
+          0.821664709208752,
+          0.966290848301184
+        ],
+        "mean_predicted_prob": [
+          0.03545770506861525,
+          0.1437372801748982,
+          0.24671069988005753,
+          0.34814692895842914,
+          0.4489718037772851,
+          0.5493239582662961,
+          0.6504352846764856,
+          0.751485701542276,
+          0.85354191677695,
+          0.9652750704686287
+        ]
+      }
+    },
+    "test_wf": {
+      "folds": [
+        {
+          "fold": 1,
+          "start": "2024-09-01",
+          "end": "2025-02-28",
+          "pr_auc": 0.9336,
+          "ic": 0.7619,
+          "brier": 0.085,
+          "n_rows": 262062,
+          "pos_rate": 0.386
+        },
+        {
+          "fold": 2,
+          "start": "2025-03-01",
+          "end": "2025-08-31",
+          "pr_auc": 0.8987,
+          "ic": 0.7044,
+          "brier": 0.0829,
+          "n_rows": 266786,
+          "pos_rate": 0.3005
+        },
+        {
+          "fold": 3,
+          "start": "2025-09-01",
+          "end": "2026-02-28",
+          "pr_auc": 0.8921,
+          "ic": 0.7011,
+          "brier": 0.0845,
+          "n_rows": 282585,
+          "pos_rate": 0.3007
+        },
+        {
+          "fold": 4,
+          "start": "2026-03-01",
+          "end": "2026-06-30",
+          "pr_auc": 0.8962,
+          "ic": 0.7157,
+          "brier": 0.0925,
+          "n_rows": 180837,
+          "pos_rate": 0.3352
+        }
+      ],
+      "mean_pr_auc": 0.9052,
+      "mean_ic": 0.7208,
+      "mean_brier": 0.0862
+    }
+  },
+  "class_balance": {
+    "total": 4231195,
+    "positive": 1701894,
+    "negative": 2529301,
+    "pos_rate": 0.4022,
+    "neg_rate": 0.5978,
+    "spw": 1.4862
+  },
+  "dataset": "/run/media/lucifer/developer/Developer/stock_model/Dataset/Processed_dataset/labeled_dataset_ml/ml_training_dataset_3.0pct_5days.parquet",
+  "run_dir": "/run/media/lucifer/developer/Developer/stock_model/models/runs/20260809_095516"
+}

versions/20260809_095516/calibration.json

@@ -0,0 +1,34 @@
+{
+  "val": {
+    "brier_score": 0.0835,
+    "brier_skill_score": 0.6068000197410583,
+    "mean_calib_error": 0.0749,
+    "calibration_curve": {
+      "fraction_of_positives": [
+        0.015583287674426738,
+        0.07435920568570255,
+        0.14871492862667884,
+        0.236217017491847,
+        0.32711561210652357,
+        0.42616482259908844,
+        0.5483246365396607,
+        0.6818195209616121,
+        0.821664709208752,
+        0.966290848301184
+      ],
+      "mean_predicted_prob": [
+        0.03545770506861525,
+        0.1437372801748982,
+        0.24671069988005753,
+        0.34814692895842914,
+        0.4489718037772851,
+        0.5493239582662961,
+        0.6504352846764856,
+        0.751485701542276,
+        0.85354191677695,
+        0.9652750704686287
+      ]
+    }
+  },
+  "test_wf": {}
+}

versions/20260809_095516/run_metadata.json

@@ -0,0 +1,346 @@
+{
+  "trained_at": "2026-08-09T12:34:20.758570",
+  "mode": "scratch",
+  "train_end": "2021-02-15 00:00:00",
+  "val_end": "2024-08-30 00:00:00",
+  "feature_cols": [
+    "log_ret_1d",
+    "log_ret_3d",
+    "log_ret_5d",
+    "log_ret_10d",
+    "log_ret_20d",
+    "log_ret_60d",
+    "ret_1d",
+    "ret_3d",
+    "ret_5d",
+    "ret_10d",
+    "ret_20d",
+    "ret_60d",
+    "ret_120d",
+    "close_sma20_ratio",
+    "close_sma50_ratio",
+    "close_sma200_ratio",
+    "close_ema20_ratio",
+    "high_20_position",
+    "low_20_position",
+    "close_to_52w_high",
+    "close_to_52w_low",
+    "position_in_20d_range",
+    "roc_10",
+    "roc_20",
+    "roc_60",
+    "momentum_10",
+    "momentum_20",
+    "ppo",
+    "dist_sma20",
+    "dist_sma50",
+    "dist_sma200",
+    "vol_10",
+    "vol_20",
+    "vol_60",
+    "vol_ratio",
+    "atr_14",
+    "atr_percent",
+    "parkinson_volatility",
+    "adx_14",
+    "di_plus",
+    "di_minus",
+    "aroon_up",
+    "aroon_down",
+    "rsi_14",
+    "rsi_7",
+    "cci_20",
+    "stochastic_k",
+    "macd",
+    "macd_signal",
+    "macd_histogram",
+    "bb_width",
+    "bb_position",
+    "bb_squeeze",
+    "intraday_range",
+    "gap",
+    "close_position",
+    "volume_ratio",
+    "mfi",
+    "body_percent",
+    "upper_shadow_percent",
+    "lower_shadow_percent",
+    "gap_up",
+    "gap_down",
+    "inside_day",
+    "outside_day",
+    "doji",
+    "distance_from_52w_high",
+    "distance_from_52w_low",
+    "rolling_drawdown",
+    "day_of_week",
+    "month",
+    "is_month_end",
+    "nifty_log_ret_1d",
+    "nifty_log_ret_5d",
+    "nifty_log_ret_20d",
+    "nifty_ret_1d",
+    "nifty_ret_3d",
+    "nifty_ret_5d",
+    "nifty_ret_10d",
+    "nifty_ret_20d",
+    "nifty_ret_60d",
+    "nifty_close_sma20_ratio",
+    "nifty_close_sma50_ratio",
+    "nifty_close_sma200_ratio",
+    "nifty_close_ema20_ratio",
+    "nifty_high_20_position",
+    "nifty_low_20_position",
+    "nifty_close_to_52w_high",
+    "nifty_close_to_52w_low",
+    "nifty_position_in_20d_range",
+    "nifty_roc_10",
+    "nifty_roc_20",
+    "nifty_roc_60",
+    "nifty_momentum_10",
+    "nifty_momentum_20",
+    "nifty_ppo",
+    "nifty_dist_sma20",
+    "nifty_dist_sma50",
+    "nifty_dist_sma200",
+    "nifty_vol_10",
+    "nifty_vol_20",
+    "nifty_vol_60",
+    "nifty_rolling_volatility",
+    "nifty_vol_ratio",
+    "nifty_atr_14",
+    "nifty_atr_percent",
+    "nifty_parkinson_volatility",
+    "nifty_adx_14",
+    "nifty_di_plus",
+    "nifty_di_minus",
+    "nifty_aroon_up",
+    "nifty_aroon_down",
+    "nifty_rsi_14",
+    "nifty_rsi_7",
+    "nifty_cci_20",
+    "nifty_stochastic_k",
+    "nifty_macd",
+    "nifty_macd_signal",
+    "nifty_macd_histogram",
+    "nifty_bb_width",
+    "nifty_bb_position",
+    "nifty_bb_squeeze",
+    "nifty_intraday_range",
+    "nifty_gap",
+    "nifty_close_position",
+    "nifty_volume_ratio",
+    "nifty_mfi",
+    "nifty_body_percent",
+    "nifty_upper_shadow_percent",
+    "nifty_lower_shadow_percent",
+    "nifty_gap_up",
+    "nifty_gap_down",
+    "nifty_inside_day",
+    "nifty_outside_day",
+    "nifty_doji",
+    "nifty_distance_from_52w_high",
+    "nifty_distance_from_52w_low",
+    "nifty_rolling_drawdown",
+    "nifty_bull_market",
+    "nifty_bear_market",
+    "nifty_strong_trend",
+    "nifty_positive_20d",
+    "nifty_positive_60d",
+    "close_1",
+    "vix_return_1d",
+    "vix_return_5d",
+    "vix_return_20d",
+    "vix_ma20",
+    "vix_volatility",
+    "vix_zscore",
+    "vix_percentile",
+    "high_vol_regime",
+    "low_vol_regime",
+    "extreme_fear",
+    "extreme_calm",
+    "gold_ret_1d",
+    "gold_ret_5d",
+    "gold_ret_20d",
+    "gold_sma20_ratio",
+    "gold_sma200_ratio",
+    "gold_momentum_10",
+    "gold_momentum_20",
+    "gold_volatility_20d",
+    "gold_trend",
+    "gold_zscore",
+    "gold_vol_zscore",
+    "gold_high_vol",
+    "brent_ret_1d",
+    "brent_ret_5d",
+    "brent_ret_20d",
+    "brent_sma20_ratio",
+    "brent_sma200_ratio",
+    "brent_momentum_10",
+    "brent_momentum_20",
+    "brent_volatility_20d",
+    "brent_trend",
+    "brent_zscore",
+    "brent_vol_zscore",
+    "brent_high_vol",
+    "brent_spike",
+    "usd_inr_ret_1d",
+    "usd_inr_ret_5d",
+    "usd_inr_ret_20d",
+    "usd_inr_sma20_ratio",
+    "usd_inr_sma200_ratio",
+    "usd_inr_momentum_10",
+    "usd_inr_momentum_20",
+    "usd_inr_volatility_20d",
+    "usd_inr_trend",
+    "usd_inr_zscore",
+    "usd_inr_vol_zscore",
+    "usd_inr_high_vol",
+    "usd_inr_appreciation",
+    "us_ret_1d",
+    "us_ret_5d",
+    "us_ret_20d",
+    "us_ret_60d",
+    "us_breadth_1d",
+    "us_breadth_5d",
+    "us_breadth_20d",
+    "us_dispersion_1d",
+    "us_dispersion_5d",
+    "us_avg_vol_ratio",
+    "us_pct_above_sma20",
+    "us_avg_rsi",
+    "us_avg_dist_52w_high",
+    "corr_nifty_20d",
+    "corr_nifty_60d",
+    "beta_nifty_60d",
+    "is_high_beta",
+    "is_low_beta",
+    "corr_breakdown",
+    "rel_strength_5d",
+    "rel_strength_20d",
+    "corr_gold_60d",
+    "corr_gold_rising",
+    "corr_brent_60d",
+    "corr_brent_rising",
+    "corr_usd_inr_60d",
+    "corr_usd_inr_rising",
+    "corr_us_60d"
+  ],
+  "total_features": 221,
+  "best_iteration": 1999,
+  "params": {
+    "objective": "binary:logistic",
+    "eval_metric": "aucpr",
+    "tree_method": "hist",
+    "device": "cuda",
+    "learning_rate": 0.02,
+    "max_depth": 6,
+    "subsample": 0.75,
+    "colsample_bytree": 0.75,
+    "min_child_weight": 20,
+    "gamma": 0.1,
+    "reg_alpha": 0.1,
+    "reg_lambda": 1.0,
+    "seed": 42,
+    "max_bin": 256,
+    "n_estimators": 2000
+  },
+  "metrics": {
+    "val": {
+      "pr_auc": 0.9042,
+      "roc_auc": 0.9472,
+      "precision": 0.7981,
+      "recall": 0.8364,
+      "ic": 0.714,
+      "n_rows": 1698058,
+      "pos_rate": 0.3062,
+      "brier_score": 0.0835,
+      "brier_skill_score": 0.6068000197410583,
+      "mean_calib_error": 0.0749,
+      "no_skill_brier": 0.21240000426769257,
+      "calibration_curve": {
+        "fraction_of_positives": [
+          0.015583287674426738,
+          0.07435920568570255,
+          0.14871492862667884,
+          0.236217017491847,
+          0.32711561210652357,
+          0.42616482259908844,
+          0.5483246365396607,
+          0.6818195209616121,
+          0.821664709208752,
+          0.966290848301184
+        ],
+        "mean_predicted_prob": [
+          0.03545770506861525,
+          0.1437372801748982,
+          0.24671069988005753,
+          0.34814692895842914,
+          0.4489718037772851,
+          0.5493239582662961,
+          0.6504352846764856,
+          0.751485701542276,
+          0.85354191677695,
+          0.9652750704686287
+        ]
+      }
+    },
+    "test_wf": {
+      "folds": [
+        {
+          "fold": 1,
+          "start": "2024-09-01",
+          "end": "2025-02-28",
+          "pr_auc": 0.9336,
+          "ic": 0.7619,
+          "brier": 0.085,
+          "n_rows": 262062,
+          "pos_rate": 0.386
+        },
+        {
+          "fold": 2,
+          "start": "2025-03-01",
+          "end": "2025-08-31",
+          "pr_auc": 0.8987,
+          "ic": 0.7044,
+          "brier": 0.0829,
+          "n_rows": 266786,
+          "pos_rate": 0.3005
+        },
+        {
+          "fold": 3,
+          "start": "2025-09-01",
+          "end": "2026-02-28",
+          "pr_auc": 0.8921,
+          "ic": 0.7011,
+          "brier": 0.0845,
+          "n_rows": 282585,
+          "pos_rate": 0.3007
+        },
+        {
+          "fold": 4,
+          "start": "2026-03-01",
+          "end": "2026-06-30",
+          "pr_auc": 0.8962,
+          "ic": 0.7157,
+          "brier": 0.0925,
+          "n_rows": 180837,
+          "pos_rate": 0.3352
+        }
+      ],
+      "mean_pr_auc": 0.9052,
+      "mean_ic": 0.7208,
+      "mean_brier": 0.0862
+    }
+  },
+  "class_balance": {
+    "total": 4231195,
+    "positive": 1701894,
+    "negative": 2529301,
+    "pos_rate": 0.4022,
+    "neg_rate": 0.5978,
+    "spw": 1.4862
+  },
+  "dataset": "/run/media/lucifer/developer/Developer/stock_model/Dataset/Processed_dataset/labeled_dataset_ml/ml_training_dataset_3.0pct_5days.parquet",
+  "run_dir": "/run/media/lucifer/developer/Developer/stock_model/models/runs/20260809_095516"
+}

versions/20260809_095516/terminal_output.md

@@ -0,0 +1,124 @@
+
+ python train_model.py
+══════════════════════════════════════════════════════════════════════
+  XGBoost — Indian Stock Market Prediction
+  Dataset : ml_training_dataset_3.0pct_5days.parquet
+  Model   : /run/media/lucifer/developer/Developer/stock_model/models
+  Chunk   : 50,000 rows/chunk  (~30MB per chunk)
+══════════════════════════════════════════════════════════════════════
+
+════ Device ══════════════════════════════════════════════════════════
+  ✅  CUDA available — training on GPU
+
+════ Splits ══════════════════════════════════════════════════════════
+  Total trading days : 5431
+  Train : 2008-09-02 00:00:00 → 2021-02-15 00:00:00  (3801d, 70%)
+  Val   : next → 2024-08-30 00:00:00  (1086d, 20%)
+  Test  : next → 2026-06-30 00:00:00  (544d, 10%)
+
+════ Schema ══════════════════════════════════════════════════════════
+  Feature columns : 221
+
+════ Dataset audit ═══════════════════════════════════════════════════
+  Rows    : 6,921,523
+  Symbols : 2,656
+  Dates   : 2008-09-02 00:00:00 → 2026-06-30 00:00:00
+  Pos     : 2,548,741  (36.8%)
+  Neg     : 4,372,782  (63.2%)
+
+  Run dir : runs/20260809_095516
+
+════ Model check ═════════════════════════════════════════════════════
+  ℹ️  No model found → TRAIN FROM SCRATCH
+
+════ Class balance ═══════════════════════════════════════════════════
+  pos=1,701,894  neg=2,529,301  pos_rate=40.2%  spw=1.49
+
+════ Building train DMatrix (chunked) ════════════════════════════════
+    Loaded 1,000,000 rows (20 chunks)...
+    Loaded 2,000,000 rows (40 chunks)...
+    Loaded 3,000,000 rows (60 chunks)...
+    Loaded 4,000,000 rows (80 chunks)...
+    Streamed 4,231,195 rows in 85 chunks (50000 rows/chunk)
+  DMatrix: 4,231,195 rows × 221 features
+
+════ Building val DMatrix (chunked) ══════════════════════════════════
+    Loaded 1,000,000 rows (20 chunks)...
+    Streamed 1,698,058 rows in 34 chunks (50000 rows/chunk)
+  DMatrix: 1,698,058 rows × 221 features
+
+════ Training ════════════════════════════════════════════════════════
+[0]	val-aucpr:0.87528
+[100]	val-aucpr:0.89288
+[200]	val-aucpr:0.89757
+[300]	val-aucpr:0.89873
+[400]	val-aucpr:0.89987
+[500]	val-aucpr:0.90073
+[600]	val-aucpr:0.90138
+[700]	val-aucpr:0.90167
+[800]	val-aucpr:0.90207
+[900]	val-aucpr:0.90227
+[1000]	val-aucpr:0.90252
+[1100]	val-aucpr:0.90273
+[1200]	val-aucpr:0.90298
+[1300]	val-aucpr:0.90316
+[1400]	val-aucpr:0.90333
+[1500]	val-aucpr:0.90351
+[1600]	val-aucpr:0.90363
+[1700]	val-aucpr:0.90377
+[1800]	val-aucpr:0.90390
+[1900]	val-aucpr:0.90403
+[1999]	val-aucpr:0.90422
+
+  Best iter : 1999  |  Best score : 0.9042
+
+════ Validation evaluation (chunked inference) ═══════════════════════
+
+  [VAL]  PR-AUC=0.9042  ROC-AUC=0.9472  Prec=0.7981  Rec=0.8364  IC=0.7140  n=1,698,058
+
+  Decile hit-rate (D10=highest confidence):
+decile  hit_rate  count
+    D1     0.004 169806
+    D2     0.009 169806
+    D3     0.017 169806
+    D4     0.031 169805
+    D5     0.066 169806
+    D6     0.146 169806
+    D7     0.311 169805
+    D8     0.605 169806
+    D9     0.888 169806
+   D10     0.984 169806
+
+  ✅  IC strong  (IC=0.7140)
+    Brier=0.0835  skill=0.6068  MCE=0.0749
+  Saved → val_predictions.parquet
+
+════ Walk-forward test evaluation ════════════════════════════════════
+
+  Walk-forward test: 4 folds × 6 months each
+  Fold  1 (2024-09-01 → 2025-02-28):  PR-AUC=0.9336  IC=0.7619  Brier=0.0850  n=262,062
+  Fold  2 (2025-03-01 → 2025-08-31):  PR-AUC=0.8987  IC=0.7044  Brier=0.0829  n=266,786
+  Fold  3 (2025-09-01 → 2026-02-28):  PR-AUC=0.8921  IC=0.7011  Brier=0.0845  n=282,585
+  Fold  4 (2026-03-01 → 2026-06-30):  PR-AUC=0.8962  IC=0.7157  Brier=0.0925  n=180,837
+  WF predictions → test_wf_predictions.parquet
+
+  IC stability:  min=0.7011  max=0.7619  mean=0.7208
+  ⚠️  IC range > 0.05 — model is regime-sensitive
+
+════ Saving ══════════════════════════════════════════════════════════
+  Phase 1 ✓ → runs/20260809_095516/
+  Phase 2 ✓ → xgb_model.json, model_metadata.json
+══════════════════════════════════════════════════════════════════════
+  ✅  DONE
+
+  Stable files:
+    feature_importance.csv                          9.1 KB
+    model_metadata.json                             7.8 KB
+    xgb_model.json                              14750.7 KB
+
+  Run files (20260809_095516):
+    calibration.json                                0.8 KB
+    run_metadata.json                               7.8 KB
+    test_wf_predictions.parquet                  7002.0 KB
+    val_predictions.parquet                     10823.4 KB
+══════════════════════════════════════════════════════════════════════

versions/20260809_095516/test_wf_predictions.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c64d99b9435c35dbbe3675fbb01271aa991ea91cc56b1e5ae1839e494acd984d
+size 7170046

versions/20260809_095516/val_predictions.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:773f16e3763716a3f0da8bd3d1b0b5d08211e1b584f64f556edb6f00bdefd4c2
+size 11083146

xgb_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b97c1d15c8924fff752cd318eac6feb641ab4368b4d5f780ad2902bebfdad77
+size 15104678