"Practice with computer" needs improvement - the computer is drawish.

Hello.

The computer will draw ASAP if we practice theoretical draw endgames with it, even if it plays the stronger side. For example, human to play:
https://lichess.org/analysis/8/2K5/8/8/3kN3/8/7p/8_w_--0_1?color=white
https://lichess.org/analysis/4k3/R7/5K2/4P3/8/8/8/1r6_b--_0_1?color=black
(From the book 100 Endgames You Must Know.)

I think the computer should try to win, or else draw by the 50 moves rule.
Thanks.

Are you suggesting that an engine should not go for a draw in a drawn position?

@pepellou said in #2:

Are you suggesting that an engine should not go for a draw in a drawn position?
I think an engine should play for a win. An opponent might not know how to force a draw and certain engine moves might increase the chances for the opponent to make a mistake.

@Youcandothis said in #3:

an engine should play for a win

In a 0.0 position, any move that keeps 0.0 is the best move.

You're asking for the engine to implement "psychological" play.
Well, Stockfish will not do that but of course, feel free to make your own bot (or find one) with those capabilities.

@pepellou said in #2:

Are you suggesting that an engine should not go for a draw in a drawn position?

If you "PLAY WITH THE COMPUTER" (strength 8) any of these positions, it will not draw ASAP. In fact, the Philidor defense can get tricky. I am suggesting the same strength to "Practice with computer".

@pepellou said in #4:

In a 0.0 position, any move that keeps 0.0 is the best move.

You're asking for the engine to implement "psychological" play.
Well, Stockfish will not do that but of course, feel free to make your own bot (or find one) with those capabilities.

Have you played any of these endings with computer strengh 8?
If you ever take some time to analyse the first position with Stockfish 14+ NNUE, you will see it does not suggest: 1...h1=Q

That's always been a problem with comps. Since in a lost position it figures that everything is equally hopeless, it will often not select the strongest defense.

The point seemingly missed here is that in order to practice and improve and learn to hold a position, we don't need or want best play, such as forcing a draw immediately. We want human like attempts. In fact the exercises themselves say "hold the draw for N moves", which clearly shows the intention... Simple.

@thascius said in #6:

Have you played any of these endings with computer strengh 8?
Of course, I have. What's your point?

@thascius said in #6:

If you ever take some time to analyse the first position with Stockfish 14+ NNUE, you will see it does not suggest: 1...h1=Q
So in your opinion is h1=Q the best attempt really?

@thisisnotme123 said in #9:

We want human like attempts.
Then go play a human.

@pepellou said in #10:

Of course, I have. What's your point?
My point was that: if you do play — not practice — the first position with the computer, and you play 1.Kf2, the computer will not reply 1...h1=Q. However, if you practice with the computer, it will reply 1... h1=Q.

1... h1=Q is a "human" move, it is "giving up" the fight. No one can learn/practice to defend a difficult position if the computer "gives up" in the first move.

I am not suggesting "phychological" moves or human moves. I am suggesting "practice with computer" be as strong as "play with the computer". E.g:
https://www.chess.com/practice/custom?color=white&fen=8/2K5/8/8/3kN3/8/7p/8%20w%20-%20-%200%201&is960=false&moveList=