What costs should your backtest assume?

Use what your account actually charges at the worst time you intend to trade, not the average and not the advertised figure.

Costs are not a haircut, they are the test

A strategy holding positions for weeks barely notices its costs. A strategy holding them for minutes is almost entirely a bet that its edge exceeds them, which means the cost assumption is not a detail of the test, it is the test.

The shorter your holding period, the more of your result is decided by numbers you typed in rather than by anything the market did.

Take the numbers from your own account

Advertised spreads are typically the best case on the most liquid pair at the quietest hour. What you need is the spread on your instrument, at the hours your strategy trades, on your account type, including the commission your account pays and the financing on anything held overnight.

Record it for a couple of weeks before you trust it. The observation costs nothing and replaces a guess in the one place a guess is most expensive.

The worst case, not the average

Averages hide exactly the conditions that hurt: the widening at the open, the session close, and the seconds around a release. Strategies that trade at those moments meet the bad case disproportionately, because that is when their signals fire.

Test at the worst spread you have observed. If the edge only survives at the average, you have not found an edge, you have found a claim about liquidity.

Slippage is directional and it is not random

It is not symmetric noise. Stops are hit in fast markets moving against you, which is when slippage is worst and most one-sided. A model that applies the same slippage to entries and exits, or that treats it as zero-mean, will read better than reality.

Charge more slippage to stop exits than to planned entries, and more to anything triggered by a sudden move.

Then test whether it matters

Re-run at double your assumed costs. If the strategy survives, your cost assumption is not carrying the result and you can stop worrying about it. If it collapses, the assumption was the strategy, and you now know the single number to get right before risking anything.

Questions

Should I use a fixed or a variable spread in a backtest?

Variable, if the tester supports it, because that is what the account trades. If only a fixed spread is available, use the widest one you have observed during the hours your strategy trades rather than the average.

How much slippage should I assume?

There is no universal figure: it depends on the instrument, the session and how your orders are triggered. The defensible method is to measure your own fills against your intended prices, and until you have that, to test at a level high enough that being wrong does not change the decision.

Do commissions matter if my broker advertises zero commission?

The cost is in the spread instead. Compare total cost per round turn across account types rather than the headline, because a raw-spread account with commission is often cheaper overall for a frequent strategy.

Sonar Sciences is a research and validation platform, not an investment adviser. Nothing on this page is investment advice or a recommendation to trade any instrument or strategy. Platform names are trademarks of their respective owners and we are not affiliated with them. Trading leveraged instruments involves substantial risk of loss.