Stockfish level 6 can't checkmate with king+rook • page 1/2 • General Chess Discussion • lichess.org

I played a terrible game off chess against Stockfish level 6 on lichess and it ended up just needing to checkmate me with a king+rook vs a king

https://lichess.org/3jpZh9Py/black#107

I was in a time scramble (it had ample time) and first played on out of petty instinct then because I realized it didn't seem to be able to checkmate me. It didn't, it played on to the point where we reached the 50 move rule. I don't have more to say just thought it was too silly not to share as one would think level 6 out of 8 Stockfish should be able to do a basic checkmate (yes it's not peak Stockfish but come on).

I played a terrible game off chess against Stockfish level 6 on lichess and it ended up just needing to checkmate me with a king+rook vs a king https://lichess.org/3jpZh9Py/black#107 I was in a time scramble (it had ample time) and first played on out of petty instinct then because I realized it didn't seem to be able to checkmate me. It didn't, it played on to the point where we reached the 50 move rule. I don't have more to say just thought it was too silly not to share as one would think level 6 out of 8 Stockfish should be able to do a basic checkmate (yes it's not peak Stockfish but come on).

Marcin2

#2

I would expect that Stockfish knows elementary checkmates at every level.

TipsChess_Official

#3

My dude, I expected Stockfish Lvl. 6 to know the simplest checkmate with king and rook, push king to a border and check him on the lowest file... dang Stockfish needs a reboot...

P7formula

#4

True, Stockfish on Lichess is falling apart, I believe the ratings are good but how they dumb down is very arbitrary and even lazy. Blame the programmers. The best logical chess engine is far from being operational mostly from funding. Programmers and logicians have to eat.

Chiksan

#5

Don't chess engines just make a mistake when the random number generator says to make mistakes? And the lower levels make mistakes more often? If that's the case, that could explain why it can't mate with the king and rook. It has to make the "right" move 20 times in a row and picks the 2nd or 3rd best move often enough that it keeps screwing up the checkmate. Kinda funny actually.

A_0123456

edited

#6

Can confirm this, it can't even checkmate you with K+R in Correspondence

https://lichess.org/QkAAPwdw/white#0

Edit: SF7 can, but with difficulty. It took 20 moves against quite suboptimal play and there were multiple moves where it randomly moved back its rook when it didn't need to

Can confirm this, it can't even checkmate you with K+R in Correspondence https://lichess.org/QkAAPwdw/white#0 Edit: SF7 can, but with difficulty. It took 20 moves against quite suboptimal play and there were multiple moves where it randomly moved back its rook when it didn't need to

VARSHANChEsS123

#7

Stockfish can be pretty wild at times!

Brian-E

#8

I don't know how reduced playing strength levels are implemented with stockfish, but my guess is a combination of a truncated search depth and reduced quality of the evaluation function. That can easily mean that the program won't know how to make progress with K+R vs K.

In the early days of chess playing software the advice was: "never resign against a computer because it often won't know how to finish you off". By playing a computer on a lower level I guess we are artificially recreating that situation.

I don't know how reduced playing strength levels are implemented with stockfish, but my guess is a combination of a truncated search depth and reduced quality of the evaluation function. That can easily mean that the program won't know how to make progress with K+R vs K. In the early days of chess playing software the advice was: "never resign against a computer because it often won't know how to finish you off". By playing a computer on a lower level I guess we are artificially recreating that situation.

corvusmellori

#9

@Brian-E said in #8:

I don't know how reduced playing strength levels are implemented with stockfish, but my guess is a combination of a truncated search depth and reduced quality of the evaluation function.

It is in fact not like that at all. Instead, the engine analyzes the top several moves, and randomly chooses a non-optimal one, with some set of criteria.

@Brian-E said in #8: > I don't know how reduced playing strength levels are implemented with stockfish, but my guess is a combination of a truncated search depth and reduced quality of the evaluation function. It is in fact not like that at all. Instead, the engine analyzes the top several moves, and randomly chooses a non-optimal one, with some set of criteria.

Toadofsky

#10

I don't think AI level 6 is using syzygybases (endgame tablebases). It may be interesting to try downloading Stockfish and see whether it needs tablebases to play well with similar randomness and depth constraints.