Play with computer is too easy

Sometimes when I finish a puzzle I want to see whether I understand how to take advantage of the winning position and I click on the bullseye symbol at the bottom of the move navigator. It has the label "play with computer".

However, I find this feature is not very useful because the level of computer play is very low. Esp. in endgames, the computer makes terrible moves of the king that just look random.

Please introduce the option to change the level of computer play so this feature becomes more useful.

p.s. I searched for this issue on github, but after searching thru several pages of issues I found nothing so I am reporting it here. Unfortunately, the last time I reported it, the thread was closed with no useful input.

Just as an example.
Solve this puzzle:

lichess.org/training/LBVgS

Then try to continue by hitting "play against computer".
The black king will move away from whites pawn, a completely nonsensical move.

Here is the same situation as a study:



Play the last move for white. Now you will see that ofcourse stockfish captures white's pawn after this. Practicing against such a weak computer is not very useful.

I think the actual problem is that it is too strong for it's moves to make sense to you. It can see the mate in 11, and that taking the pawn won't delay the mate, so it doesn't bother to take it.

Ok, maybe i selected a bad example. But my impression is that its moves are not strong. I regularly defeat 'play against computer' and I feel like SF14+ would make the situation much harder. Do you have a different experience?

It uses SF14+, idk what you're talking about.

Usually in "Practice with computer" the computer plays very strong. Even with an evauation of +4 I seldom win. --- But as @ProgrammerAngrim said, when the situation is lost, it is lost - and the computer's moves seem indeed to be "random". So you just can't try to pratice a pawn endgame because he never seems to defend the best possible.

There have been a couple of threads about this, I never got a real explanation yet. My guess is: When in tablebase mode (less then 8 pieces left) the engine just chooses number 1 from the list of all equally lost variations.

In dozens of different forum posts I keep explaining that Stockfish with tablebases produces more interesting endgame moves than Stockfish without tablebases, and while I can't guarantee that in every position and while people contrive counterexamples, over my years of using it (on my PC, not in a browser) that's proven to be the case and in other cases Stockfish endgame moves tend to be fatalistic like this.

But why, @Toadofsky ?

When I just go in analysis mode and let me show the best move, then engine shows good moves. For example, recapture a rook or something (even if I can win even then).
But in "practice" mode, the engine often makes moves that are not even among the best three Stockfish moves. And this only happens in the endgame.

It looks like the "practice" mode uses another algorithm to pick its moves in the endgame. Hence my suspicion it might be a tablebase move. And I really see no reason for that.

Thankyou @Sybotes . I was beginning to doubt myself. I'm not a strong player and its certainly possible that in some cases I misjudge the best moves in a position. This discussion has added insight for me.

Encouraged by your statement re. tablebase , I went back to the example puzzle i posted and let SF14+ run for a while.

After I make the best move (completing the puzzle), these are the best three continuations:

Kxf4 (eval +99). This is the move I expected the computer to play. It does not.
Ke4 (#17).
Kd4 (#16).

"Play against computer" plays none of these three moves.
It plays Kd2 (#14).

Therefore the move it plays actually accelerates the mate.

This seems to disagree with @ProgrammerAngrim 's statement that there is a mate in #11 and it doesn't matter whether the computer captures the pawn. But maybe that was just a hypothetical example?

Anyway, I stand by my original statement that "play against computer" to practice is not much fun in endgames since the computer often just runs into a sword.

@EmaciatedSpaniard said in #9:
> Anyway, I stand by my original statement that "play against computer" to practice is not much fun in endgames since the computer often just runs into a sword.
Yes, stay with that statement. We had the discussion a couple of months ago, e.g. lichess.org/forum/lichess-feedback/why-does-lichess-stop-playing-the-best-move-in-a-lost-position

@Toadofsky keeps telling us that there is no problem then and today.