Open microprediction opened 2 years ago
The set of stacking methods here is rather limited.
There's a nice survey in this paper and it would be pretty quick to add a few more.
Judging by the stacking success here it is worthwhile.
There's no need to do much testing independently as the Elo ratings take care of it.
Any takers?
This issue is probably superceded by the https://github.com/microprediction/precise package
However leaving here as we need to incorporate it into the skatertools/ensembling
The set of stacking methods here is rather limited.
There's a nice survey in this paper and it would be pretty quick to add a few more.
Judging by the stacking success here it is worthwhile.
There's no need to do much testing independently as the Elo ratings take care of it.
Any takers?