LEAPS Paper Trader — validation loop
Every time you click Paper on a scan row (or from a detail page), the contract + all scores + regime state are persisted to KV. A nightly grader marks each open position to the current option mid-price. Positions close automatically at expiration or after 365 days held. This page shows the running results — and, most importantly, whether higher opportunity scores actually predicted higher realized P&L. That's the learning loop.
Score-decile P&L The learning loop
Closed paper trades bucketed into deciles by their overall opportunity score at entry. A monotone-increasing bar chart (left low → right high) means the scoring model has real predictive power. A flat or U-shaped chart means the score isn't sorting outcomes and the weights need re-tuning.
Why this matters
The scoring engine is a weighted linear composite over ~20 factors (see methodology). Weights are hand-set from Van-Tharp-style prior knowledge. They may or may not actually reflect what predicts LEAPS returns on THIS universe in THIS regime.
Every paper trade records the full scoring_version tag at entry. When enough trades have graded out (target: 30+ per decile), we can:
- Check rank correlation (Spearman) of entry score vs realized P&L.
- Regress realized P&L on each sub-score to identify which factors carry signal and which are noise.
- Re-weight the composite and bump the scoring_version tag. The next generation's outcomes are graded against the new weights — an honest walk-forward validation.
Fair warning: with fewer than ~50 settled trades, decile buckets are noise. Wait for a meaningful sample before reading the chart.
Paper-trade history
| Opened | Ticker | Side | Contract | Score | Regime | Entry $ | Mark $ | P&L $ | P&L % | Days | Status | Grade |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading history… | ||||||||||||