Hi,
Are there any threads or help guides (strategic or tactical) to triangulate a "more" optimized strategy configuration in a given optimization period via a walk forward optimization (WFO) that would produce better long term results?
For example, I am running a WFO between 01/01/2022 to 12/31/2022 with an Optimization Period of 200 days and a Test Period of 30 days, optimizing on Profit Factor.
The strategy will optimize on my inputs for 200 days after 01/01/2022 and select the first strategy with the highest Profit Factor. The WFO will then use that configuration for the next 30 days (as a back test) to determine how well that strategy would have performed. It will then iterate until 12/31/2022.
If there were, let's say, 4 strategy configurations that returned the same profit factor, the settings will default to the first one.
However, I have found with some testing, the first strategy may actually return worse results in practice when compared to the other 3 strategies that were skipped.
1. What are the exact rules in place that determine which is the "best" strategy? It looks like there is perhaps some sort of stack ranking system that weighs number of trades, profit factor, win rate, etc... for each strategy. Is that the case? If so, can that be further investigated and/or adjusted?
2. Is there a way to systematically and statistically identify better strategies (i.e., how can we adjust the WFO strategy selection process so that we incorporate and utilize the information from, again, let's say, any of the 2 - 4 strategy options in my example above)?
3. Are there any known best practices or guidance that the team could provide in this regard?
Happy to provide more context if needed. Thank you in advance.
Are there any threads or help guides (strategic or tactical) to triangulate a "more" optimized strategy configuration in a given optimization period via a walk forward optimization (WFO) that would produce better long term results?
For example, I am running a WFO between 01/01/2022 to 12/31/2022 with an Optimization Period of 200 days and a Test Period of 30 days, optimizing on Profit Factor.
The strategy will optimize on my inputs for 200 days after 01/01/2022 and select the first strategy with the highest Profit Factor. The WFO will then use that configuration for the next 30 days (as a back test) to determine how well that strategy would have performed. It will then iterate until 12/31/2022.
If there were, let's say, 4 strategy configurations that returned the same profit factor, the settings will default to the first one.
However, I have found with some testing, the first strategy may actually return worse results in practice when compared to the other 3 strategies that were skipped.
1. What are the exact rules in place that determine which is the "best" strategy? It looks like there is perhaps some sort of stack ranking system that weighs number of trades, profit factor, win rate, etc... for each strategy. Is that the case? If so, can that be further investigated and/or adjusted?
2. Is there a way to systematically and statistically identify better strategies (i.e., how can we adjust the WFO strategy selection process so that we incorporate and utilize the information from, again, let's say, any of the 2 - 4 strategy options in my example above)?
3. Are there any known best practices or guidance that the team could provide in this regard?
Happy to provide more context if needed. Thank you in advance.
Comment