What Is the Best Timeframe for an EA? What Our Backtests Say

Publicado: 2026/11/22 UTC Actualizado: 2026/11/22 UTC Enlace permanente

“What is the best timeframe for an EA?” has a short answer and a useful one. The short answer: there isn’t one. The useful answer: a strategy has a timeframe where its edge per trade is bigger than its cost, and often only one. You find it by testing the same rule on several timeframes and looking for the peak.

We know this because we did exactly that and kept every result, including the ones that failed. This article shows the numbers: one strategy tested on four timeframes, one entry that a higher timeframe made worse, and one EA where the timeframe hardly matters at all.

One Strategy, Four Timeframes

Our range-reversion EA buys short dips in the London morning, long only. We filed the same rule on M1, M5, M15 and M30, on the same pairs. When we rejected the last of them, the record summed up the whole map:

Timeframe map is now complete: M1 0.59, M5 ~1.0, M15 ~1.5, M30 ~0.8. M15 is the peak, not a point on a rising curve.

TimeframeTypical holdProfit factorVerdict
M1about 7 minutes0.59rejected
M5about 35 minutes~1.0rejected
M15about 100 minutes~1.5went on to the live EAs
M30about 3½ hours~0.8rejected

The shape is the lesson. If shorter timeframes were better, M1 would win. If longer ones were safer, M30 would. Neither is true. The result rises to a peak and falls away on both sides.

Why the Short Timeframes Lost: the Spread

On M1 we used real recorded ticks, 100% real, so the result is not a modelling artefact. USDJPY made a profit factor of 0.65 on 123 trades, EURJPY 1.04, EURUSD 1.20; pooled, 336 trades lost money. The note on the rejection says why: the edge per trade was below the spread.

M5 shows the same thing more slowly. On 2026 real ticks it made about +$0.05 per trade, against a spread of 0.3 to 0.5 pips. On M15, over the same window, the same rule made about +$0.36 per trade.

That is the whole mechanism. Every trade pays the spread once, whatever the timeframe. A shorter timeframe makes each trade’s expected move smaller, but it does not make the spread smaller. Below some timeframe, the move a reversion trade catches is about the size of what it costs to enter, and the edge disappears into the cost. (Which tick mode you test that on matters a great deal, as Every Tick vs Real Ticks explains.)

Why the Long Timeframe Lost: the Pattern Itself

M30 trades less often and holds longer, so the spread weighs less on each trade. It lost anyway. On the same London window, with the settings that worked on M15, EURUSD made a profit factor of 0.28 on 28 trades, EURJPY 0.99, USDJPY 1.27. Our reading: the dip the strategy is built to buy is short-lived, and on half-hour bars much of it is over before a bar closes and the rule can see it.

So the lesson is not “higher is better” either. A timeframe is the scale at which the strategy looks at price, and a pattern lives at a scale. Look too closely and the cost swamps it. Look too far and it is not there.

A Higher Timeframe Will Not Rescue a Bad Entry

We also tested the opposite hope. Our trend-following entry lost on M15: profit factors of 0.92, 0.85 and 0.93 on EURUSD, GBPUSD and USDJPY over 846 trades. A natural thought is that the spread was eating it, and that H1 would fix it, because on H1 the spread is about 5% of the stop instead of about 15%.

It did not. On EURUSD H1 the same entry made a profit factor of 0.53 on 126 trades. The stop took 70% of exits. The higher timeframe removed the cost and exposed the real problem: the entry had no edge to begin with. Both versions are rejected, and their numbers stay on the site.

If a strategy loses on one timeframe, moving it up or down is a test, not a fix. Run it, and believe the result even when it is worse.

When the Timeframe Hardly Matters

Our one proven mechanism, the Tokyo fix on gotobi days, runs on M5 charts. But its rule is a clock, not a bar pattern: sell at the first tick after 09:55 Japan time on a gotobi or month-end day, close 25 minutes later. The chart’s timeframe barely enters into it. The 25-minute hold came from measuring the move after the fix at holding times from 5 minutes to two hours.

That is the third kind of answer. For a strategy driven by a known time or event, ask what the right holding period is, not what the right chart is.

How to Find the Right Timeframe for Your EA

  1. Test the same rule on three or four timeframes, with the same pairs, dates and settings, and change nothing else.
  2. Compare profit per trade with the spread, not just profit factor. If a timeframe’s average trade is close to one spread, it will not survive live costs.
  3. Look for a peak, not a trend. If the results rise steadily in one direction, you have not found the edge yet; test one step further.
  4. Then test the winner out of sample. A peak found on one window can be luck. The live M15 versions of this strategy passed 2023–24 and then lost on 2019–22, so they run only at minimum size under our regime tier. How we hold windows back is in Forward Testing in MT5.
  5. Keep the losers. The rejected M1, M5 and M30 versions are what made the M15 result believable. A single good timeframe with nothing to compare it to proves nothing.

There is no best timeframe for EA trading. There is the scale at which your strategy’s edge is bigger than its cost, and it can only be found by testing, not chosen in advance.

Nothing on this page is investment advice. Every figure is a backtest on one broker’s data, published so it can be checked.

Si te ha servido para tu EA, compártelo.
X Facebook LinkedIn

Keisuke Kurosawa
Hello

Comentarios

0
Todavía no hay comentarios.

Para comentar, inicia sesión.
Compartir
https://1kpips.com/en/blog/best-timeframe-for-ea-trading
Categorías
MT5 Strategy Testing & Optimization
Etiquetas
EA timeframe,M15,Spread,Strategy Tester,MT5 Backtesting,techniques

Artículos relacionados

Siguiente paso
Llévalo a tu EA: añade un filtro de sesión y haz backtest con y sin él para ver la diferencia.