Why Backtest Profits Rarely Survive Live Spreads

A pip of spread sounds trivial next to a twenty pip target. It is trivial next to the target and it is not trivial next to the thing that actually pays you.

Why backtest profits rarely survive live spreads comes down to that distinction, and the arithmetic is worth running once properly.

Spread Is Measured Against the Edge, Not the Trade

Take a strategy winning fifty five percent of the time, with twenty pip winners and fifteen pip losers.

Expectancy is a fraction over four pips per trade. That is the number the strategy generates, and it is considerably smaller than either the winner or the loser that produced it.

Now add spread. One pip costs you a pip on every winner and adds a pip to every loser, which reduces expectancy to three and a quarter. A reduction of nearly a quarter from a single pip.

At a pip and a half, expectancy falls to two and three quarters, down more than a third. At two pips it is halved. At three pips nearly three quarters of the edge is gone.

Nothing about the strategy changed. Entries, exits and win rate are identical. Only the cost of participating moved, and because it comes out of the edge rather than the trade, a small absolute figure produces a large proportional loss.

Why Testing Understates It

Several reasons, and they compound.

Most testing applies a fixed spread rather than the varying one the market supplies. Some modes apply the current symbol spread across the whole historical period, which is the spread today rather than then.

Genuine historical spread data is rarely available, so even careful testers model rather than reproduce.

Raw accounts complicate it further. If your account quotes tight spread plus commission per lot, a test modelling only spread has omitted a cost entirely.

The Variable Spread Problem

Average spread understates your real cost whenever your entries cluster in expensive periods.

Spread widens at rollover, around releases and in thin hours, so a strategy trading those windows pays well above the average your test assumed.

Session open strategies suffer most. The opening minutes carry the widest spreads of the day, which is exactly when they are designed to act.

So the relevant figure is not the instrument’s average spread but the average during the minutes your strategy enters, which is usually larger.

Measure It Rather Than Assuming

Log the spread at the moment of every entry over a few weeks of live or demo running.

Then recompute your expectancy using the observed distribution rather than the modelled figure. That calculation tells you whether the strategy survives its own costs, and it is the single most informative thing you can do with a logged dataset.

If expectancy after real costs is marginal, the conclusion is clear. Optimisation does not fix a cost problem, because cost is not a parameter. Either the targets get larger, the frequency drops, or the strategy is not viable on that instrument.

The Implication for Strategy Design

Spread sensitivity scales with turnover. A system taking two trades a week barely notices a pip. A system taking ten a day is paying that pip ten times daily against an edge measured in single pips.

Which is why small target strategies rarely survive real costs, and why a longer timeframe sometimes rescues a method that looked broken.

Conclusion – Why Backtest Profits Rarely Survive Live Spreads

Why backtest profits rarely survive live spreads is that spread is subtracted from the edge rather than from the trade. A pip and a half can remove a third of a perfectly reasonable expectancy. Measure the spread your entries actually encounter, recompute, and accept the answer.

FAQ – Why Backtest Profits Rarely Survive Live Spreads

1.How much does spread actually cost me?

Compare it against your expectancy rather than your target. A pip and a half against a four pip edge removes more than a third of it.

2. Why does my backtest not reflect this?

Most testing applies a fixed or current spread rather than historical variable spread, and raw accounts add commission that spread modelling omits entirely.

3. Can optimisation fix a cost problem?

No. Cost is not a parameter. The answer is larger targets, lower frequency, or a different instrument.

We have helped thousands of traders reach funding at TTT Markets from account sizes of $5k upwards to $500k. Check out our programs. 

Additional resources:

Why a Profitable Backtest Loses Live: Reconciliation Guide 

Backtesting vs Live Trading: 4 Reasons Why Your Results Don’t Match | EBC Financial Group