I just had the second occurrence of the server analysis giving an evaluation that is clearly wrong, with the local analysis finding the correct evaluation after a few seconds.
Here the server analysis estimates +0.5 despite the position being a theoretical draw.
The first one I found even had the evaluation going from 0 to #-19
In both cases it seems to be caused by a late game inaccuracy which seems to make the server focus on the played line, and be lazy about computing the optimal line to the end.
There are also lots of positions where the local and server recommendation differ, and I prefer the local line. But for those it is possible that this is caused by my own lack of skill.
Personally I think that the best solution would be to allow the local analysis to overwrite the server analysis once it gets deeper.
Here the server analysis estimates +0.5 despite the position being a theoretical draw.
The first one I found even had the evaluation going from 0 to #-19
In both cases it seems to be caused by a late game inaccuracy which seems to make the server focus on the played line, and be lazy about computing the optimal line to the end.
There are also lots of positions where the local and server recommendation differ, and I prefer the local line. But for those it is possible that this is caused by my own lack of skill.
Personally I think that the best solution would be to allow the local analysis to overwrite the server analysis once it gets deeper.