How Far Does an EA's Performance Fall When Spreads Double — A Sensitivity Analysis Method

Bottom line: whether an EA can withstand it comes down to "how much it pockets per trade"

When spreads double, one EA barely changes its performance while another flips from profitable to unprofitable. This gap has nothing to do with which strategy is "better" — it comes down to simple division: what fraction of the average profit per trade gets eaten up by the added cost.

If an EA that averages 10 pips per trade sees its spread go from 1 pip to 2 pips, the added cost equals 10% of its average profit. For an EA averaging 3 pips, that same increase wipes out 33%. Even though the "spread doubles" in both cases, the amount shaved off is completely different.

Before we publish any EA, we always run this spread sensitivity analysis. Here's the reasoning behind it, along with a procedure you can run yourself.

The mechanism: degradation is just "added spread × number of trades"

The total loss caused by spread deterioration follows a surprisingly simple formula.

Degradation ≈ Added spread (in currency terms) × Total trade count

In other words, only two variables matter: trade count and profit per trade. From here, it becomes clear which strategy types are more fragile.

To put it concretely: GOLD KING v6 (XAUUSD M1) recorded 2834 trades over a 7.5-year backtest, and GOLD NEURON (XAUUSD M15) recorded 765 trades over 7 years. By contrast, BITCOIN GLACIER (BTCUSD D1) had only 28 trades over 8 years, and BITCOIN COIL (BTCUSD H4) had only 57 trades over 8 years. Even with the same spread widening, paying it 2834 times versus 28 times means a 100x difference in total cost.

What makes this even trickier is that high-win-rate, thin-margin EAs tend to be the most exposed. GOLD PIVOT (XAUUSD M1) has a 90.6% win rate, and GOLD KING v6 has an 83.1% win rate — but a high win rate is usually the product of a "small gains, taken often" design. When those small gains get eaten by rising costs, the win rate itself starts to fall. Winning trades slip into breakeven exits, and breakeven exits slip into losing trades.

Trend-following EAs like MEGAMAX DON EURJPY (EURJPY H1, 25.2% win rate, 571 trades), on the other hand, are designed from the start to make up for a low win rate with large profit margins. Since a handful of big winners carry the returns, the return structure holds up even if entry costs rise somewhat. Here, a low win rate is actually a form of defense.

Procedure: running a spread stress test in MT5

Once you understand the logic, it's time to verify it with actual numbers. This isn't a difficult process.

1. Establish a baseline test

First, run the Strategy Tester using the "Every tick based on real ticks" model and save the baseline results. Be sure to record four figures: net profit, PF, total trade count, and average winning trade amount. The average winning trade in particular becomes the denominator for later estimates.

2. Raise the spread in fixed steps

Change the tester's spread setting from "current" to a fixed value, and run it at four stages: the assumed real-world value, 1.5x, 2x, and 3x. For volatile instruments like gold or Bitcoin, it's worth checking up to 3x — spreads can genuinely widen to several times their normal level around news releases or in early-morning hours.

3. Watch how PF declines, not the absolute PF

What matters isn't the absolute value at each stage, but the slope. An EA whose PF barely moves at 2x has plenty of margin relative to costs. An EA whose PF drops toward 1.0 at 2x means its current profitability depends on the spread environment. That isn't necessarily a flaw in itself, but you need to recognize it as an EA that can only run well on a favorable-condition account.

4. Add in commissions too

ECN accounts offer narrow raw spreads but charge a round-trip commission. If you don't enter the actual commission value in the tester's commission field, you'll misjudge "narrow spread = advantageous" by looking at spread alone.

5. Measure the real-world spread on the account you'll use

You need to measure the assumed value yourself. Install a monitoring script on the account you plan to run and record the average and maximum over one week of weekday trading. Points to consider when choosing a broker are summarized in Broker Comparison. If you're about to set up an EA, see the EA Setup Guide for the procedure.

Limitations: what this analysis cannot tell you

To be candid: a spread stress test only reproduces part of the real-world operating cost.

• Slippage is a separate issue. Fixed spreads in the tester don't account for execution delay or slippage. In reality, cost deterioration comes from spread widening and slippage happening together.
• It can't reproduce the timing of widening. A fixed-value test applies the added cost evenly across the whole period, but in reality spreads widen selectively during volatile price moves. If that coincides with the conditions an EA tends to trade, the real-world damage will be worse than the fixed-value test suggests. We haven't been able to quantify this gap.
• Instrument specification differences need separate checking. Contract size, minimum stop level, and swap rates vary by broker and affect performance independently of the spread.
• Past spread conditions don't guarantee the future. If the underlying liquidity structure changes, the normal baseline level itself can shift.

So use this test not as "a tool for finding a safe EA," but as "a tool for understanding how dependent an EA is on the cost environment." Measured figures for each EA are published in our Full EA Performance Ranking.

Summary

• The impact of spread deterioration is determined almost entirely by "added spread × trade count." It's a matter of design, not strategy quality.
• EAs with many trades and small per-trade margins are the most fragile. GOLD KING v6's 2834 trades versus BITCOIN GLACIER's 28 trades mean the total cost differs by an order of magnitude even under the same widening.
• High-win-rate EAs tend to be thin-margin by design, so rising costs can drag the win rate itself down.
• Trend-following EAs earn through profit margin rather than win rate, giving them relatively more resilience.
• The procedure: baseline test → 1.5x, 2x, 3x → watch the slope of PF decline → factor in commissions → measure real-world spread on your actual account.
• That said, slippage, widening timing, and instrument specification differences can't be reproduced. The test isn't a cure-all — it's simply a tool for measuring dependency.

You'll find our list of free EAs here, and backtesting and MT5 operation know-how in our Knowledge base.