strange behavior of genetic engine

from B125 we can use in genetics IS Validation data…for me its nothing more, only another type of OOS type of date, and with only one thing added, where i can restart genetics algorithm by the ISV

BUT

if i have the same building config, in one i am using only OOS, on second one, i will change OOS to ISV and restart genetics by ISV and ranky by ISV – so i am changing only one thing – the restarting of evolution

i am getting very different number of accepted strategies

OOS – 2500 accepted strategies per hour

ISV – 6200 accepted strategies per hour

AND I DONT GET IT WHY THIS SIGNIFICANT CHANGE, because the restarting of evolution happens only the same for ISV or IS whole AND I DONT THINK, THAT I AM GETTING WITH ISV MORE DUPLICATES TO DATABANK

SOMETHING IS VERY STRANGE

take a look at those buidling configs a try for yourself

build configs

but if I’m not mistaken they have different building blocks…148   and   119

I checked this  – the difference is in filters. With the ISV setting you have also conditios for IS (In Sample).

But – if you don’t have Out of Sample part defined, then In Sample part is the whole data. In Sample Validation (ISV) is a subset of this.

So in your configs you:

• have OOS defined, hence IS part is about half the data
• don’t have OOS defined, instead of this you have In Sample – Validation defined. Hence IS is full range of data.

If you want to use your config with ISV as same as your OOS config you have to replace the conditions where you use whole IS with IST.

understood – so i need for example

IS/OOS or IST/OOS – this will be the same

and if i want to use ISV, i need to set

IST/ISV

right?

yes.

IS part consits of IST + ISV parts.

So genetic builder is checking it,s own building process performance using ISV against a fitness moving average Which seems to improve amount of found strategies per hour.