Questions to ask a strategy that only works on one timeframe
Posted: Sat Nov 14, 2026 9:44 am
It's common to find a setup that looks clean on M5 and falls apart on M1 and M15. Sometimes that's fine, sometimes it's a warning sign, and it's worth working out which before putting money behind it.
A few questions that help separate the two:
Does the logic explain why M5 is special? If the setup is built around something with a natural time scale, like the first 15 minutes after a data release, it makes sense that only one timeframe captures it. If the logic is just "price bounces off the 20 EMA", there's no obvious reason it should care about the chart setting.
Does it degrade gradually or collapse? A real edge usually weakens as you move to neighbouring timeframes rather than vanishing. If M4 and M6 (or a 3-minute and 7-minute chart) still show something, that's encouraging. If only exactly M5 works, that leans toward luck in the sample.
How many trades is it based on? A one-timeframe result built on 40 trades is very fragile. Sample size matters even more when there's a reason to be suspicious.
Were the parameters tuned on that timeframe? If the stop, target and indicator lengths were all optimised on M5, then naturally M5 looks best. The fair test is to keep the rules fixed and check other periods and other pairs too.
None of this proves a strategy wrong. It just helps decide how much trust to give it, and how small the first live size should be.
Do people here deliberately test on neighbouring timeframes, or do you treat each timeframe as its own separate system?
A few questions that help separate the two:
Does the logic explain why M5 is special? If the setup is built around something with a natural time scale, like the first 15 minutes after a data release, it makes sense that only one timeframe captures it. If the logic is just "price bounces off the 20 EMA", there's no obvious reason it should care about the chart setting.
Does it degrade gradually or collapse? A real edge usually weakens as you move to neighbouring timeframes rather than vanishing. If M4 and M6 (or a 3-minute and 7-minute chart) still show something, that's encouraging. If only exactly M5 works, that leans toward luck in the sample.
How many trades is it based on? A one-timeframe result built on 40 trades is very fragile. Sample size matters even more when there's a reason to be suspicious.
Were the parameters tuned on that timeframe? If the stop, target and indicator lengths were all optimised on M5, then naturally M5 looks best. The fair test is to keep the rules fixed and check other periods and other pairs too.
None of this proves a strategy wrong. It just helps decide how much trust to give it, and how small the first live size should be.
Do people here deliberately test on neighbouring timeframes, or do you treat each timeframe as its own separate system?