Description
Catch the bugs in your backtest before your buyers do.
Runs 8 automated checks on any backtest CSV export and flags the statistical patterns that almost always point to a logic bug — before a customer finds them for you.
What it checks (8 checks):
1. Sample size is large enough (roughly 100-300 trades minimum)
2. Win-rate plausibility (a win rate above ~75% is almost always a bug, not a real edge)
3. Directional asymmetry (long vs short results that are suspiciously different)
4. Trade frequency per session (catches uncontrolled re-entries)
5. Time coverage (a single market regime, or several?)
6. Transaction-cost heuristic (P&L that looks ‘too clean’)
7. Look-ahead bias heuristic (only if the CSV includes trade high/low)
8. Walk-forward vs in-sample check (only if the CSV distinguishes both)
Each check returns PASS / WARNING / FAIL / SKIPPED with a plain-English explanation — you don’t need to be a quant to read the result.
Input format: works with any CSV that has a date column and a P&L column. Column names are auto-detected (no renaming needed) for date (date, entry_time, timestamp), P&L (pnl_usd, pnl_pct, pnl, profit), and direction (direction, side).
Native TradingView support: upload the CSV exported directly from TradingView Strategy Tester’s ‘List of Trades’ — untouched — and the tool automatically detects and normalizes its native 2-rows-per-trade format for correct analysis.
Honest limitation: MetaTrader and NinjaTrader don’t have dedicated normalization yet — you may need to rename columns manually for those platforms. If your CSV has no recognizable P&L column, the tool tells you so explicitly instead of failing silently. Checks 7 and 8 (look-ahead bias, walk-forward) only run if your CSV has the required columns — the tool never fakes a result it can’t compute.
Requirements: Python 3.8+, pandas, numpy. No internet connection required — runs 100% locally, your data never leaves your machine.
Note: this tool doesn’t replace a code audit and doesn’t guarantee a strategy is profitable — it flags statistical patterns commonly associated with backtesting bugs, for you to review yourself.
Reviews
There are no reviews yet.