b6k/1b6/2b5/3b4/P3b3/P4b2/P5b1/RRR1K2b w - - 0 1
Is apparently somehow good for black.......
Humans are smarter than machines......
Is apparently somehow good for black.......
Humans are smarter than machines......
To make it easier for people viewing the thread:
en.lichess.org/analysis/standard/b6k/1b6/2b5/3b4/P3b3/P4b2/P5b1/RRR1K2b_w_-_-_0_1
It is definitely true that there are positions that humans evaluate more accurately than the best engines, and this is an example of one.
Of course, I always find it a bit amusing that we humans these days still get really excited when we find a position that we are able to evaluate better than the engines.
After all, these days there are far, far more positions that engines can figure out that we can't than there positions like this, so it's really not much of a victory :)
Of course, we don't like to think about it, so we just ignore those positions :)
Still, finding these sorts of positions is interesting, since finding them is the first step towards creating engines that can correctly evaluate them, which is just a matter of time :)
en.lichess.org/analysis/standard/b6k/1b6/2b5/3b4/P3b3/P4b2/P5b1/RRR1K2b_w_-_-_0_1
It is definitely true that there are positions that humans evaluate more accurately than the best engines, and this is an example of one.
Of course, I always find it a bit amusing that we humans these days still get really excited when we find a position that we are able to evaluate better than the engines.
After all, these days there are far, far more positions that engines can figure out that we can't than there positions like this, so it's really not much of a victory :)
Of course, we don't like to think about it, so we just ignore those positions :)
Still, finding these sorts of positions is interesting, since finding them is the first step towards creating engines that can correctly evaluate them, which is just a matter of time :)
Texel Chess Engine by Peter Österlund does correctly evaluate this position.
But I'm pretty sure it has some special draw detection hard-coded.
Shredder has something similar for some pawn fortresses but fails at this position.
But I'm pretty sure it has some special draw detection hard-coded.
Shredder has something similar for some pawn fortresses but fails at this position.
I just want to draw attention to this hoping that some Stockfish dev might see and correct this.
#4 why should they?
It's no big deal to implement the draw check and override the evaluation. But it slows the engine down. And what is the use? You have ONE more accurate evaluation but you lose overall ELO (-speed) for highly artificial positions, which - most likely - will never happen in a real game.
Remember that this totally wrong evaluation does not cost any ELO. The engine will not play weak moves just because the evaluation isn't 0.0.
Furthermore think of it that way: an engine analysis is never correct unless it displays mate for one side, or a draw score in drawn positions.
It's no big deal to implement the draw check and override the evaluation. But it slows the engine down. And what is the use? You have ONE more accurate evaluation but you lose overall ELO (-speed) for highly artificial positions, which - most likely - will never happen in a real game.
Remember that this totally wrong evaluation does not cost any ELO. The engine will not play weak moves just because the evaluation isn't 0.0.
Furthermore think of it that way: an engine analysis is never correct unless it displays mate for one side, or a draw score in drawn positions.
This is such an artificial position that nobody should care.
However, in Forum General Chess Discussion there is a thread with "Stockfish can't see win for white here. And how about you?" which looks like a more serious flaw.
However, in Forum General Chess Discussion there is a thread with "Stockfish can't see win for white here. And how about you?" which looks like a more serious flaw.
@brobot What's incorrect?
This topic has been archived and can no longer be replied to.