Outlytic Outcomes REAL DATA · v1.0.28

Interpretation: P coverage and price/RV errors test the forecasting layer. P−Q edge buckets test whether larger disagreement with market-implied probabilities corresponds to larger realized advantage. Vol Edge tests whether forecast realized volatility correctly falls above or below market IV. These are kept separate from executable option P/L so forecast quality and option translation can be diagnosed independently.

How Quant Intelligence can become more profitable: do not optimize for direction accuracy or trade count alone. First calibrate P width so P EV is not inflated by over-wide tails; then require executable value to survive bid/ask and slippage; then use Outcomes to discover whether edge is stable inside particular P−Q, health, regime and robustness buckets; finally keep only selectors that add realized P/L versus AI #1 out of sample. The panel remains research-only until samples are large enough.

• Keep raw P live, but test the calibrated shadow width before trusting large P EV; over-wide tails can manufacture option value.

• Do not use direction as a hard profitability gate yet; emphasize value/volatility edge and executable P/L until directional calibration improves.

How to read it: Actual coverage above Expected means the forecast interval may be too wide; below Expected means it may be too narrow. A well-calibrated model should approach each target as the sample grows. Mean interval width is the standard sharpness diagnostic: narrower is better only if calibration is preserved. Coverage above target suggests intervals may be too wide; coverage below target suggests they may be too narrow. Mean interval width is the standard sharpness diagnostic. The displayed sharpness score is an Outlytic convenience index: 100/(1 + mean width % of P50), so higher means tighter. The dispersion multiplier is a retrospective research estimate only and is not applied to live forecasts, rankings, or P EV.

Research guardrail: P CALIBRATED is a shadow, retrospective rescaling of each forecast around its own P50. It is shown to measure whether a narrower or wider P would have improved historical coverage. Option AI may now display and log the CAL shadow beside RAW, but it does not change the official live P forecast, RAW Pure Quant winner, Final Outlytic winner, or trade routing. The shadow fit is in-sample, so improvement here is not proof of out-of-sample edge. Scalar log-distance rescaling around each forecast P50; retrospective coverage fit across available P intervals.

Data hygiene: Simulator runs without a usable market Q distribution are now skipped automatically. Existing bad/no-data Simulator P/Q rows can be removed without touching Option AI, My Predictions, Gap/Pivot, or other valid Outcomes data. Purge invalid Simulator P/Q runs

Why this exists: Option AI remains the candidate generator. P/Q is judged as a second-stage reranker. RAW and CAL selectors are tracked independently so Outlytic can prove whether Outcomes calibration improves selection instead of assuming that it does.