Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Final Report on Draw Problem in Correspondence Chess
- - By Ciron (**) Date 2015-08-26 18:03
Some days ago ChessBase published my Final Report in the matter OPEN LETTER to ICCF and SURVEY on my proposal of an enhanced score system:
http://en.chessbase.com/post/how-many-points-should-you-get-for-a-draw

The feedback on my Open Letter was remarkable, same as the number of entries to the survey (172 regarding all / 123 with proven ICCF membership).

I discussed all proposals, whether modifying my own suggestions or counter proposals. There are also some interesting new suggestions.

A clear majority favours the idea of a pilot project in corresponendence chess with the enhanced score system (=3/4 for stalemating opponent or being up a minor piece against the bare king).

One slight correction or addition to the ChessBase article:
There are 2 overviews showing the proportions of yes/no/abstention to the 5 main questions.
The first overview is related to ICCF members only (123), the second one to all entries (172). This info got lost for any reason.
Parent - By Ozymandias (****) Date 2015-08-26 23:32

>Another interesting proposal is to reward draws with black before move 40 with a small bonus.


+1

Edit: the bonus should be detracted from white's score.
Parent - - By Gr00vy (**) Date 2015-08-27 01:32 Upvotes 1
I think the matter of win-light is not going to make more decisive games, but instead direct the game to certain types of draws.  Especially opposite color bishops and a single pawn, and increase the "value" of a and h pawns, that draw because of stalemate.   And will tend to invalidate and need a change of egtb to show these as being win-light rather than simply drawn.  For awhile this will invalidate most engines, but surely now there will be a correspondence setting to include the new values created by win-light,   

The whole concept of win light will make a ton of games less interesting, because the point of the game will be directed to certain draw types that are weil known and avoided..   And may make the game closer to solved, in that win-light may be an easier threshold to get to than full practical wins, and may significantly increase the theoretical advantage for one side or the other making certain types of openings over earler, because the opening results significantly either win or win-light.   Where the opening would be ignored, because the other the opponent can always easily steer to draw, but now it's a penalized draw.  And it may ALWAYS be at LEAST a penalized draw.  At the end of the day, by changing the meta conditions of the game, you may instead of getting to interesting battles, it will devolve into known lines that result in at least penalized draws.
Parent - - By MarshallArts (***) Date 2015-08-27 01:48
You may well be right, Gr00vy. Very good post.
We shouldn't mess with the stalemate rule.

All these suggestions to alter the scoring of draws tend to have unintended consequences,
with an ugly impact on how the game will be played.

:neutral:
Parent - By Ozymandias (****) Date 2015-08-27 08:31
The only consequence registered so far, is the steering of usually drawn games, into stalemate positions. A stalemate may also be a draw, but it's more pleasant to some eyes, than throwing away material to reach a forced draw.
Parent - By Ozymandias (****) Date 2015-08-27 08:29
You're assuming:
1- A degree of control in the opening, I've only seen from the Anson-Evans team. Even then, they took some risks (with black) that could point otherwise.
2- That someone with said amount of opening knowledge, would settle for a "sure" stalemate, instead of gambling on a win (with white).
Parent - - By Kreuzfahrtschiff (***) Date 2015-08-27 08:32 Edited 2015-08-27 08:39
totally agree with your (Gr00vy) statement

i want no changes at all for chess

my opinion is to invent a new more complex version of chess, like 12x12 boards or so
Parent - - By Ozymandias (****) Date 2015-08-27 11:19

>i want no changes at all for chess


>my opinion is to invent a new more complex version of chess, like 12x12 boards or so


And that'd be a no-change-version of chess?!
Parent - - By Kreuzfahrtschiff (***) Date 2015-08-27 13:26
because it changes not the old game, its a new one

you have then 2 chess versions, thats NO change with original chess, do you understand? ;)
Parent - - By Ozymandias (****) Date 2015-08-27 15:58
Ok, but if it's a different game, how does it address the draw problem in chess?
Parent - By Kreuzfahrtschiff (***) Date 2015-08-27 18:03
i dont see it as problem, its just solved like checkers

go will be longer unsolved, because it has a 19x19 board, so we need more squares
Parent - - By Ciron (**) Date 2015-08-27 08:36 Edited 2015-08-27 08:43

> I think the matter of win-light is not going to make more decisive games, but instead direct the game to certain types of draws.


Pure speculation. The only way to find out the truth is to play a large number of test games under proper conditions.

In your scenario the only active part is White. It seems Black can only react on what White has in mind, who is going for promising lines that might either win or guarantee him a 3/4 point. But chess isn't that simple. Neither will you win tournaments with such approach, nor even win many 3/4 points. Black's ressources to achieve normal draws are still quite rich, even with an enhanced score system.

You miss the point that chess would become slightly more complex, more complicated, as both sides have to consider different strategies aiming for mate / stalemate / material win or draw. And I don't think engines will easily solve those long-term questions. Users might rather experience that engines for a while show the same values for many moves, that actually might lead to different results. This would be the time for human decision making, either in freéstyle chess or in correspondence chess.

Of course will there be changes in opening theory. But again, why assuming those will only help White?

As far as I could see from tests it seems to me you have to play for a win in order ro achieve even a 3/4 point. The other side has to make mistakes, if you want to achieve more than a normal draw. So the question might be put even this way: How can I complicate the game in a way that causes great danger for my opponent? Is that possible by avoiding danger for myself?

Well, as said before, there is no other way than to try it out. And it's not the only idea about decreasing the draw rate.
Parent - - By Gr00vy (**) Date 2015-08-27 15:06
It is pure speculation.   But to say that we have no way of predicting without experimentation is a bit specious. 

And I made no claim about white being superior.   What I said is that certain drawing lines have been avoided because the are drawish.     The deal with bunch of draws lines, is that they are draws not usually because of position repetition, but because they are known to be draws early and a huge chunk of those precisely because they logically resolve to drawn end games.    However, a ton of those drawn end games don't come down to bare king v king, but indeed to not having enough material to mate, or because the position resolves into common inevitable stalemates,   Those positions largely now become win-lights.

Imagine that this is a target.   A practical win is the smallest of bullseyes.     A win-light is a significantly larger bullseye, and getting to that bullseye is going to be such a large target that it will likely resolve to at least win-light will be known much earlier in the game.    Whether or not this is an advantage for white or black is going to be learned relatively early.    The assumption for white, is that there is an opening line, or lines regardless of blacks response always results in win-light or better.   In which case in this scoring system, the game is solved.   The assumption for black is that instead the game is zugzwang, and that no matter the opening line, the game can be steered to a win-light or better for black.   In which case the game is solved.   Mistakes will happen, but instead of a draw in the pocket, its a win light in the pocket.

It is true that for the experimentation period this is yet to be discovered, but once discovered it is going to be done.   The assumption is that it will be significantly easier to steer the game to win-light, and that it will be discovered that ease indeed makes the game inevitably steered to win light.   But chess has always been about victory, and that steering to draw has not been the desire of the stronger aggressor, but the defiance of the defender.   Under the new system the defender scoring .25 instead of .5 is a major change in the underlying meta-game.

The assumption that people will play for wins instead of playing for win-light, I think is a pretty big statement.    I think the decision to play for win v win-light is going to be a higher hurdle, most lines are going to be more drawish not less, but more win-lighty.    Expect to see lots of opposite colored bishops games.   Both sides are going to be willing to go there because of the scoring of win-light.   More careful pawn setups.   More value to A and H pawns.  The striving to stalemate instead of being seen as a mistake for the aggressor, and a victory for those behind snatching draw from the jaws of a loss.    Will instead being seen as a victory for the aggressor (win-light), and determined way earlier in the game, possibly so soon that the game becomes solved from opening to early middle game.  

An advantage for the interim, if you're trying to lessen engine influence, is that engines will take a while to be able to properly steer the game to win-light.   The balance of evaluation and egtb currently steer away from win light.  The most obviously "drawn" positions, and the ones to be avoided now, are going to instead change into win-light positions, and instead of being avoided will be sought out. I am willing to bet that currently the vast majority of 5 piece tables that have KB v KBP opposite colors are marked draw.   Most of these will need be marked Win-light, and the engine will need to be able to understand win-light as a setting.  So for some period of time there will be relative chaos because of the new conditions.  Those most able to deal with that will be most successful during the time period, even if they are not necessarily the "best" players.
Parent - - By Ozymandias (****) Date 2015-08-27 16:17

>for some period of time there will be relative chaos because of the new conditions.  Those most able to deal with that will be most successful during the time period, even if they are not necessarily the "best" players.


So far, in three tournaments, only six stalemates have been achieved. Only Arno has been able to repeat the feat. That hasn't brought him a lot of success, even though he's one of the best players.
Parent - - By Gr00vy (**) Date 2015-08-27 17:58
It is way too soon to have anything approaching arguable data.   I think this could be begun to be explored reasonably quickly by creating a special version of Stockfish v.regular stockfish, and give them a few thousand games or so to see what happens.   This would test whether the metagame change is and how exploitable.   You could start by simply making a stalemate a winning condition, and play that against standard stockfish to see if that does anything practical and how much.   IE just one search SEEKING out stalemates.. Then give preference to opposite color bishops and play that against the "stalemate is a winning condition version" and the regular version.    This would tend to give credence to the amount of variance and value this may create in number of draws v value of win-light vs winning.    Ideally you would want to do this with another engine as well, but it would be awhile to get someone else to give it a go with a credible engine.   And you would have to not use egtb's, and play to the end.    There would also need to be some form of adjudication to hard draws to determine if it is a win-light.  But I think these are relatively small technical hurdles to at least begin asking some of the questions.   The ultimate question of whether exploitation can achieve solvability of win-light will obviously take tons more work, and thought about opening types that tend to win-light endings, and how forcible they are.   But it will be a start down the path of answering the question.   And in the interim, human correspondence players can provide some interesting if largely anecdotal evidence to what it may mean.
Parent - - By Ozymandias (****) Date 2015-08-27 20:12
Are you saying that "Those most able to deal with that" can already give it a go, so that hey can "be most successful"? Or do they have to wait and see how it pans out?

BTW, a Stockfish version was already shown going for the stalemate (with some code modification), but IMO, engines aren't the main beneficiaries, of this rule change. It's aimed at corr and freestyle chess. Humans are the ones, better suited to exploit it. In that vein, only corr and freestyle chess tournaments, can shed some light on the matter.
Parent - By Chaotic Chess (****) Date 2015-08-27 20:41
Anyone can share the modified SF engine that can understand/go stalemate position? thank you
Parent - By Ciron (**) Date 2015-08-28 17:29
Thanks for your input. I don't agree with your assumptions, but I appreciate your contributions.

>What I said is that certain drawing lines have been avoided because the are drawish. The deal with bunch of draws lines, is that they are draws not usually because of position repetition, but because they are known to be draws early and a huge chunk of those precisely because they logically resolve to drawn end games.    However, a ton of those drawn end games don't come down to bare king v king, but indeed to not having enough material to mate, or because the position resolves into common inevitable stalemates,   Those positions largely now become win-lights.


Again, if a player tries to avoid drawish lines, it's mostly White. But anyway, opening lines are not chosen by only one side, both sides try to enter lines they prefer. And I am even not sure, whether your statement about people avoiding drawish lines is true in any respect. There are quite a lot of rather drawish lines in e.g. Italian Game, Ruy Lopez, Petroff, Scotch, Queen's Gambit, Queen's Pawn Openings, English Game, that are played over and over again despite the fact that they are drawish according to data bases. I guess, you heard of "Carlsen Openings". That's another modern idea. People don't look anymore for theoretical opening advantage, but for practical chances. And practical chances are not less (for both sides) in games with an enhanced core system. Btw, I made some tests with a modified Stockfish version, which values stalemating about +3, against the normal Stockfish. The question was: Can we keep on playing Gambits without taking too high risks? I wasn't too much surprised to discover that all sound (though rare) Gambits like Evans Gambit, Budapester Gambit, Benko and many others are still playable without significant difference. Yet, I don't take these tests too seriously; of course does the modified SF have no clue about strategies to go for a stalemate win right from the opening or the early middlegame.

> Expect to see lots of opposite colored bishops games. Both sides are going to be willing to go there because of the scoring of win-light.


Sorry, I can't follow you. Most games with opposite coloured bishops just lead to blocked positions. Just draws, no way to simplify and go for an endgame K+B+P vs. K. Such an endgame would rather occur after a real battle with lots of complications, but not just following the logical course of any opening line.

> More careful pawn setups.   More value to A and H pawns.


I don't think such general statement is right. You underestimate the dynamics in chess. Or do you think, people will switch to more static instead of dynamic systems? I would rather say, all pawns get slightly more value because of the chance to stalemate your opponent in the basic endgame K+P vs. K. But as said earlier, it's far from easy to achieve such an ending. The opponent has to make mistakes before. And if so, why not award the better player with a 3/4 point?

Btw, what is the exact meaning of your term "win-light"? Could you please define? I never heard this before.
Parent - - By Loboestepario (****) Date 2015-08-28 19:08
Hi Arno,

I do not think the increasing draw-percentage is a critical problem at this time. World Champions in ICCF tend to be able to achieve 3+ scores in the final or higher.
Do we play Finals to find a winner or to simply have more wins?
However if we find in the future that Finals were won with only a 2+ score then I would suggest to increase the reflection time from 10/50 to 10/90 or even 10/100.
I strongly believe the better player will squeeze out the win if given more time.
If this does not work then we have Chess960 or your proposal, but we cannot call them standard chess.

Gino Figlio
Parent - - By MarshallArts (***) Date 2015-08-28 20:48
Another idea that might work even better in corr chess (than in OTB events) is giving 3 points for a win and 1 for a draw, as in the soccer scoring system. The yearly London superGM tournament uses this scoring system.

:neutral:
Parent - By Ozymandias (****) Date 2015-08-29 07:37
Even though the problem may seem to be the same one, OTB and corr-freestyle face two different problems. Both have the same symptom (high draw rate), but the illnesses that cause it are distinguishable, although related. In OTB chess, we see players seeking or agreeing to a draw, because they're afraid of losing (money, ELO…), or have other non chess related circumstances, that make them want to shorten the game. In corr and freestyle, we see players constructing their whole opening repertoire, with the sole intent of drawing, knowing it can be done, and also knowing they aren't well equipped to fight for a win. This is a decision they take unilaterally, without the other player having much say in the matter, while in OTB, only GMs resort to that, as part of their tournament strategies, working only when their opponents also wants a draw. Yes, there's also a fear of losing (reputation), but this fear has its roots in certainty, rather than chance. To avoid draws, we're faced with a real risk of losing (specially with black), and not every player wants to do that, because they've already worked out the probabilities of winning versus losing. In OTB, you don't need to risk at the opening, to fight for a win, just take a look at Carlsen's games, if you don't want a draw, you can press on. In corr and freestyle, you may press as much as you want, while your opponent may as well be eating popcorn, if that's what he wanted to do.
Parent - - By Ciron (**) Date 2015-08-28 20:58 Edited 2015-08-28 21:00
Hi Gino,

> World Champions in ICCF tend to be able to achieve 3+ scores in the final or higher.


I doubt, we can speak of "World Champions" in the meaning of No. 1 players like Carlsen in o-t-b chess. There are a couple of players in correspondence chess who are able to win a Final. I guess, not less than 50 players, just to give any figure. It depends, whether those players qualify for a final. Sometimes weaker players qualify, whereas stronger players fail to qualify or make no efforts to qualify because of lack of motivation. With such a high rate of draws like in the two last finals luck is getting a more and more decisive factor to achieve even 1+. Six of eight players, who had White against Papenin in the 28th Final, won against him. He was completely disorganized (or should I say "out of sorts"), but managed to draw at least all his white games. So wasn't it pure luck to be among those players having White against him?
Table: https://www.iccf.com/event?id=37632
May be, there will be a lucky winner with +3, may be not... (I hope, not one of those players who beat Papenin), but that doesn't change much. Even participants of the final find it boring to play such tours with a high drawing rate. Why waiting for even more boring tours, before taking measures?

It's not only a problem of World Championships, it matters for all high-class tournaments. You know that the Bielecki Memorial was even 2 categories higher than the WCC Final.

More reflection time will not help, I guess. Those players at top usually rather profit from the fact that other players lose games because of "time trouble". More reflection time means even more draws and even more boring and longer tours. I wonder, whether you meant it seriously, it sounds like kidding ,-)

Anyway, it's time for a change and for experimenting before accepting a longer or shorter period of agony.

Arno
Parent - - By Loboestepario (****) Date 2015-08-29 00:28
Hi Arno,

There is no doubt that the last 10 finished finals had a winner with a +3 score at least.
The 28th is unfinished but the candidate winner has +2 and 1 unfinished game.
I am not aware of any surveys involving finalists that report boredom.
Is the problem extensive to other classes of players? For sure but we have to start somewhere.
The data you presented mostly involved WC Finals, if I had time I could dig stats on lower rated events. Not today.
Of course I was serious about the reflection time. If you agree that a faster time control would increase the draws even more then it is not counter-intuitive to suspect that increasing time control might be beneficial. Think of it as providing the better player more time to find the win.

Gino
Parent - By Ozymandias (****) Date 2015-08-29 07:38

>If you agree that a faster time control would increase the draws


I doubt he will, because it's the other way around.
Up Topic The Rybka Lounge / Computer Chess / Final Report on Draw Problem in Correspondence Chess

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill