Claes Fransson

Now, I guess it remains to be seen how or if this affects this flattening issue. I noticed flattening also with `normalize_y=false` [before](https://github.com/kiudee/chess-tuning-tools/issues/118#issuecomment-775865410)...

> edit: To answer the question: We need to scale `alpha` since `y` has been scaled. Hmmm, I guess my logic of alpha only being added to non scaled K...

Trying to understand how the variance/uncertainty of the score is estimated/calculated by Chess Tuning Tools when there is only one game pair (1 round per iteration = 2 games). If...

> > Trying to understand how the variance/uncertainty of the score is estimated/calculated by Chess Tuning Tools when there is only one game pair (1 round per iteration = 2...

For the record, I now get flattening also with noise=0 for the scores, which I guess hints that the noise estimation of two game scores, is not responsible for the...

Thanks for your comments and analysis. > ...and the middle plot is definitely the most sensible. Still, the y-scales get shorter each continuous iteration. Here is the iteration before, iteration...

After implementing the possibility to save the intermediate results to disk after each game pair, I have now run a similar tune as before (no opening book, 4 opponents chosen...

[In this CLOP software](https://github.com/erbsenzaehler/clop) the game results are saved after each played game, and you can abort and resume the tuning after each game. I have no idea if that...

Thanks, I will continue to use 2 games per iteration then, hoping that the total required number of games won't increase too dramatically compared to more games per iteration. In...

For the record: Briefly mentioned a very general idea of how to do this feature [in this comment](https://github.com/kiudee/chess-tuning-tools/issues/118#issuecomment-888167317) (some loop inside main optimization loop), but haven't yet figured out how...