Testing Portfolio Success Rate

[Anonymous]

Hey team,

Thought I’d just pop in another post here. I’m quite new to EA Studio – however, I quite often see that others will remove poorly performing strategies from their accounts and not move them to live. I was wondering whether this was the best move given that you develop a portfolio so that one strategies losing streak can be compensated by another strategies winning streak. Surely by removing some strategies that are performing poorly at that given moment, we are altering the portfolio and it’s purpose.

A bit of a ramble, but I thought I would try the following as a proof of concept – and to check whether anyone has done a similar process.

I was going to produce a portfolio of 30 EA’s for a currency pair. Let’s use EURUSD as an example. I’ll perform the following:

• When generating the strategies I’ll make sure that I set the END DATE in the horizon tab to be one month prior to today’s current date. I’ll then perform a generation test with 10% OOS. I’ll pick the best 30 strategies from this result. Especially those that showed promise when OOS.
• I’ll now load that collection and change the horizon date to be TODAY and then refresh the collection to see how it would’ve performed for the last month.

I’ll be interested to then see what percentage of strategies in that collection performed well. Whether it was only 30% of them, 50% etc. From this perspective, I should gain a better understanding of how quickly you can determine whether a strategy is going to perform well or not.

Has anyone performed this before, and can shine any light on the results? I am currently going through the generation process so it’ll be another 1-2 days before I can test this theory.

Thanks,

Stephen

[Author]

Posts