Alpha testing procedure v0.9

In the course of finding a temporarily working strategy, I have come to the conclusion that I need to do more than just back test the system. There must be more than just a strong back test ( possibly curve fitted ) to give me confidence in the systems longevity and effectiveness .

Finding Alpha:

The first test should be a three¬†month ( for 1H and less ) or eight¬†month (4H and higher) from today, back. More than likely, the shot in the dark variables will be a wash or less. In which case, a slight tweak in optimization is okay. Per strategy rule, there should be no more than 5 possible optimizations per rule, multiplied by the number of rules. So two rules should have no more than 5 * 2 possible optimizations – in any variance. This can not include risk variables, as before where I would check for stop ranges, if the strategy can’t find a good entry, there is no point in continuing the idea.

In this preliminary test, we are looking for a break even or better. Break even could be worked on further ( not optimized ), and obviously a better would be even better.

Walk Forward:

In a walk forward, I will perform 6 sets of two year in sample with one year out of sample . This isn’t traditional, and it will force the optimizations to expose each other, especially if I have already lost my way and over optimized them.

In each set, I am looking to get break even to higher, allowing for the final set to be the optimal ( and highest odd ) for whatever might happen in the next year for that particular strategy .

Final Year Set Data with Monte Carlo:

I got this idea from a book which I thought was ingenious. First, Using the settings from the final set in the walk forward, run a back test for the last year. Take the entire wins and losses, and push the losses altogether. Use this to calculate the risk of ruin, obviously having a less than 25% RoR is satisfactory, and everything in between…. less is best regardless

Incubation Period:

For at least 6 months the strategy shall run on a demo account. It is not super important the strategy is better than expected, or even the same,…. or even break even. What is most important here is that we don’t jump to a live account. We need the dust to settle, for our attention to be drawn to other ideas and strategies , as so we can see this system from a different perspective.

Be the first to comment

Leave a Reply

Your email address will not be published.