The problem with forex backtests that omit real costs

A backtest shows what a rule set produced under specified data and assumptions. Omitting or oversimplifying trading costs can move results away from feasible execution. The difference is not always a fixed deduction; some costs also affect order counts, fill feasibility and strategy behaviour.

Decision prices and execution prices differ

A midpoint or last-traded price may inform analysis, while transactions face defined bid and ask conditions. Identify the prices used for entry and exit and where spread enters the model. Using one cost-free price for both sides is a material assumption, not an unexplained fact.

A hypothetical fragile gross result

Suppose one hundred trades produce 100 units of gross profit. With a hypothetical average cost of 1.2 units per trade, total costs are 120 and the net result is minus 20. This simplified example assumes constant costs and predicts no market outcome. It shows how small per-trade differences can change aggregate results.

Which costs should be visible?

• Spread appropriate to time and instrument rather than only the minimum.
• Commission with correct units and charged sides.
• Slippage and possible partial execution or rejection.
• Holding charges, currency conversion and related constraints.
• Excluded assumptions and their unknown effects.

One average is not always enough

Costs may differ around news, thin liquidity or the trading-day boundary. If the method concentrates activity at those times, a broad average can mislead. Record sources, periods and scope for each estimate. Do not substitute zero for unavailable data; distinguish bounded scenarios from unknown effects.

Test sensitivity rather than tune the conclusion

Review outcomes under several defensible cost assumptions and show where signs or behaviour change. Selecting only the assumption that preserves profitability is not independent review. Make assumptions clear before interpreting results and record reasons for later revisions.

Costs are only one backtesting limitation

Future-information leakage, selected periods, overfitting and live-execution differences still matter. Adding commission does not resolve them all. Out-of-sample testing guarantees no future outcome either. Read it separately from the rule-selection period and label it as simulation.

Professional reporting preserves sources and assumptions

Keep rule versions, data, costs and limitations alongside the chart. Do not combine backtests with actual accounts or verified performance without labels. Managers can clarify tests versus live history in a cooperation inquiry. In account information, request evidence appropriate to each figure. This article approves no strategy, recommends no live execution and guarantees no predictability of returns.