Curve Fitting
Tuning a strategy's parameters until it fits historical data closely, producing results that do not repeat on new data.
How it is identified
Test: performance degrades sharply when parameters are perturbed slightly or when the rules are applied to unseen data
Unit
qualitative
In depth
Also called overfitting, this is the central failure of quantitative strategy development: with enough parameters, any historical series can be fitted almost perfectly while containing no predictive content whatsoever. The diagnostic is parameter sensitivity — a robust rule performs similarly at 18, 20 and 22 periods, while a fitted one collapses outside its tuned value. Testing many combinations and reporting the best is a form of multiple-comparison error, and the more combinations tested, the more certainly the winner is noise. The defence is fewer parameters, out-of-sample testing, and an economic reason for the rule to work.
Worked example
A system returns 34% at a 19-period setting, 6% at 18 and 4% at 20. The peak is almost certainly an artefact of the specific history tested; a robust parameter would show a broad plateau rather than a single spike.
Illustrative figures, chosen so the arithmetic is easy to follow. Not a live price and not a valuation of any company.
Educational reference only
This entry explains what “Curve Fitting” means. It is not investment advice and not a recommendation to buy or sell any security. Any numbers above are illustrative, not live prices, and nothing here predicts price direction or rates a stock. Consider your own circumstances and consult a SEBI-registered investment adviser before acting.