Candlestick patterns, tested like your account depends on it
Every trading course starts here. The bullish engulfing. The doji. Three white soldiers. Patterns with three-hundred-year-old Japanese names, taught as if they were laws of nature. "When you see this candle, buyers have taken control."
We coded them exactly as the textbooks define them — no creative reinterpretation — and ran them on index intraday data through our full engine: out-of-sample windows, brokerage and slippage, fills only at achievable prices. Long side, short side, multiple timeframes.
The scoreboard
| Pattern (as taught) | Instrument scope | Out-of-sample result |
|---|---|---|
| Bullish engulfing → long | Both major indices | −₹0.9L to −₹2.0L |
| Doji breakout | Both major indices | −₹2.6L to −₹2.9L |
| Three white soldiers | NIFTY | −₹1.7L |
| Three white soldiers | Bank index | positive recently — but negative in training era: luck, not edge |
The doji breakout deserves special mention: in walk-forward testing on NIFTY it lost −₹1.4L — a strategy whose entire premise is "indecision resolves into direction" turned out to be a machine for donating the spread, roughly 45% win rate with losers outrunning winners.
The one that "worked" — and why it's the most dangerous row
Look at the last row. Three white soldiers on the bank index was profitable in the recent period. A course seller would show you exactly that window. Our engine flags it differently: the same rules lost money across the training era. When a strategy only works in the most recent slice, that's not an edge — that's a coin that came up heads lately. This single distinction — "recent-only" versus "survives all eras" — is most of what separates testing from marketing.
Not just indices — commodities agree
We repeated the exercise on five global commodities — gold, silver, copper, crude oil, natural gas — on 1, 5 and 15-minute charts, using fourteen years of 1-minute history. Result: the standard candlestick patterns lost money intraday on every timeframe tested. On daily charts a few patterns crawled to marginal breakeven — nothing tradeable after costs. Two markets, one conclusion.
What would change our mind
Good research stays falsifiable. If a specific pattern, on a specific instrument and timeframe, with specific entry/exit rules survives our engine — all eras, honest fills, costs — we'll publish that too. That's exactly what the submission box below is for. So far, the textbook is 0 for everything intraday.
Send us the exact rules — we'll test them on 800 million bars, free, before you spend more on it.