Cursed win at TCEC

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
syzygy
Posts: 4253
Joined: Tue Feb 28, 2012 10:56 pm

Re: Cursed win at TCEC

Post by syzygy » Wed Nov 16, 2016 8:14 am

syzygy wrote:
Jouni wrote:Note, that in TCEC Rapid there was also same ending round 20 Arasan - Protector. No discussion then. But engines were not reporting 0,00 - does it change situation?
Arasan reported 0.00. And Arasan had committed to this path to a draw 15 moves earlier or so. So Arasan made a bet which was fully supported by the rules of chess but still lost it.
Actually, you are right that Arasan too did not report 0.00.

However, if you look at Arasan's earlier evaluations you can see it correctly reported 0.00 until it entered a 6-piece position. From that point on Arasan reported its "swindle score". This is a bit dangerous as a very negative swindle score coupled to a very positive score reported by the other engine could have triggered the TCEC win rule, but the point is that Arasan was "aware" of the draw and would have been able to hold the draw.

If you replay the game with SF and Syzygy50MoveRule disabled, you can see it goes from a drawn position to a lost position about 15 moves before the end of the game. That is the point where Arasan committed to the 50-move rule draw escape. If it had ignored the 50-move rule, it would have played another move and probably have held the draw (not certain of course).

User avatar
hgm
Posts: 22306
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Cursed win at TCEC

Post by hgm » Wed Nov 16, 2016 9:24 am

mwyoung wrote:This is what happens when you have rules to adjudicate games, and not let the programs play out the moves.
There is of course always some unfairness in adjudication, in that you assume perfect play for engines that might not be able to deliver that. But you could argue that with participants like these sub-optimal play would be very rare anyway.

But this is much worse. It is whet happens when you have ERRONEOUS rules to adjudicate games.The rules assume the programs are stupid and will blunder,and decide the result accordingly.

This shows how much better the WinBoard approach to tablebase adjudication is: rather than assigning a result, it would just force the engines to move practically instantly (from their own tablebase), to see what they would make of it. That achieves the desired result (namely almost instant termination of the game) without introducing artifacts.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Cursed win at TCEC

Post by Lyudmil Tsvetkov » Wed Nov 16, 2016 9:56 am

gladius wrote:
Houdini wrote:
whereagles wrote:Have a look:

http://tcec.chessdom.com/archive.php?se=9&sf&ga=17

Engines showing 0.00 due to 50-move rule, but position was auto-adjudicated as an M72 TB win :D

Discuss :)
Not much discussion possible.
Both engines know that it's a draw (0.00) and play accordingly.
Suddenly the GUI decides otherwise and is clearly not following the rules of chess as implemented in the engines.
It's kinda ridiculous, but not very important.
Agreed, it should be a draw. As I posted in the TCEC chat, the adjudication should match the result if the engines had played the position out. In this game, it was a 50 move draw.
I would not agree here with Gary.
A tablebase win is a tablebase win. The position is simply won for white, so why declare it a draw? If both engines assume it is 0.0, that is only their fault they still have not implemented the much more relevant 100-move draw rule instead of the well-outdated 50-move rule. (or, what is the longest tb win without captures/promotions/pawn move?)

I am not certain what FIDE says about the 50-move/100-move draw rule, but why should engines follow FIDE? Engines are at the cutting edge of progress and progress says abovementioned position is simply a win for the stronger side. It is simply time to implement longer draw rule than 50-moves.

That should be specified in some protocol though, I agree it was not quite fair to both Houdini and SF in terms of their lack of knowledge, but a win is a win.

User avatar
hgm
Posts: 22306
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Cursed win at TCEC

Post by hgm » Wed Nov 16, 2016 10:33 am

You seem to miss the point. The position is not a win, but a draw. And both engines saw that correctly.

So the relevant question is: if the position is a draw, why declare a win?

Nay Lin Tun
Posts: 273
Joined: Mon Jan 16, 2012 5:34 am

Re: Cursed win at TCEC

Post by Nay Lin Tun » Wed Nov 16, 2016 10:54 am

hgm wrote:
So the relevant question is: if the position is a draw, why declare a win?
That is why ARB accused as SCAM years ago. :oops:

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Cursed win at TCEC

Post by Lyudmil Tsvetkov » Wed Nov 16, 2016 11:14 am

hgm wrote:You seem to miss the point. The position is not a win, but a draw. And both engines saw that correctly.

So the relevant question is: if the position is a draw, why declare a win?
as said, it is a tb win in 72 moves. engines go to move 50, 100 plies no captures, pawn moves, so 0.0 is returned, and with that the game ends for the engines. However, 22 moves later on, the win arrives. you know that much better than me, so I wonder why you take this stance.

if you mean, TCEC adjudication rules are not clear enough, that certainly migth be the case, though the position is still won for white.

That is what TCEC rules state regarding Adjudication:

A game can be drawn by the normal 3-fold repetition rule or the 50-move rule. However, a game can also be drawn at move 40 or later if the eval from both playing engines are within +0.05 to -0.05 pawns for the last 5 moves, or 10 plies. If there is a pawn advance, or a capture by any kind, this special draw rule will reset and start over. In the website this rule is shown as "TCEC draw rule" with a number indicating how many plies there are left until it kicks in. It will adjudicate as won for one side if both playing engines have an eval of at least 6.50 pawns (or -6.50 in case of a black win) for 4 consecutive moves, or 8 plies - this rule is in effect as soon as the game starts. In the website this rule is shown as "TCEC win rule" with a number indicating how many plies there are left until it kicks in. Cutechess will also adjudicate 5-men or less tablebase endgame positions automatically.



Problem is, how do you reconcile 1) the 5-fold scores within 0.05/-0.05 of both engines, with 2) the adjudication by Cutechess of tablebase positions?

Both drawing factors actually occured in the game, but I guess Cutechess adjudicated before the draw according to the special TCEC drawing rule could come into force. Anywhere written that GUI adjudication takes precedence to the special TCEC drawing rule, or this occurs automatically due to some technical stuff?

In any case, for me it would be wrong to adjudicate an objectively won position as a draw.

User avatar
Guenther
Posts: 2271
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Contact:

Re: Cursed win at TCEC

Post by Guenther » Wed Nov 16, 2016 11:36 am

Lyudmil Tsvetkov wrote:
hgm wrote:You seem to miss the point. The position is not a win, but a draw. And both engines saw that correctly.

So the relevant question is: if the position is a draw, why declare a win?
Lyudmil Tsvetkov wrote: That is what TCEC rules state regarding Adjudication:

A game can be drawn by the normal 3-fold repetition rule or the 50-move rule.
in contradiction to (if adjudicated by non-syzygy tablebases, what happened...)
Lyudmil Tsvetkov wrote: Cutechess will also adjudicate 5-men or less tablebase endgame positions automatically.
You again misssed the point. TCEC plays by the 50 moves rule... see again above. They just had set up another rule inforced by tablebases incapable of doing what a main rule says.

User avatar
hgm
Posts: 22306
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Cursed win at TCEC

Post by hgm » Wed Nov 16, 2016 11:41 am

Lyudmil Tsvetkov wrote:as said, it is a tb win in 72 moves.
Wrong. It is a tablebase draw in 50 moves. It is only a tablebase win in the sense that bare King versus bare King can be a tablebase mate in 3 if you use a flawed-enough tablebase.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Cursed win at TCEC

Post by Lyudmil Tsvetkov » Wed Nov 16, 2016 11:48 am

Guenther wrote:
Lyudmil Tsvetkov wrote:
hgm wrote:You seem to miss the point. The position is not a win, but a draw. And both engines saw that correctly.

So the relevant question is: if the position is a draw, why declare a win?
Lyudmil Tsvetkov wrote: That is what TCEC rules state regarding Adjudication:

A game can be drawn by the normal 3-fold repetition rule or the 50-move rule.
in contradiction to (if adjudicated by non-syzygy tablebases, what happened...)
Lyudmil Tsvetkov wrote: Cutechess will also adjudicate 5-men or less tablebase endgame positions automatically.
You again misssed the point. TCEC plays by the 50 moves rule... see again above. They just had set up another rule inforced by tablebases incapable of doing what a main rule says.
It is you and Harm that miss the point, not me.

[d]K5Q1/8/8/8/5bb1/6k1/8/8 b - - 0 72[/d]

this is the end position, and it is a white win, anyone can confirm.


for example, see wikipedia article on pwanless endgames:
https://en.wikipedia.org/wiki/Pawnless_chess_endgame

Specifically, for queen vs 2 bishops, the article says:
Queen versus two bishops: A queen has a theoretical forced win against two bishops in most positions, but the win may require up to seventy-one moves (a draw can be claimed after fifty moves under the rules of competition, see fifty-move rule); there is one drawing fortress position for the two bishops

So, it is a win, but the engine 100-plies counter gets it as a draw.

Nay Lin Tun
Posts: 273
Joined: Mon Jan 16, 2012 5:34 am

Re: Cursed win at TCEC

Post by Nay Lin Tun » Wed Nov 16, 2016 11:48 am

The point is simple. Before the adjudication, there was KQPP from stockfish and KRBB from move 55. [d]1r6/K2Q1b2/8/6k1/PP3b2/8/8/8 b - - 1 55
If there were no tablebase and no 50 moves rules there will be no way Houdini lose the game. However due to 50 moves , Houdini(same to stockfish) accessed TB partially only and Houdini chose the inferior moves from the tables and lose the game. However the moves Houdini chose were all TB draw within 50 moves of tablebase.

For example, the conditions are similar to announcement to the students like that
1. All questions will come from this book
2. You dont need to read beyond 50 pages of this book and
The exam questions come from beyond 50 pages

And the teacher said you have responsibility to read this book thoroughly you failed the exam cos you didnt read the book.

So condition 2 contradict/mislead condition 1 and the game finally ended unfavorable/unfair result.
Last edited by Nay Lin Tun on Wed Nov 16, 2016 11:56 am, edited 2 times in total.

Post Reply