What stupid rules Lichess uses regarding draws on timeout

@Akbar2thegreat said in #50:
> What about the evaluation then?
Modern competitive engines (like stockfish) are designed for best play, not for "accurate" evaluation. The stockfish devs, in particular, have been very clear that they are not focused on having an evaluation that is helpful to humans, just in having an engines that wins chess games (and not even an engine that solves studies).

Even once it found the solution, stockfish evaluated the position as -2.2 or so, and evaluation that in most positions would be completely winning. The evaluation would likely drop from -2.2 to 0 sharply only at the very end of the 50 move rule (and I wouldn't be surprised if stockfish chose to sacrifice the queen to prolong the end of the game).

It's of course theoretically true that any engine that performs infinite search will *eventually* play the best line, but for some puzzles (much harder than this one), it may take a very, very long time, and many compute resources.

@JesusIsLord906 said in #47:
> What part of burning up their clock while still staying alive isn't a skill to you?
I don't really understand what you mean. By the end all your were doing is surviving, there was no way for you to win. The payoff of your skill in this game was you did not lose, you ended up drawing, which is better than losing. What more could you want? You didn't play well enough to win.

@Buttercup22 said in #42:
> I agree, it's like they have a rule but they don't enforce it. Running out of time in a timed game means you lose. Except it doesn't because we don't take our own rule seriously.

What a circular argument... "I don't like the rule, therefore it isn't being followed."

@corvusmellori said in #51:
> Modern competitive engines (like stockfish) are designed for best play, not for "accurate" evaluation. The stockfish devs, in particular, have been very clear that they are not focused on having an evaluation that is helpful to humans, just in having an engines that wins chess games (and not even an engine that solves studies).
>
> Even once it found the solution, stockfish evaluated the position as -2.2 or so, and evaluation that in most positions would be completely winning. The evaluation would likely drop from -2.2 to 0 sharply only at the very end of the 50 move rule (and I wouldn't be surprised if stockfish chose to sacrifice the queen to prolong the end of the game).
>
> It's of course theoretically true that any engine that performs infinite search will *eventually* play the best line, but for some puzzles (much harder than this one), it may take a very, very long time, and many compute resources.

To supplement this... I think our topic shifted to fortress detection, but nobody did a Google search for "stockfish fortress detection" and instead we decided to reinvent the wheel. Anyway, here's what I found after 10 seconds of effort, sparing us 10 years of argument:

> Dec 26, 2021 — Crystal is a Stockfish fork that specializes in detecting things such as endgame fortresses that other engines may miss.

www.reddit.com/r/chess/comments/rornau/crystal_40_has_just_been_released_crystal_is_a/

@Toadofsky said in #54:
Surprisingly, I found that Crystal performed worse on the two fortress puzzles than Stockfish. I've always had difficulties with Crystal - it so often seems to go into a search explosion forever and stop reporting back evaluations. Stockfish is known to do the same in some positions, but at least in my experience far fewer. Or maybe it's just me :)

@Toadofsky said in #53:
> What a circular argument... "I don't like the rule, therefore it isn't being followed."

No he was saying, Logically when we run out of time, we lose, but under some circumstances we don't follow our rule and magically just make up new rules.

@Quantum_Immortal said in #52:
> I don't really understand what you mean. By the end all your were doing is surviving, there was no way for you to win. The payoff of your skill in this game was you did not lose, you ended up drawing, which is better than losing. What more could you want? You didn't play well enough to win.
Didn't play well enough to win?

This is my whole point, If you can sacrifice your whole army and burn up their entire clock you certainly deserve to win.

Imagine staying alive against a freaking army until some sort of clock ran out in any other sport or war that would be considered epic.

When playing against really good players this is really really hard too!

Trust me, it is one heck of a skill to burn up the enemies time and still be surviving even when your army is gone, insane skill.

If you can't compute/understand that 1 guy vs loads of guy and still surviving, plus you've used up all the enemies time is a huge skill, then I don't know what to tell you, and I'm telling you, Clock dies, you should too, regardless of material or winning chances, the whole point of the clock is to declare the other person the winner if yours runs out.

I mean think about this logically, I'm in a war, with my enemy, I've planted a bomb with a timer, but all my men in my squad/army get killed but me, and finally the timer/clock on the bomb goes off, but the bomb just decides, Hey man, nah I'm cool, I'm not going off, because that guy only has 1 man/1-piece left, lets call this a draw.

Utter nonsense.

The clock/timer runs out, the bomb goes off and in this case it is big enough to destroy the enemies king/army.

Point is, Clock dies, You die.
Not this fairy tale rubbish where oh well the clock went off but uhm, yeah the guy is safe with a draw.
No, clock dies, you die, it's like oxygen, what are you going to do when you've reached absolute 0? You'd die.

Anyway this is where I'm resting my case, we should be more logical and consistent, clock dies, you die, period, no exceptions.

That's my opinion and I feel it is much more sensible and it is observed in nature so to speak.

I mean imagine telling someone, now don't run out of men/pieces or when his clock dies, he's saved! It becomes a draw!

Like what the..

@JesusIsLord906 said in #29:
> So somehow you think material is more important than the person running out of time.
i dont say it actually you will say it if this happens to you

if fide rule changes

after one month of rule changing:
@JesusIsLord906 said hey i got timed out and lost the game but it was mate in 1 this is unfair i should win if i run out of time because i was winning this rule of lichess is stupid

after 2 years
hey i lost 15 rating after losing to 1000
and i got only 2 rating after winning to 1000 this is unfair
reason: you are higher rated than 1000 and you should be lucky that u lost only 15 points

@JesusIsLord906 said in #58:
> The clock/timer runs out, the bomb goes off and in this case it is big enough to destroy the enemies king/army.

I've had friends IRL play a game using a chess clock but without any board or pieces; they attempt to mash the clock so the opponent loses on time. This rule was something the Lichess team discussed and after a year of meditating on it I aggressively lobbied for a FIDE-like rule because we're here to play Chess, not Clock.

The problem you're seeking to solve is this: players agree to a time control without an increment, one of the players loses enough material that it's easily proven that a checkmate is no longer possible, and the opponent's clock elapses. A FIDE-like rule discourages "dirty flagging" attempts and encourages players to either study endgames (thereby not losing on time) or to play with a +1 or longer increment.

The only factor in favor of a USCF-like rule (flag = loss) is that it would be simple to understand. And that's why over the course of a year I repeatedly considered which rule I wanted to advocate for, did my research of FIDE and USCF rules, ran some thought-experiments about possible scenarios under each rule set, considered positive and negative aspects of each rule set, and advocated for an informed albeit surprising-to-me position. At first, being a USCF member I thought rules I've seen on so many platforms (flag = loss) would be the best, being simple to implement and simple to explain & understand; but the added complexity brought by a nuanced rule actually promotes deepening an understanding of endgames (being able to convert won positions and put up resistance in bad positions), of time management (not trying to "dirty flag" without having any pieces), and an appreciation for just how absurd 5+0 and slower "+0" time controls are (having frequently won OTB blitz, rapid, and even classical games by "dirty flags" before USCF popularized delay clocks). Rules which promote players learning rather than randomly shuffling pieces for an unearned win are good rules.