43998

[drahacikfm]

Richard:

First problem I have ever seen where the Toga evaluation of the solution (1....Rxb4 +1.54) is less than +1.75.

It's clearly the best move and it's clearly a winning K+P ending to a human.  But is this a generator bug, or is it expected behavior?  If it's expected behavior, why aren't there a lot more problems like this where the winning move is less than +1.75?

[richard]

I think in this case what happens is that I do the first analysis which crosses over the "good enough to be a tactic" threshold (this is different from the 1.75 second best move threshold).  When I do the ambiguity check with multi variation mode turned on (which is where I take the evaluation presented to the user from) the evaluation changes slightly (I think the threshold here is 1.80 so it moved at least .26 between the initial check and the ambiguity check).  This might be a difference in the time taken or a change brought about by multi-pv mode being used.

Richard.