Reversal Candle Win-Rate Backtester
The Range Edge Candle Grader scores one setup at a time. This tool scores your history instead. Paste OHLC bars and it walks a rolling Donchian channel (Richard Donchian, the 20-bar lookback made famous by the 1980s Turtle system) across the data, flags every hammer, engulfing bar, star, tweezer, harami, piercing line, and inside-bar reversal that printed inside the edge zone, then races each one toward the channel midpoint against a tight stop over the next 5 bars. The result is a ranked win rate per pattern on your own data, plus a call on the setup sitting on the edge right now.
One bar per line as open,high,low,close. No dates needed. 150 bars or more gives each pattern a fair sample. The channel needs the lookback in bars before scoring starts.
No live signal
Paste bars and hit run. The chart fills in with every pattern that reached your edge zone, graded on the 3 bars that followed it.
How it backtests. The range is a rolling Donchian channel: the highest high and lowest low of the prior N bars (default 20, recomputed at every bar so nothing here sees ahead). A pattern counts as an edge instance when its extreme sits inside the edge zone you set (default 20% of the channel width) from the rolled high or low. Pattern rules are the same 8 shapes as the Range Edge Candle Grader, following Nison's Japanese Candlestick Charting Techniques (1991): hammer and shooting star, bullish and bearish engulfing, piercing line and dark cloud cover, tweezer top and bottom, bullish and bearish harami confirmed by the next close, morning and evening star, and the inside-bar reversal. Doji prints are skipped, same as the grader treats them: a pause, not a signal.
Outcome races 2 levels over the 5 bars after the signal bar closes. Win: a close reaches the channel midpoint, the same target the Partial Exit Plan Backtester calls P1. Loss: a close breaks the edge by more than 0.25x the average bar range over the lookback window, the same stop-buffer multiple the Range Edge Candle Grader uses against ATR, substituting the average range since this tool takes no separate ATR input. Whichever level gets hit first inside the 5 bars decides it; neither hit in time is chop. Instances inside the last 5 bars of your paste can't be graded yet and are dropped from the count, except the final bar, which becomes the live read. A pattern needs 5 or more graded instances before its win rate counts as a read; fewer shows as thin in the chart. Keyboard: Enter runs the backtest, 1 and 2 load the examples.
Educational tool only, not financial advice. A pattern with a strong win rate on your pasted history can still lose on the next print. Size for the stop, not for the backtest.
Built by Jason Parker, founder of Trading Ranges.