Optimization and fitness function for robustness vs for performance

Optimization and fitness function for robustness vs for performance

Kiran
2015-04-11 20:56:59

Several systems end up being over-optimized and seldom perform out-of-sample.

Is there a fitness function that would help ensure robustness of the system (vs performance) and would yield the best out-of-sample performance?

Another way is to have the PBIL/Genetic optimizer to "hunt down" combinations that also perform well out-of-sample to ensure robustness, before presenting the prioritized results.

Kiran

QuantShare
2015-04-13 02:43:24

Unfortunately, there is no such magic fitness function. Your systems end up being over-optimized because of its rules/logic. That is what you need to change, not the fitness function.

Get more messages

No more messages

Back


Trading Items Average Stock Performance for each U.S Exchange Advance-Decline-Unchanged issues for NYSE, AMEX and NASDAQ 5-Minute Historical Intraday Data for Forex, Indices and Futures/... EOD Quotes Data for Hong Kong Stocks and ETFs Historical EOD data for LSE-listed Stocks and ETFs	How-to Lessons How to download EOD quotes for active and valid stocks only How to screen for stocks having a high correlation with the Dow J... How to get fundamental data for U.S. Stocks How to get stocks for a particular index using the global script How to add future bars to plan for market scenarios
Related Forum Threads In-Sample and Out-Of-Sample Dates for Optimization Custom Fitness Functions for Walk Forward Testing Custom Multi-Dimensional Fitness Formula (CMDFF) for Optimizer indicator for 1 symbol in screener and simulation Use AI optimization (PBIL/Genetic) for walk-forward optimization	Blog Posts Rules performance for different volatility regimes How to Scan for Parallel Resistance and Support Lines How to get buy and sell orders for a portfolio based on a trading... Step by step on how to get free realtime/delayed data for stocks,... Creating and managing notes for your stocks