Hello,
I have two questions concerning Walkforward Optimization, and I'm specifically seeking insights from seasoned backtesters. While opinions are valuable, I'm particularly interested in feedback from those with years of hands-on experience who can offer well-informed perspectives.
I'd love to hear your expert opinions on the best practices for setting the intervals and steps for each parameter when running Walkforward Optimization, especially in the context of intraday strategies.
Thank you for your time, and I eagerly await your insights.
What are the Best Practices for Interval and Step Settings in Walkforward Optimization (for intraday strategies)?
In Walkforward Optimization, parameters are first optimized using in-sample data during the "Optimization Period (days)." These optimized parameters are then tested on out-of-sample data in the "Test Period (days)." My initial hypothesis is that using smaller steps for the optimization could risk overfitting. However, since the highly optimized parameters are ultimately validated on out-of-sample data, it may not be a significant concern.
Small steps:
Large steps:
What is best practice on "Optimization Period (days)" and "Test period (days)" when running WFO if you build intraday strategies?
Turning to the specifics of the "Optimization Period (days)" and "Test Period (days)" in Walkforward Optimization, I primarily focus on intraday trading strategies. My hypothesis is to keep these periods relatively short, given that market conditions can change rapidly. I've experimented with both 90-day and 365-day durations for the "Optimization Period (days)" and "Test Period (days)." Interestingly, my strategies have yielded better results in the WFO tests when optimized over 90 days as opposed to 365 days. Additionally, aligning the 90-day cycle with contract rollovers offers a practical advantage.
I have two questions concerning Walkforward Optimization, and I'm specifically seeking insights from seasoned backtesters. While opinions are valuable, I'm particularly interested in feedback from those with years of hands-on experience who can offer well-informed perspectives.
I'd love to hear your expert opinions on the best practices for setting the intervals and steps for each parameter when running Walkforward Optimization, especially in the context of intraday strategies.
Thank you for your time, and I eagerly await your insights.
What are the Best Practices for Interval and Step Settings in Walkforward Optimization (for intraday strategies)?
In Walkforward Optimization, parameters are first optimized using in-sample data during the "Optimization Period (days)." These optimized parameters are then tested on out-of-sample data in the "Test Period (days)." My initial hypothesis is that using smaller steps for the optimization could risk overfitting. However, since the highly optimized parameters are ultimately validated on out-of-sample data, it may not be a significant concern.
Small steps:
Large steps:
What is best practice on "Optimization Period (days)" and "Test period (days)" when running WFO if you build intraday strategies?
Turning to the specifics of the "Optimization Period (days)" and "Test Period (days)" in Walkforward Optimization, I primarily focus on intraday trading strategies. My hypothesis is to keep these periods relatively short, given that market conditions can change rapidly. I've experimented with both 90-day and 365-day durations for the "Optimization Period (days)" and "Test Period (days)." Interestingly, my strategies have yielded better results in the WFO tests when optimized over 90 days as opposed to 365 days. Additionally, aligning the 90-day cycle with contract rollovers offers a practical advantage.
Comment