Choosing a timeframe without fooling yourself
The same rules can pass on 4h and fail on 5m. Why that happens, and how to compare timeframes fairly.
Run one rule set across every timeframe and the spread is usually enormous. That spread is real information — as long as the comparison is fair.
Fair means one identical window
If the 1-hour test covers two years and the 1-minute test covers three weeks, you have not compared timeframes. You have compared market regimes. Every timeframe must be replayed over the same start and end instant.
Lower timeframes pay more in costs
Costs scale with trade count. Drop from 4h to 5m and you may multiply your trade count by fifty — and your total cost with it. A strategy whose gross edge is 0.15% per trade is already gone.
Higher timeframes pay in sample size
Go the other way and costs stop mattering while evidence dries up. Weekly rules produce so few trades that the confidence range swallows the result.
How to read a timeframe table
Rank on expectancy after costs, then sanity-check three things:
- Is the trade count large enough to trust the rate?
- Is the drawdown survivable on your account, not just small in percent?
- Does the winner still win on held-out candles?
The best timeframe is rarely the one with the highest win rate. It is the one where the edge is big enough to clear costs and frequent enough to be measured.