---
name: marketpilot-time-machine
description: "Run reproducible historical and bootstrap scenarios for a user-specified entry, stop, and target."
---

# MarketPilot AI Time Machine

## Workflow
1. Require an explicit plan (entry, stop, target), instrument, interval, and historical OHLCV file. Verify `0 < stop < entry < target` for the supported long-only mode.
2. Run `marketpilot time-machine --csv <BARS.csv> --entry-index <N> --entry <ENTRY> --stop <STOP> --target <TARGET> --simulations 10000 --seed 19`.
3. Confirm that the sampling pool consists only of bars preceding the entry index. Report the actual historical path separately from the bootstrap distribution.
4. Report target-first, stop-first and still-open frequencies; average R; percentiles; seed; method; and every limitation returned by the tool.
5. Explain same-bar ambiguity: if both levels are touched in one bar, the engine assumes the stop first. Note that the method bootstraps independent historical OHLC bar-return vectors and ignores dependence, market impact, spread, halts and news jumps.

## Interpretation
- Frequencies are conditional on this dataset, plan, resampling method and horizon—not objective probabilities of future market outcomes.
- Do not call the output a forecast, an edge, or proof that a strategy works.
- A small or unrepresentative sample should be labeled insufficient; do not replace missing history with synthetic examples.
- Always show actual sample counts and the seed so the run can be reproduced.