Low Expected Payoff EAs: Profit Buried by Costs

Bottom Line: Expected Payoff Shows How Well an EA Withstands Costs

The "Expected Payoff" in an MT5 backtest report is net profit divided by the number of trades, showing how much was left on average per trade. The conclusion first: an EA with a small value here can easily lose its profit when live trading costs run just a little higher than in the backtest.

Even if PF and cumulative profit look impressive, a thin margin per trade means a slight misjudgment of costs can change the results dramatically. When we build EAs, this is something we look at even before PF.

Why a Small Expected Payoff Is Dangerous

Costs Are Deducted Every Time, Without Fail

Trades win and lose, but spreads and commissions are deducted on every single trade. Expected Payoff is the figure after costs, so as a formula it looks like this:

• Actual Expected Payoff = average pre-cost price move − cost per trade

The smaller the pre-cost move, the larger the share that costs take up. As a general illustration, if an EA's pre-cost margin is only about the same as its costs, doubling the costs would push Expected Payoff down to nearly zero. If an EA's margin is several times its costs, the same increase barely changes its results.

Backtest Costs Are Only an "Average Condition"

The spread in a backtest is either the value stored in the historical data or a fixed value. In live trading, deviations like these occur:

• Spreads widen around economic releases and in early-morning hours
• Slippage appears on market orders
• Commission structures differ by broker and account type

Almost all of these deviations work in the direction of reducing Expected Payoff. An EA with a small Expected Payoff has no room to absorb them.

With Many Trades, the Impact Simply Accumulates

For an EA with short holding times and many trades, the cost difference per trade accumulates across the whole trade count. Even among our own EAs, the character of the backtests differs considerably.

Even when PF is similar, a design with many trades and short holding times, like GOLD VIPER, is structurally more sensitive to per-trade cost differences. On the other hand, designs with long holding times, such as DONCHIAN ENGINE and ETHEREUM EA, tend to have larger price moves per trade, so the cost ratio is relatively lower. However, with fewer trades, statistical reliability is lower. This is not about which is better; it is that the type of weakness differs.

Steps to Check Before Choosing an EA

1. Compare Expected Payoff with costs: Put the report's Expected Payoff next to the per-trade cost of the account you plan to use (spread plus round-trip commission converted by lot size). If Expected Payoff doesn't reach several times the cost, be cautious.
2. Re-test with wider spreads: Set the tester's spread larger than the current value and run it again. Check whether the EA stays profitable and where it turns into a loss.
3. Look at average holding time and trade count: The shorter-term and higher-frequency the EA, the more weight you should give the result of step 2.
4. Verify with your own broker's conditions: Even for the same symbol, costs differ by broker. It is safest to check the differences between account types in the broker comparison before deciding.
5. Compare with other EAs in a list: Use the measured ranking to line up trade count, holding time, and PF and see where an EA stands relatively.

Limits of This Metric and Cautions

Expected Payoff alone does not decide whether an EA is good or bad. Here are the inconvenient points as well.

• It is a monetary amount, so it depends on lot size and starting capital. For EAs whose lot size grows through compounding, later trades push the average up, so they cannot be compared simply with other EAs.
• It is an average, so the distribution is hidden. The average may be lifted only by a few large wins. It needs to be read together with maximum consecutive losses and win rate.
• Even with a large Expected Payoff, there are losing periods. We also forward-test our EAs on Exness demo accounts, and some EAs that are profitable in backtests are trending negative in demo trading. Demo results are a different thing from backtests, and they are not the same as live-account execution either.
• How much cost increase an EA must withstand to be considered sufficient is something we cannot define uniformly, either. It varies by symbol, time of day, and broker, so in the end you will need to verify it in your own environment.

Summary

• Expected Payoff shows how much remains per trade after costs
• Trading costs are deducted every time, so the smaller an EA's Expected Payoff, the more vulnerable it is to changes in costs
• Spread widening, slippage, and commissions in live trading mostly work in the direction of reducing Expected Payoff
• The shorter-term and higher-frequency the EA, the more the per-trade cost difference accumulates over the trade count
• Always re-test with wider spreads and verify under your own broker's conditions
• Expected Payoff is not all-purpose; judge it together with distribution, losing streaks, and how lot size is handled