Excellent.
Stockfish Resigns In "Winning" Position!
Sort:
Thanks for sharing; interesting position. It seems that these positions must not come up too often in Stockfish's own games anymore given all the *%! it kicks, including still dominating Leela which would probably come up with a different result here.
Yeah it's funny how the computer can't see that the position is drawn even after playing for 100+ moves when a human can recognize it instantly. I don't think Leela or Alpha Zero would play 41. a6 because the way they analyze positions is much closer to how humans analyze positions. They also learn from past experience.
Jan 16, 2019
0
#6
yes. garry kasparov i think said that what was so amazing about alpha zero's moves is that they seemed so much more "human" than those of stockfish or houdini for example
Try to follow a6 and lets see what happened then.
Stockfish is the top publicly available engine, dominating the computer chess test rankings. There is not necessarily *any* "better" engine. Even Alpha Zero may not be truly better as it crushed Stockfish playing on custom & monstrously powerful hardware. But Leela and Alpha Zero, to the extent they play more "human" may have avoided a6.
@Geodexic
I did try the computer's line with 41. a6 and the computer couldn't make any progress whatsoever.
@ArtNJ
Yes that's true thanks for your comments.
@Phoenyx75
Sorry I must have initially misread your comment. For some reason I thought that you said your engine suggested something else besides 41. a6. But it looks like we're on the same page.
@IMRonilm1204
Really? Where can I find it?
Stockfish on other servers is 150 moves, beyond that limit , it is a well known fact that SF will resign to save server resources! As you know a strength of computer program depends on available hardware resources and time. Even if the server supports 100 cpu, if 1000 people are playing/using server resources, the strength of Server Stockfish is 0.1 cpu strength. I guess those SF are not 3500 Stockfish, but about 2600 I guess.
@drmrboss
Yeah Stockfish resigned at about the 150 move mark so that holds up. I once saw a game where John Bartholomew played against the "Impossible" computer in chess.com live chess and it also resigned in a drawn position. He was also confused, haha. Now I know why it does that, thanks.
Ok thanks. And yeah I know about the quotes being disabled.
Jan 16, 2019
0
#18
Good point here! Note also that Stockfish and other engines will consider some moves blunders (several Pawn equivalents worse than the best move) even though the moves are ones that clearly lead to a victory, like keeping the opposition in a King and Pawns endgame - that happening with one of my wins.
The computer doesn't understand that low-complication sure moves are often best for human minds!
@MickinMD
Yeah I was very surprised to find out how many of my moves were counted as mistakes or inaccuracies according to the engine. I thought that I had played very well. It's hard to know when to trust the computer in a position like this. I take the engine's evaluation with a grain of salt.
You are right that sometimes the best move for a human is not always the top move recommended by the engine. The best move for a human player is the one that makes the most sense to that player. If you are winning then you want to make things as simple and as straight-forward as possible. If one move is +4 but requires a lot of accurate moves and one move is +2 but is an easy win with no risk then you should almost always go for the sure thing. If you complicate things then you give your opponent chances to get back in the game.
For trusty analysis of SF, just analyse a position for 1 billlion nodes( 3 mins for 4 cores pc). SF strength vs search nodes is a bell shape curve, 1 billion node is a sweet point for SF. .SF strength massively improves up to 200m nodes. but from 1 billion nodes to 100 billion nodes , the search doesnt get like x100 benefits, cos SF already see most important nodes within 1 billion nodes search.
For Leela analysis, even a single node is very powerful and , is around 2100 strength., but doesnt get much better after 1 million nodes.
Yesterday I watched some videos on chess.com in which one of my favorite players, GM Simon Williams, shows some of the instances where computers get things wrong. One of the things computers get wrong is that they don't understand blockades.
I recently finished a game that shows a perfect example of this. On move 41 the computer completely misunderstands the position because its top move is a6. It even gives White a +2.8 advantage! During the game I immediately dismissed a6 because I knew that if I lock up the Queenside I would drastically reduce my chances of winning.
I played out the engine's moves and eventually got to this position:
The computer was just shuffling its pieces around aimlessly all the while saying White had a winning advantage. So I decided to play against the engine as Black to see what it came up with.
For some reason Stockfish actually resigned even though the evaluation said White was winning!
To quote Simon Williams:
"Computers are not infallible and we’ve got to remember that. We’ve got to rely on our own mind. We’ve got to be confident in our own abilities. We’ve got to have our own views on a position. A famous story which I had is I went to a tournament in Vienna, a beautiful tournament at the Town Hall with another Grandmaster friend of mine and a weaker player friend who was only about 1700. We got in a routine after the game of having a bottle of wine, sitting outside and looking at the games on a board. Me and my Grandmaster friend wouldn’t use a computer, we would just suggest ideas. Our friend had his eyes fixed to the computer on his phone. If we suggested a move he would be saying “well my computer says -0.21, that’s not a good move”. And we would be saying “turn your computer off!” And eventually we nearly ended up grabbing his phone and throwing it across the square into the nearest water feature. The moral is this guy would not improve because he was so transfixed on what the computer said on his phone. He didn’t actually learn anything himself. He didn’t use his own mind. He didn’t try to work things out for himself. So all I’m saying is use computers to point you in the right direction. Use computers sometimes in very tactical positions. Use them in openings but don’t rely on them 100%."