I backtested all candlestick patterns across every instrument to see if they are actually profitable in any way

TL;DW: I backtested 24 candlestick patterns to see if any actually predict price.

Data:

Bitcoin and Nasdaq futures from 5-minute up to daily (\~1.4M candles), plus 23 instruments across stocks, indices, forex, commodities, and crypto on the daily timeframe going back to 1970 (\~180k bars). Each pattern was scored against a baseline of random trades taken the some number of times as the strategy.

Results:

\- Across 1,186 cross-asset tests, 117 patterns beat the market by a significant margin and 103 did significantly worse, they were roughly balanced, which is what you'd get from random noise. As a group, candlestick patterns add basically nothing.

\- Bullish patterns win about 54% of the time, but the market is up \~54% of the time anyway, so they're not beating anything. A few (bullish engulfing, bull marubozu) actually do worse than a random entry.

\- Bearish patterns win less than 50% of the time — shorting them loses money on average, because you're betting against a market that mostly drifts up.

\- The famous ones are duds: Morning Star ≈ coin flip, Three White Soldiers didn't beat a random trade in a single market, and the move after a Doji is slightly smaller than a normal day (not bigger).

\- There was only exception - which is the inverted hammer (small body, long upper wick after a sell-off). It had a small but statistically real edge (\~+0.3% over 10 days), significant in about half the markets tested. This pattern also showed a some profit on 15m BTC chart. However, the average trade edge was only about (0.03%) so its only usable if you can get your fees below that. (might be possible with maker rebates, maybe, not sure)

NOTE: Most of these backtests were scored against random trades taken for the same duration and strategy. This is because for most strategies, lets take inverted hammer for example, cash was just sitting around 88% of the time. The risk profile of most of these strategies are also much lower than buy and hold due to smaller drawdowns, so leveraged trading is possible, so comparing to buy and hold won't be a fair comparison.