HKJC · local network
Analysis dashboards
Single Flask app for predictions payout analysis, ROI trends, meta-prediction plots, historical dividend distributions, feature importance / SHAP, NN top-3 backtest ROI, lock strategy odds work, and HK Racing Pro application logs. Accessible from other machines on the same network via this host’s LAN IP and the listen port.
Meeting status
Dashboards
ROI Trends
Same 17.8 Predictions filters and series, plotted as ROI through race dates instead of accumulated payout.
/roi/Lock Strategy
Accum-payout locks and money-weighted odds rules for PreRace1 Win/QP — playbook, t+1/t+2 detail, and ROI by odds gate.
/lock-strategy/Bets
Race-day ledger: T-30 vs T-5 tickets, skip/place, official dividends, and meeting P&L. Jason decides stake — no HKJC auto-bet.
/bets/Combo Payouts
Per-combo payout grid across race dates for T-30, T-5, and PreRace1 — stakes, payouts, and ROI side by side.
/combo-payouts/Predictions
Cumulative payout graphs (17.8) with cascading filters, ROI bands, and date ranges.
/predictions/Meta
Meta-prediction cumulative dividends (18.6) with strategy filters and venue summary.
/meta/Feature Importance
LightGBM gain rankings and SHAP views by feature set, selector, top finish, and params — with stability across runs.
/feature-importance/Dividends
WIN · PLACE · QUINELLA · QUINELLA PLACE distributions by course, race type, track, course, going, class, draw, and distance.
/dividends/NN Backtest
Deep-learning top-3 backtest series: cumulative dividends, unit ROI, stake ROI, and net profit.
/nn/App Logs
Analyze hk_racing_pro/run.py logs: access traffic, latency, status codes, and errors.
Pred Run Logs
Analyze run_predictions_analysis.py session logs: feature sets, platforms, command timings, and errors.