EXPLORATORY_REAL_MARKET

One dataset.
Declared comparisons.

840,736 samples · 24 usable features · 551,698 shared forecasts. Models, the fixed momentum baseline and equal-weight baseline share dates, outcomes, execution lag and cost rules.

All methods and cost scenarios

MethodCostsRunsNet CAGRNet Sharpe mean ± seed stdMax drawdown
equal_weight1×112.75%1.141 ± 0.000-18.70%
equal_weight2×112.70%1.137 ± 0.000-18.71%
lightgbm1×318.23%1.337 ± 0.022-21.46%
lightgbm2×317.51%1.290 ± 0.023-21.64%
ridge1×312.85%1.005 ± 0.000-22.06%
ridge2×312.20%0.960 ± 0.000-22.20%
ridge_without_size_feature1×312.24%0.955 ± 0.000-21.68%
ridge_without_size_feature2×311.58%0.909 ± 0.000-21.84%
single_momentum1×111.38%0.871 ± 0.000-23.40%
single_momentum2×110.95%0.842 ± 0.000-23.47%

Source identity, exact protocol and all runs · Metrics CSV · Fold results · Daily returns (gzip CSV) · Hashes

This is a protocol comparison. Its real-market variant remains exploratory until the declared source limitations are resolved. No test result selects a model or seed for deployment.