Platforms & AutomationOver-OptimisationOverfitting
Curve Fitting
Tuning a strategy so closely to historical data that it captures noise instead of a real edge, and fails on new prices.
What Curve Fitting means
Curve fitting is what happens when a strategy is adjusted until it fits one particular history extremely well. Every optimisation pass, every added filter and every hand-picked parameter value increases the risk. The tell-tale signature is an equity curve that looks almost perfect in the tested period and then breaks down immediately afterwards, because the rules encoded the accidental details of that sample rather than any repeatable market behaviour. It is the single most common reason automated retail strategies fail after launch.
Several habits reduce the risk without eliminating it. Keep the number of free parameters small relative to the number of trades in the sample, since a system with eight tuned inputs and forty trades is fitting almost by definition. Prefer parameter values that sit on a broad plateau of acceptable results rather than a sharp peak, because a peak means small changes in the market destroy performance. Hold back out-of-sample data, use walk-forward analysis, and check that the rules still make economic sense.
The honest caveat is that curve fitting cannot be fully avoided, only limited. Any strategy chosen because it performed well on history has been selected using that history, and testing many candidates on the same data set is itself a form of fitting even when each individual test looks clean. This is why forward testing on unseen prices matters more than any in-sample statistic, and why a modest, robust edge is usually more durable than a spectacular optimised one.
Worked example
A system optimised to buy when RSI drops below 27.3 and exit after exactly 43 bars is almost certainly curve fitted; if changing the threshold to 30 or the exit to 40 bars destroys the results, the parameters were fitted to noise.
Related terms
- BacktestingReplaying a strategy's rules over historical price data to estimate how it would have performed before risking real money.
- Forward TestingRunning a strategy on live, unseen prices in demo or at minimal size to check that backtested behaviour survives real conditions.
- Algorithmic TradingUsing coded rules rather than discretion to generate signals, size positions and route orders, with the computer executing decisions.
- ExpectancyThe average profit or loss a system produces per trade given its win rate and its average win and loss sizes.
- Maximum DrawdownThe largest peak-to-trough equity decline recorded over a given period of trading or testing.
Frequently asked questions
What does Curve Fitting mean in forex trading?
Tuning a strategy so closely to historical data that it captures noise instead of a real edge, and fails on new prices.
How does Curve Fitting work in practice?
Several habits reduce the risk without eliminating it. Keep the number of free parameters small relative to the number of trades in the sample, since a system with eight tuned inputs and forty trades is fitting almost by definition. Prefer parameter values that sit on a broad plateau of acceptable results rather than a sharp peak, because a peak means small changes in the market destroy performance. Hold back out-of-sample data, use walk-forward analysis, and check that the rules still make economic sense.
What is an example of Curve Fitting?
A system optimised to buy when RSI drops below 27.3 and exit after exactly 43 bars is almost certainly curve fitted; if changing the threshold to 30 or the exit to 40 bars destroys the results, the parameters were fitted to noise.
Trade with a regulated broker
Understanding the terminology is the cheap part. The expensive part is choosing a counterparty whose execution, financing and withdrawal behaviour match what the marketing implies. Every broker below has been reviewed with a funded live account, and each review states which legal entity and which regulator applies to the account you would actually open.
Check the licence on the regulator's own register before you deposit — our regulators directory explains what each authority enforces, from leverage caps to compensation limits.