>>About the position you gave, that didn't work because White plays Rh2:g2.<<
you're right, I was blind
>>getting away with draws on timeout when their opponent could still checkmate them. That is the current system.<<
this system definitely needs a fix, we agree
>>So I suppose it's not possible.
But hey, we can add it to the list!
on (TimeOut)
{
if (enemy has nothing but a king) draw
....<<
I don't like the idea of suchs lists: it takes just some time and then some freak found another situation not yet in the list.
still my suggestion: ignore the immediate draw rule, let the game go on until the draw can be detected without accounting future moves.
and on timeout: request the engine to check all future positions whether helpmate would be possible.
don't worry, there's a reason why this check can't consume infinite time: if checkmate is possible it is always possible within a maximum number of moves, so the BFS algo can stop after this maximum depth.
I think the list idea is cool because so far it is very small. Nothing but a King left, vs. nothing but King and bishop left and only if you have at least a rook or Queen.
To complete that list would be an interesting experiment, though I think the most likely circumstance is that it is already complete (those are the only two ways to draw by timeout).
We can ignore the immediate draw rule, but there seems to be a small recursion in what you're saying.
You're saying, until the draw can be detected.
Well, the "immediate draw rule" is the reason King vs. King is a draw (no pieces). If we ignore that rule then both sides just have to move their kings a shit load of times until somebody gets a heart attack and dies (or hopefully just falls asleep). I know you said "until the draw can be detected", but, how would you define that? Any human can always detect a permanent draw pretty much but, the draw gets forced eventually or you allow both sides to wander.
What's a BFS algo? Sounds new to me.
It can be done for timeout as well; it's just difficult because for every game that times out, the server does a game analysis, which uses up bandwidth. I would think it more efficient to just check a small list of cases (only 2 so far) where it is definite the other side could not have won if time were not an issue.
>> I know you said "until the draw can be detected", but, how would you define that?<<
I said: "until the draw can be detected without accounting future moves."
and I simply wanted to say: "until there's just king vs king, king vs king + bishop, etc..."
and I should better have said: "ignore the exact FIDE definiton of the immediate draw rule" instead of "ignore the immediate draw rule"
>>What's a BFS algo?<<
just means that you call a function looping over every next move; within the loop: call the function again, recursivly iterating the moves after the next moves... and so on....
it's a concept to implement an AI in game theory.
>>it's just difficult because for every game that times out, the server does a game analysis, which uses up bandwidth. I would think it more efficient to just check a small list of cases (only 2 so far) where it is definite the other side could not have won if time were not an issue.<<
I see the problem with the server load... why don't we "outsource" this game analysis to the brain of the player who would be in favour of the win (in contrast to draw)?
"outsourcing" could work like the following:
if your opponent times out, you are allowed to do your opponent's moves until you get them checkmate.
if you can do that, you win. if you can't, it's a draw.
sound's funny, but definitely less server load than requesting a chess engine for helpmate.
>> I would think it more efficient to just check a small list of cases (only 2 so far) where it is definite the other side could not have won if time were not an issue.<<
I see your point, and yes, the "outsourcing" suggestion would annoy people in cases where the outcome is obvious... but how about this:
if draw/loss can be decided upon a simple list, like you propose, the game's provisional result is like you say,
but, if the player who didn't time out says: "why is it a draw, I would have won this game!!" then okay, allow them to make the moves of the timed out player - if helpmate is possible, then the timed out player loses.
why do I discuss this at length? simply because I don't want the case that the "else"-clause of your "on (TimeOut)"-rule gets triggered although there isn't a path of valid chess moves leading to defeat, neither do I want one of the "if/else if"-clauses to get triggered although such path does exist.
=> that would hurt the rules, and people justifiably complain: "hey, why did I lose? the opponent can't checkmate me" or "what? just draw? the opponent timed out and I know how I'd get 'em!!!"
it's me again, I signed up now, really needed an identity.
>>To complete that list would be an interesting experiment, though I think the most likely circumstance is that it is already complete (those are the only two ways to draw by timeout).<<
have a look at this:
lichess.org/analyse/4xzz1vttwhite timed out, and current lichess rules say white lost.
BUT: there's no combination of valid chess moves that make black win this game (white has to take the black rook), so the game has to be a draw.
so your list of exceptions needs to contain at least three items...
you might say: such situations are just artificial!
but I'm really worried that there are real life situation as well, which both of us just can't think of.
at the beginning you said only-a-king is the only exception, then there was the bishop thing, and now this example - the list gets longer and longer, I think I have a good reason to dislike the list idea.
Maybe, but I look at it this way.
In a position like that, White could have resigned thinking that was checkmate. :)
So actually, Black could still have won.
It's my cruel way of saying quite honestly I don't really care about the evaluation on time-out. In fact up until I started playing here on Lichess I have never heard of such a rule before, that the game gets scored as a draw on time-out in any circumstances at all. It's so rare that I would rather just drop it and give the full point over to the one player who could keep up with the clock. :)
Sorry, but it's just a totally new, arbitrary-seeming rule to me that I've never heard of. Not saying it doesn't exist, but if I had to choose between unable to completely support the Insufficient Material automatic draw rule, vs. unable to completely detect if the opponent was unable to checkmate to know whether to count timeout as a draw vs. a loss, I'd rather just stick to Insufficient Material and let the less, smaller rules take care of themselves. :)
I don't mean to walk away I have just been busy with other things and had less time to chat here. Anyone can discuss it, but I have a feeling thibault thinks it's tl;dr and won't do anything.
Yes and no;-)
I agree that it's difficult to implement these two rules:
Immediate draw as soon as no checkmate possible for either,
and correct draw/lose if one site runs out of time.
But I have to counter you when saying:
>>but if I had to choose between unable to completely support the Insufficient Material automatic draw rule, vs. unable to completely detect if the opponent was unable to checkmate to know whether to count timeout as a draw vs. a loss, I'd rather just stick to Insufficient Material<<
It's not like one rule gets more difficult to implement if we decide for the other - both rules are difficult, anyway.
Consider these two examples:
Example #35:
lichess.org/analyse/4xzz1vttExample #37:
lichess.org/analyse/j0y2ry8nIn example #35, there's absolutely no situation where one site couldn't get a checkmate against the other, meaning that all difficulties of implementing this are unrelated to the immediate draw rule. It's just a timeout thing.
Example #37 is played completely without timeout, it's an unlimited game, and so it's not related to the draw-or-loss rule on timeout. It's just about the immediate draw rule.
Both examples are artificial, but there could be real life situations similiar to them.
What I want to say: Thibault will have to face major difficulties whenever he wants to implement one of the two rules, no matter how he deals with the other one.
My attitude towards real life situations similiar to example #37 is that we can rely on the reasonability of the two player to agree it's a draw. This is why I'd be contented if the implementation just checks for standard draw situations like king vs. king, etc.
With real life situations similiar to example #35 I see a problem: It's not fair to say a player lost the game even though there's physically no way for their king to get checkmated - This is the reason why the timeout rule was introduced at all. You said, the player could resign, but why should they resign if there's just one move and the opponent has nothing but a king?
About the existence of the timeout rule: I thought Dougthehead already pointed that out in the original post? I can show you that, too, if you don't belive that.
And I can't imagine that the people who indruduced the timeout rule did so just for fun; there are serious considerations that a game shouldn't be lost if checkmate is impossible, except in case of resignation.
"In example #35, there's absolutely no situation where one site couldn't get a checkmate against the other, meaning that all difficulties of implementing this are unrelated to the immediate draw rule. It's just a timeout thing."
Not sure what you mean. The draw at the end of #35 is in every way because of the immediate draw rule. The problem is of course implementing its detection.
#37 is no different. Whether or not there is a clock has no effect on whether draw by insufficient material exists or not.
Unless you're saying that the draw rule of insufficient material should be completely ignored and just let draws happen on timeout, which is equally as unfair as me favoring the opposite because it's choosing one rule being implemented and the other being ignored.
"About the existence of the timeout rule: I thought Dougthehead already pointed that out in the original post? I can show you that, too, if you don't belive that.
And I can't imagine that the people who indruduced the timeout rule did so just for fun; there are serious considerations that a game shouldn't be lost if checkmate is impossible, except in case of resignation."
Well since this is "the Internet" and internet chess, I don't think that such "real-life" or "serious" situations exist. It is still very rare that your clock runs out in a game where your opponent is completely powerless, over 95% of the time having nothing but a King. If you can't keep up with your opponent and the clock when your opponent has nothing left but the King left you probably don't deserve even half a point anyway.
So yes, in the way you're presenting it, you're suggesting one rule is more important to implement than the other (which is yet another thing we agree on :P) You were just saying earlier that the automatic draw by insufficient material should be ignored and leave both players to offer a draw generously and accept; I'm saying that that's no less biased than following the rules and drawing by timeout, giving 0.5 points to both players for fucking up in a way where one player is too slow to conquer somebody who has nothing.
>>Not sure what you mean. The draw at the end of #35 is in every way because of the immediate draw rule. The problem is of course implementing its detection.<<
Luckily, all moves are numbered :-)
From the beginning of example #35, up to the situation of the black move #19: rook e7 -> e2, both sides can technically checkmate the other, assuming their worst play. Up to that move, black has still has a king and a rook, so helpmate against white would be possible. There's no insufficient material rule applying to this situation, right?
Now, how would the game go on, if white didn't time out? Easy: White captures the black rook (e.g. move #20: queen d1 -> e2), and then can easily kill that lonely king. So, from this move onwards, black can't helpmate white, but white can checkmate black. So the insufficient material rule doesn't apply here either.
Hopefully this convinces you that the immediate draw rule (due to insufficient material) can't be applied to example #35. If not, please post it along with what move number you say the insufficient material rule should be triggered, and why.
So example #35 is really only about what should happen after a timeout, and doesn't involve the insufficient material rule.
Example #37, you admit, can't be about what happens after a timeout, as there are not timeouts in this game at all - it's unlimited!
>>which is equally as unfair as me favoring the opposite because it's choosing one rule being implemented and the other being ignored.<<
These examples should show you that the what-should-happen-after-timeout rule and the immediate-draw-if-no-helpmate-possible-for-either rule (a.k.a. "Insufficient Material" rule) are two completely different animals.
About the "Insufficient Material" animal:
I think the current Lichess engine already does a good job in detecting those, and that there's no urgent need for improvement. This is what I wanted to say with: "Ignore the exact FIDE definition of the immediate draw rule", because that's exactly what the engine currently does. (If Lichess really exactly obeyed the immediate draw rule, then example #37 would be an automatic draw, rather then waiting for the players to agree on this.)
=> So, no, I did not suggest to ignore, weaken, whatever, the current implementation of the insufficient material rule.
I say: The current implementation of the insufficient material rule is good as it as, it already covers the standard situations like king vs. king, etc.
It's just that it can't handle situations like example #37; if you know how to detect situations like example #37 to properly trigger an automatic draw, then post it - otherwise let's agree that the current implementation is good and doesn't need to be changed.
About the timeout animal:
I DO request a change here, because the current implementation doesn't obey the rules in two ways:
1.: In the original poster's case (example #1), white should have won the game, because black timed out and helpmate is possible. However, the engine decided it's a draw.
2.: In example #35, the engine decided white lost the game. But that's not possible, because there's absolutely no way for black to helpmate white after the shown position.
Both of these cases aren't rule conforming. The worst of this is that the rule gets hurt in both directions: Sometimes games are treated lost, although there should a draw, and vice-versa.
If no rule conforming implementation is possible, then there should be at least a simple "always draw" or "always lose". The latter one is probably the better one;-)
I meant #37.
At first when I clicked #35 and #37 I thought they were the exact same game, except one of them was timed and the other was untimed.