Every time you enter and exit a position you pay something. On most retail products it arrives in three pieces.
The spread is the gap between the price you can buy at and the price you can sell at. You pay it the instant you open, because you buy at the higher number and the position is immediately valued at the lower one. Commission is the explicit fee, if your account has one. And overnight financing, called swap on many platforms, is what you pay or receive for holding a leveraged position past the daily rollover.
Backtests are often silent on all three, or use a default that has nothing to do with your account. That is the single most common reason a strategy that looked profitable in testing loses money live.
The arithmetic, worked
Take a simple example with round numbers. Say a strategy averages a 10 pip gain per winning trade and risks 10 pips per loser, and wins 55 percent of the time. Before costs, 100 trades gives you 55 wins and 45 losses, so 550 pips in and 450 pips out. You are 100 pips ahead.
Now add a round-trip cost of 2 pips. Every one of those 100 trades pays it, wins and losses alike, so that is 200 pips. Your 100 pip profit is now a 100 pip loss, and the strategy that looked like it had a 55 percent edge is simply a way to pay your broker 200 pips.
Nothing about that example is unusual. It is the ordinary situation for short-horizon strategies, and it is why so many of them are hopeless in a way no amount of parameter tuning can repair.
The rule of thumb
The move you are trying to capture needs to be substantially larger than the round-trip cost of capturing it. Several times larger, not slightly larger. When the target is close to the cost, small changes in your fill quality decide whether you make money, and fill quality is the thing you control least.
This is also why the same idea can work on one timeframe and be impossible on another. Costs are roughly fixed per trade, so a strategy holding for two days pays the same spread as one holding for two minutes while aiming at a far bigger move. Trading more often does not multiply an edge. It multiplies the toll.
What to check on your own backtest
- Find the cost assumption in your testing software and compare it with your broker’s actual figures for the instrument you trade. Many defaults are optimistic, and some are zero.
- Check the spread during the hours you actually trade, not the advertised average. Spreads widen around news and at session boundaries, and if your entries cluster there, the average is not your cost.
- If you hold overnight, get the financing figures for both directions. They are usually asymmetric, and on a strategy that holds for days they can matter as much as the spread.
- Re-run at double your assumed cost. If the result falls apart, your edge was a cost assumption rather than a strategy.
There is a longer write-up of a real case on the blog, where a strategy’s losing years turned out to be almost entirely a cost problem rather than a market one: the bad years were a toll booth. The mechanical version of the same check, run on an uploaded trade log, is the transaction-cost test.
Next: how many trades you actually need before a result means anything.