lidraughts.org

Scan - wrong evaluation

Hello Notice this kind of bugs few times already. for one move Scan show 0 evaluation draw and just move later realized its lost.

I noticed similar problems in different positions before.

lidraughts.org/study/yNFunceV

@bogomol Scan solves tactical and short-term tasks efficiently, but Scan can't evaluate long-term plans, and without tablebases Scan can even make inaccurate predictions on the nature of eventual endgame. If a player follows to indications of Scan without critical thinking, he or she might get materially advantageous endgame position, which unfortunately is not sufficiently for winning. It seems to me that Scan is indicating a plan where an opponent can run into positional inaccuracy or tactical blunder easier.

Hi kalnap,
Not sure if my link worked to show right moment. Can you check game 59 and move 82 for white. in below study
lidraughts.org/study/yNFunceV
Scan missing simple shot, maybe it some problem in the engames?

@bogomol Scan doesn't do well multiple king endgames: sometimes it is suggesting inaccurate moves or doesn't recognize correct number of moves for winning. You might see some examples at my study of "Multiple king endgames" : lidraughts.org/study/YiAZbWM6 .

Thanks for the study, will have a look.

But in this case Scan missing simple one move shot - one for two. And evaluating it as a draw. Its no complicated must be some serious bug in the program engine.

Scan gives 0.0 on 81...25-9 because this position is a repetition of 79...4-9, and shows the same move that was played there in its pv.

If you analyse move 79 of black in the same study, or setup the position on the analysis board, it gives a normal evaluation:
lidraughts.org/analysis/standard/W%3AWK5%2C15%3ABK9%2CK24%2C27%2CK35#0

So it's not a Scan evaluation bug, but a somewhat confusing way of showing the repetition.

Ah I see, thanks for explanation! Now all make sense.

But maybe in the analyses the rule of repetition (or any other 15 moves without moving pieces etc) should not be taken on board. As it can make confusion in position valuation? Just a suggestion

Move by repetition is actually key in evaluation some positions though. If you are winning except your opponent found one move to get repetition then your position is not winning. Think king queen endgames in chess.

It would be nice if there was a #0 meaning forced draw by repetition which would be useful for many players just like the #forced win

I think a reference to threefold repetition draw or inability to convert material or positional advantage within prescribed number of moves should be considered as a part of FEN string.

This topic has been archived and can no longer be replied to.

Reconnecting