Lila: AI threefold repetition draw claim

[ddugovic]

1. After 82... Rg3 I see a "Claim a draw" button. Clicking it does nothing: http://en.lichess.org/EGrA3Zf8EsoA
2. By FIDE laws of chess, a player has a right to claim threefold repetition prior to playing 82... Rg3 because they are able to cause threefold repetition (and if the arbiter asks, only then the player indicates Rg3).

[Unihedro]

1 - Is a bug, but I can't reproduce it.
2 - There is no arbiter and computing every move whether a threefold repetition can be _caused_ isn't very constructive. This rule is difficult to implement in online chess (the same with when time out with K + N can still be a win if in a forced mate sequence) and therefore is neglected.

[ornicar]

lichess is supposed to support this. Maybe the takebacks confused it.

[niklasf]

I can consistently reproduce 2. This is a real problem, because you have to pass the turn to your opponent, who might move, before you manage to claim.

[Unihedro]

On that note, maybe it's more practical to keep the button to claim draw there for an additional move (when I move, not when my opponent moves)?

[ddugovic]

Ah, perhaps simpler (and FIDE-rules appropriate) would be:

1. Player clicks "Offer Draw"
2. Player plays move causing threefold repetition, thereby claiming it

[Vinvin20]

I just saw an inconsistent eval with 3-fold rep today : https://fr.lichess.org/M0eOwUgm#69
The local engine gives +6 with 35.Qf6 but the Lichess analyse gives 0 .

[Vinvin20]

The local eval is right because it's only 2 times the same position.

[ddugovic]

I don't think it's quite correct to "keep the button to claim draw there for an additional move" since this could lead to angle shooting (in the form of a player seeing the opponent's winning move and then deciding to claim a draw in a position where threefold repetition no longer applies).