@mCoombes314 #10
Can you please read the entire thread before posting next time? I said the exact same thing in post #1.
Can you please read the entire thread before posting next time? I said the exact same thing in post #1.
As MessyAnswer said, I also agree, that local evaluation works really bad. And here comes the proof for you:
pl.lichess.org/study/oinf0KUg
Game is loaded from FEN, becouse it's not my game, so i only want to show you the key position.
Here, Local engine claims -6.8 - corectly finding, that black is on high adventage here. She doesnt change a mind even when evaluating this really long.
But well, she misses forced mate in 5. And the move she propose (bishop drop) is not the correct one.
As i could easly find this watching my friend's game,
the engine shuld really find this in microseconds.
Something is really wrong with local evaluation if she misses the mate in 5 lol.
Server side engine, fortunetely gives a correct answer here.
pl.lichess.org/study/oinf0KUg
Game is loaded from FEN, becouse it's not my game, so i only want to show you the key position.
Here, Local engine claims -6.8 - corectly finding, that black is on high adventage here. She doesnt change a mind even when evaluating this really long.
But well, she misses forced mate in 5. And the move she propose (bishop drop) is not the correct one.
As i could easly find this watching my friend's game,
the engine shuld really find this in microseconds.
Something is really wrong with local evaluation if she misses the mate in 5 lol.
Server side engine, fortunetely gives a correct answer here.
@CrazyModeOn: I want to exclude that this is a problem with position setup. Do you have the full game by any chance?
Yes I do, but i feel like i shuld ask first the guy who play that one if i can share this ;) Well he was not too proud of that game. Especially when i showed him that here he shuld be mated by much lower rated opponent.
Maybe I can send you the game in private way?
I think local analysis is time limited. Perhaps less than 20s. I am not sure on this.
The engine is not God. There are positions that it does not evaluate accurately. Added to that its time is limited. But the fact that it sees the advantage for black proves that it is doing well. I don't think it is bad.
The engine is not God. There are positions that it does not evaluate accurately. Added to that its time is limited. But the fact that it sees the advantage for black proves that it is doing well. I don't think it is bad.
@Durupo.
At some point she may all the sudden miss the mate for the teoreticly loosing side. Especially when she is missing so short mating sequence.
At some point she may all the sudden miss the mate for the teoreticly loosing side. Especially when she is missing so short mating sequence.
@CrazyModeOn: Sure. Send me a private message on lichess or email me at niklas@backscattering.de. I'll make sure only other trusted lichess developers see it. That would be really helpful. If you're paranoid my PGP key is 0x2eca66c65b255138 ;)
the sign of a true professional! an invitation to encrypted communications. well done.
well done, sunsetter ai level 8 is already very strong, i checked ten blitz games for a quick test and the result is: the rating of this engine @ level 8 looks 2800-2900 at first glance and not the given 2500.
This topic has been archived and can no longer be replied to.