Why the Same EA Looks Great on M15 and Terrible on M1

Why the Same EA Looks Great on M15 and Terrible on M1

19 十一月 2026, 02:03
Xiang Jun Jiang
0
3
A user sent me two tester reports last month for the same EA: one on M15, one on M1. The M15 report showed a smooth equity curve and a profit factor of 1.6. The M1 report had the same code, the same symbol and the same period, and lost money in every month. He asked which report was "real". The honest answer is that both were real, and that the difference is not the code but what the timeframe actually changes.

Timeframe changes four things at once

When you move an EA from M15 to M1, you are not just giving it more bars. You are changing the noise-to-signal ratio, the number of decisions per hour, the cost per decision, and the definition of a "bar close". Each of those matters, and they push in different directions.

Noise first. On M1, a large share of bars are pure microstructure: a few ticks up, a few ticks back. Any rule that interprets a bar break as a breakout will fire constantly on M1 and rarely on M15. If the rule has no additional confirmation, the extra signals are not extra opportunity, they are extra transactions paying spread.

Cost per decision second. This is the part most people underestimate. A scalping EA on M1 can take twenty trades an hour. At 0.01 lots on gold, a round turn at 25 points of spread costs about 0.25 USD. Twenty trades an hour is 5 USD an hour in spread alone. On M15 the same rules may take two trades an hour, which is 0.50 USD. The strategy did not change, the tax on it changed by a factor of ten.

Bar-close logic third. Many EAs wait for a bar to close before confirming a signal. On M15 that means the EA sees a decision every fifteen minutes. On M1 it sees a decision every minute, and each decision is based on far less information. A filter designed to reject weak setups on M15 has almost nothing to work with on M1.

Position management fourth. Trailing stops and break-even logic are expressed in points, but the distance they need to travel is a function of volatility, not of the chart. On gold, an average M1 range might be 15 to 30 points while an M15 range is often 100 or more. A 60 point trailing stop that gives the trade room on M15 will be hit within minutes on M1, converting a trend trade into a series of small losses.

What to check before blaming the code

Run the same EA on three timeframes and compare four numbers: trade count, average trade duration, average profit per trade, and the ratio of spread cost to gross profit.

If the trade count multiplies by five when the timeframe drops by fifteen, you are trading noise. If the average profit per trade falls faster than the trade count rises, the extra trades are cost, not edge. If the average duration drops below two bars of the tested timeframe, the exit logic is probably too tight for that resolution.

Then do the arithmetic explicitly. Take the spread in points, multiply by the number of round turns, and compare it to the gross profit in the report. If spread consumes more than a third of gross profit, the strategy is fragile no matter how good the equity curve looks.

What actually helps

There are only three honest fixes. First, scale the stop and trail distances with ATR of the tested timeframe instead of using fixed points. Second, add a minimum bar-range or volatility filter so that M1 bars that are mostly noise are ignored. Third, accept that some strategies only work on one resolution and stop trying to force them elsewhere.

The reverse is also true: an EA that looks mediocre on M15 can look excellent on H1, not because it got smarter but because the noise it was fighting disappeared. That is not a reason to move it to H1 forever; it is a clue about what the strategy actually needs.

One practical rule

Before changing the timeframe, write down what the strategy needs: how much room it needs for a stop, how long it expects a trade to last, and how many decisions per day it can afford at your spread. Then pick the timeframe that satisfies all three. If no timeframe satisfies them, the problem is the strategy, not the chart.

The M1 report was not a lie and the M15 report was not a marketing trick. They were two different businesses: one with ten times the transaction tax and one tenth of the information per decision.