Why transaction costs change a backtest
A line that looks profitable on paper can disappear once you pay to get in and out of the trade.
What a “cost” is
Every completed round trip has a price that is not the mid-market print you see on a chart. The usual pieces are commission, the bid-ask spread, and slippage (you get a worse fill than the signal price). Borrow fees matter if you short. Overnight financing matters if you use leverage.
Researchers often quote a single round-trip number, for example 0.10% or 0.15% of notional, and subtract it from every entry-plus-exit. That is a model, not a promise that your broker will charge exactly that.
Why small edges die first
If a study shows an average gain of 0.20% per trade before costs, and you assume 0.15% round-trip costs, most of the edge is gone. If the true cost is 0.25% on a jumpy name, the same study is a losing process.
High-turnover ideas are more sensitive than low-turnover ones. A strategy that trades once a month can survive a 0.15% haircut. A strategy that trades ten times a day usually cannot, unless the average win is large enough to pay the toll.
How to read a result
Ask three questions of any backtest you are shown. Was a cost applied on every trade? Was the cost large enough for that name and that venue? Did the author also show the zero-cost version so you can see how much of the headline return is just “free fills”?
A honest study will say the cost assumption in plain language and will not hide that some names are more expensive to trade than quiet large-caps.
Key takeaways
- Costs are part of the result, not a footnote.
- Small average wins plus high turnover is the first place a study breaks.
- The cost number is an assumption. Treat it as a stress test, not a forecast.