Data Range parts and Fitness rank
6 months ago #282346
I noticed that once the available data has been partitioned into IS and OOS (random generation), the Strategy Quality ranking (Fitness) function which selects, once the databank has already reached its capacity, which strategies to add and which to discard according to the chosen criterion, does on the totality of the data, i.e. IS+OOS and not just IS.
Isn’t there a lot of bias introduced this way?
What sense then does it make to divide the available data into IS and OOS?
Am I missing something? Maybe it’s possible (there are sooo many options available) to select somewhere the portion used for fitness rank?
we have identified an issue when strategies are being replaced in a full databank. As you mentioned fitness was not evaluated correctly using IS-only data but using full sample. This will be fixed into the next SQX update 137
Viewing 1 replies (of 1 total)