Definitions for "Overfitting"
The parameters of a trading system are selected to return the highest profit over the historical data. A model developed with rules tailored to fit the historical data precisely.
Using too many covariates for the number of outcome events in a multivariable predictive model; can lead to spurious or incorrect associations.
Using the highest profit point for the selection of parameters of a trading system over a specific time period without determining the likelihood that the parameters will retain their profitability over other periods of price data. Also known as "Curve Fitting".
Behavior displayed by ANNs when they are unable to generalize.