Description
Just curious if anyone else has these issues. The eval swings from white to black in almost equal measure and then only corrects itself once the next move is played. Makes it impossible to keep track of the actual evaluation of a move.
I also haven't been able to get book moves to show up. For example, the engine is evaluating c6 in the Caro as in inaccuracy.
Using on Mac OS Sequoia 15.4.1
Edit: to update, book moves was completely on me. Needed to enable one of the APIs and now it works perfectly. However the evaluation is still weird. E.g before playing a move it will evaluation 1. e4 as 0.26 but as soon as it is played the evaluation drops to -0.17 if e5 is played, which is obviously an error. Once 1... e5 is played, says that 2. Nf3 will have an eval of 0.23 and once it's played, the eval swings again to -0.22 for 2...Nc6 which swings back again to 0.24 if 3. Bb5 is played etc.
Edit2: And the eval issue has now been solved. The issue was with Nibbler itself. Under analysis, the option to select the POV for win rate and centipawn was set to current. I took this to mean, the current POV of the board, in this case white. What it seems to mean is the POV of which piece's turn it is. Setting both manually to the perspective you are analysing from solved the issue.