Hey Mark can we get a very simple function built into crosschecks?
This would be: Number of additional markets tests resulting in greater than zero $. Value is "3" for example
I just did a little study and found that if a strategy can be greater than zero in net profit on 3 additional markets it has a much higher probability of passing MC and other tests. The key is that is doesnt really matter which markets pass, just that 3 do out of 6 for example.
Attachment additional_markets_filtering.png added
It would also be very useful if we could time-slice each of those markets into say 5 or 10 equal times periods. We could then say: Passes x% of time-slices in a market, for x number of markets. Or simply say of all time-slices across all markets, X% of the time slices must pass.
By using time-slices you can derive greater statistical significance from the same number of markets. For example, performing this test on 3 markets, gives 3 results. But performing this test on 3 markets, each time-sliced into 5 equal times periods, you now have 15 tests and you can filter strategies that pass say 10 of these 15 tests. This also helps to avoid book end profits that can give false-positives (pasess) when looking at just 3 markets and their overall result.
Or another extreme example. As we slice smaller and smaller we would converge on win%/loss%.
But I guess slicing is good as long as we still have reasonable statistical significance in each slice...
If we could simply say 3/4 additional markets in the retest must pass this would cut down on a very significant amount of development time.
Time slices would be even nicer, but if the 3/4 (x/y) markets filters cannot be done within 6 months of the original OP then I am not hopeful.