Kill Criteria — final version v1.0
Finalized: 2026-07-07 Status: final (the Master Plan Phase 0.1 example values adopted as the v1 final values) Amendment rule: no amendments while in a loss. Amendments are allowed only when all positions are closed + there is a written rationale + the change is recorded in this document's revision history.
1. Trigger criteria
| Scope | Criterion (final value) | Action when triggered |
|---|---|---|
| Strategy level | Out-of-sample Sharpe ratio < 0.5 or live-vs-backtest deviation > 50% | Discard the strategy. No reviving it by re-tuning parameters |
| Account level | MDD reaches -15% | Close all positions → 2-week cooling-off → resume only after documenting a root-cause analysis |
| Project level | Paper trading fails 3 times in a row (deviation criterion not met) | Live trading entry postponed indefinitely; simulation model redesigned |
| Daily level | Daily loss of -5% | Bot stops automatically for the day (no manual restart; resumes automatically the next day) |
2. Success criteria (not profit)
- Backtest ↔ paper trading deviation < 20% [inference — a reasonable threshold, not an industry standard]
- 3+ months of uninterrupted automated operation
- Automatic recording of every trade log and decision rationale (track record)
3. Measurement definitions (removing ambiguity)
- MDD: drawdown from peak on account equity. Includes unrealized P&L.
- Daily loss: versus starting equity of the day at UTC 00:00. Realized + unrealized combined.
- Deviation: the mean of
|backtest expected return − measured return| / |backtest expected return|per trade. Trades whose denominator is near 0 are aggregated separately by absolute difference. - 1 paper failure: mean deviation ≥ 20% after running at least 2 weeks.
4. Revision history
| Date | Change | Rationale | Position status at the time |
|---|---|---|---|
| 2026-07-07 | v1.0 finalized (plan example values adopted) | Initial finalization | No positions |
Source and migration evidence
Source: backtest/docs/kill-criteria.md
Original SHA-256: 9be3b056f47c13541b9031fd668916d7dba8a11c81a3fd70b7c278f6e5041988
Raw data and account information are not part of the public package. The original results do not replace statistical or scientific revalidation. The English text is a translation; numbers, tables and charts are carried over unchanged.