ML Monitoring Dashboard

Performance monitoring for P4 OTM Predictor

Market Days Only: Data is collected only during US market trading days (Monday-Friday, excluding holidays). No data is shown for weekends or market holidays like Presidents' Day, Memorial Day, etc.
Total Decisions

Passed Filter

Rejected

Acceptance Rate

Daily Metrics (Last 7 Market Days)

Excludes weekends and market holidays (e.g., Presidents' Day)

Date Model Version Total Decisions Passed Rejected Acceptance Rate Conf (Passed) Conf (Rejected) Total Opportunities Opps/Symbol
Loading...

Recent Filter Decisions (Last 50)

Date Symbol Decision Confidence VIX RSI ATR % Beta ETF Mega Cap High Vol
Loading...
-->

P4 OTM Probability Predictor

XGBoost model predicting Out-of-The-Money probability

Checking outcome pipeline health…
Total Predictions

all options scored
Outcomes Collected

Calibration Error

Actual OTM Rate

Base Rate

83.6%

OTM rate in training data

Accuracy Tracks

Recommendations = options actually sent in alerts / auto-executed (Phase 7 winners, identified by tiebreaker metadata). All Predictions = every option P4 scored including non-winners. Use "Recommendations" for business performance; "All Predictions" for model diagnostics.
Note: "Recommendations" track uses p4_changed_winner IS NOT NULL as a proxy for Phase 7 winners until Step 4 logging is fully deployed.
Track Total Predictions Settled Outcomes Actual OTM Rate Calibration Error vs Base Rate (83.6%)
Loading…

Calibration Summary

Evaluated Predicted OTM Avg Actual OTM Rate Calibration Error Brier Score
Loading…

Performance by Probability Bucket

Probability Range Count (settled) Avg Predicted Actual OTM Rate Error Status
Loading…
Segmented Accuracy Step 3 — Data Collection In Progress

Data collection in progress — available after Step 3 logging columns are populated (blend_regime, execution_path, symbol_in_training_set, dte_in_training_range).

Once live, this section will show accuracy breakdowns by:

  • Execution path: auto-execute vs client alert
  • Training coverage: in-training symbols (AAPL, AMZN, GOOG, HOOD, META, MSFT, NFLX, NVDA, PLTR, SPY, TSLA, VOO) vs extrapolated symbols
  • DTE range: DTE ≥ 14 (in training distribution) vs DTE < 14 (extrapolated)
  • Blend regime: model-heavy (RSI ≥ 40, 60% model / 40% delta) vs delta-heavy (RSI < 40, 40% model / 60% delta)
P4 Decision Impact Step 4 — Data Collection In Progress

Data collection in progress — available after Step 4 Phase 7 tiebreaker instrumentation (p4_changed_winner) is fully deployed across all scan paths.

Once live, this section will show:

  • How many selections had multiple candidates in the same RI bucket
  • How often P4 changed the winner vs confirmed the RI-only pick
  • Outcome comparison: P4-changed picks vs RI-only picks (did overriding RI actually help?)

Recent Predictions (Last 50)

Yellow DTE = out-of-distribution (DTE < 14, below training range). Yellow symbol = not in training set (extrapolated). ★ Winner = Phase 7 selected option.

ID Symbol Strike Expiry DTE P4 Prob Stock Price Safe Strike IV Delta Blend Regime Exec Path Outcome Date
Loading…