Anchored Walk Forward Optimization to Avoid Curve Fitting

[fiverr]

Gents,

I have the following WFO result and it looks promising. What parameters should I pick and how often should I re-optimize my parameters?

 

[fiverr]

In addition to the out-of-sample testing, I would suggest some bias-correction measures, which are rooted in the information theory.

For example, you may consider something like this:
BCP = (P * SQRT(T)) / (D * D)
where
BCP is bias-corrected performance
P is the raw in-sample performance, such as Sharpe ratio
T is the number of trades
D is the dimension of the model (6 in your case)

So, in-sample, you want to find a set of parameters which maximize BCP.

For motivation behind this formula, take a look at the Akaike Information Criterion.

Many thanks, you are awesome.

 

[nonlinear5]

Gents,

I have the following WFO result and it looks promising. What parameters should I pick and how often should I re-optimize my parameters?

It's difficult to see what's going on here in your tabulated results. You have a 6-parameter optimization space, and I would suggest you find a way to visualize that parameter space. Below is how I do it, using the heat map. The x-axis is a strategy parameter, the y-axis is another strategy parameter, and the color indicates the system performance. The "hotter" the color, the better the performance.

What you want to find is what I call a "plateau of elevated performance". That is, you want to see a sustained, continuous region of out-performance (as opposed to random spikes).