Draws are getting cheaper, are 15W enough?

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
lkaufman
Posts: 3725
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Draws are getting cheaper, are 15W enough?

Post by lkaufman » Mon Sep 16, 2019 11:42 pm

Ozymandias wrote:
Mon Sep 16, 2019 10:35 pm
lkaufman wrote:
Mon Sep 16, 2019 5:14 pm
Stalemate is too rare for that rule to do much, although of course it will change some basic endgames. Much more significant is addressing bare king (making king and minor vs king a win), and doing something about repetitions, at the very least those due to perpetual check should be either banned or scored something like .3 for the side giving perpetual, .7 for the other.
King plus piece should score better than Stalemate, but I wouldn't give it the full point. Perpetual... scoring like you propose is sensible, but how many checks are needed? Three repetitions I would disincentivize, by scoring it as only .4 (for both sides). The 50 moves could be .4 for the side making the last move, .6 for the other.
Well, to avoid new rules as much as possible I would just say that if a triple repetition occurs with one player in check each time, that is "perpetual check" and gets the special scoring. The exact scores given for these "draws" isn't so important, as long as the inferior side (or the one resorting to perpetual check) gets less than the other side it gives the players something to play for when a win isn't possible.
Komodo rules!

EroSennin
Posts: 125
Joined: Fri Apr 09, 2010 1:26 am

Re: Draws are getting cheaper, are 15W enough?

Post by EroSennin » Tue Sep 17, 2019 12:51 am

How drawish is Fischer random in correspondence?

User avatar
Ozymandias
Posts: 1134
Joined: Sun Oct 25, 2009 12:30 am

Re: Draws are getting cheaper, are 15W enough?

Post by Ozymandias » Tue Sep 17, 2019 6:23 am

lkaufman wrote:
Mon Sep 16, 2019 11:42 pm
Ozymandias wrote:
Mon Sep 16, 2019 10:35 pm
lkaufman wrote:
Mon Sep 16, 2019 5:14 pm
Stalemate is too rare for that rule to do much, although of course it will change some basic endgames. Much more significant is addressing bare king (making king and minor vs king a win), and doing something about repetitions, at the very least those due to perpetual check should be either banned or scored something like .3 for the side giving perpetual, .7 for the other.
King plus piece should score better than Stalemate, but I wouldn't give it the full point. Perpetual... scoring like you propose is sensible, but how many checks are needed? Three repetitions I would disincentivize, by scoring it as only .4 (for both sides). The 50 moves could be .4 for the side making the last move, .6 for the other.
Well, to avoid new rules as much as possible I would just say that if a triple repetition occurs with one player in check each time, that is "perpetual check" and gets the special scoring. The exact scores given for these "draws" isn't so important, as long as the inferior side (or the one resorting to perpetual check) gets less than the other side it gives the players something to play for when a win isn't possible.
Summarizing:
  • Stalemate gives an edge to the stalemating side.
  • The bare King gives a more unbalanced score.
  • 3 repetitions by check gives an edge to the checked side.
  • 3 repetitions without check underscore for both sides.
  • The 50-move rule underscores for the side making the last move.
  • Mutual agreement should be checked by the TD, to be a position were a win or an unbalanced score can't be forced.

lkaufman
Posts: 3725
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Draws are getting cheaper, are 15W enough?

Post by lkaufman » Tue Sep 17, 2019 4:11 pm

Ozymandias wrote:
Tue Sep 17, 2019 6:23 am
lkaufman wrote:
Mon Sep 16, 2019 11:42 pm
Ozymandias wrote:
Mon Sep 16, 2019 10:35 pm
lkaufman wrote:
Mon Sep 16, 2019 5:14 pm
Stalemate is too rare for that rule to do much, although of course it will change some basic endgames. Much more significant is addressing bare king (making king and minor vs king a win), and doing something about repetitions, at the very least those due to perpetual check should be either banned or scored something like .3 for the side giving perpetual, .7 for the other.
King plus piece should score better than Stalemate, but I wouldn't give it the full point. Perpetual... scoring like you propose is sensible, but how many checks are needed? Three repetitions I would disincentivize, by scoring it as only .4 (for both sides). The 50 moves could be .4 for the side making the last move, .6 for the other.
Well, to avoid new rules as much as possible I would just say that if a triple repetition occurs with one player in check each time, that is "perpetual check" and gets the special scoring. The exact scores given for these "draws" isn't so important, as long as the inferior side (or the one resorting to perpetual check) gets less than the other side it gives the players something to play for when a win isn't possible.
Summarizing:
  • Stalemate gives an edge to the stalemating side.
  • The bare King gives a more unbalanced score.
  • 3 repetitions by check gives an edge to the checked side.
  • 3 repetitions without check underscore for both sides.
  • The 50-move rule underscores for the side making the last move.
  • Mutual agreement should be checked by the TD, to be a position were a win or an unbalanced score can't be forced.
I would prefer that the total score given for the game always be 1 point, for several reasons. So I would say that for repetitions without check the side repeating the position for the third time gets less than half point, the other gets more, same as for fifty move rule.
Note that with these rules an unbalanced score is guaranteed, so mutual agreement to a 0.5 score makes no sense. I suppose with TD approval players could agree to a specific draw, for example if K plus rook pawn vs. K, agree to score it as stalemate.
One further improvement: for all draw scores, Black gets 0.1 more and White gets 0.1 less than the stated amounts. This should make the game roughly fair.
Komodo rules!

User avatar
Ozymandias
Posts: 1134
Joined: Sun Oct 25, 2009 12:30 am

Re: Draws are getting cheaper, are 15W enough?

Post by Ozymandias » Wed Sep 18, 2019 6:31 am

lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
I would prefer that the total score given for the game always be 1 point, for several reasons. So I would say that for repetitions without check the side repeating the position for the third time gets less than half point, the other gets more, same as for fifty move rule.
That's more neat, although somewhat unfair. Unlike the 50 move rule, where is very difficult to anticipate what will happen down the road, in the case of repetition without checks, the players accept the outcome willingly, which is why I proposed to have the same penalty for both sides. But ok, I will amend that.
lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
Note that with these rules an unbalanced score is guaranteed, so mutual agreement to a 0.5 score makes no sense. I suppose with TD approval players could agree to a specific draw, for example if K plus rook pawn vs. K, agree to score it as stalemate.
I hope it is, the whole point of revising the draw rules, is to avoid most of the games to fall under the same category when the scoreboard gets filled. As it stands, only a handful of games still reflect in the outcome, the struggle that lead there and who it was that led the charge. Unfortunately, there are tournament situations where both sides may be happy with half a point and shake on that, well before it becomes necessary. I wasn't thinking that the TD should be bothered, shortening the length of games that are headed towards an unbalanced score, but rather to make sure that the games are playing in good "fighting spirits". To that end, either the TD keeps an eye on the games, or the draw offer is simply taken out of the equation. What do you think about that last option?
lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
One further improvement: for all draw scores, Black gets 0.1 more and White gets 0.1 less than the stated amounts. This should make the game roughly fair.
You're probably right, but without hard data at hand, I don't know how the actual numbers would play out. I would rather see the rules in action and, after a sufficient number of games have been played, make the necessary adjustments. I will pass the proposal with a corollary.

User avatar
Ovyron
Posts: 2754
Joined: Tue Jul 03, 2007 2:30 am

Re: Draws are getting cheaper, are 15W enough?

Post by Ovyron » Wed Sep 18, 2019 7:48 am

lkaufman wrote:
Mon Sep 16, 2019 2:27 pm
Does anyone run correspondence or freestyle events with rules designed to avoid or minimize draws, such as specified one-sided openings or startpositions, or modifications of the normal draw rules, or anything else?
Yes, Russian website crazy-chess.ru did this on 2017, they managed to find positions where it was unclear who had the advantage, and when people played them out with engine assistance, there were no draws, and white wins and black wins were close to 50%.

Here are the positions:













(credits to Programmist)

What was fun about some of these is that engines (of 2017) would give a huge advantage to some side. However, if the side managed to defend well then the advantage would jump out to the other side, and the chances of drawing would remain low.

If these positions are useful then there's probably a way to find other ones.
Great spirits have always encountered violent opposition from mediocre minds.

lkaufman
Posts: 3725
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Draws are getting cheaper, are 15W enough?

Post by lkaufman » Wed Sep 18, 2019 3:25 pm

Ozymandias wrote:
Wed Sep 18, 2019 6:31 am
lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
I would prefer that the total score given for the game always be 1 point, for several reasons. So I would say that for repetitions without check the side repeating the position for the third time gets less than half point, the other gets more, same as for fifty move rule.
That's more neat, although somewhat unfair. Unlike the 50 move rule, where is very difficult to anticipate what will happen down the road, in the case of repetition without checks, the players accept the outcome willingly, which is why I proposed to have the same penalty for both sides. But ok, I will amend that.
lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
Note that with these rules an unbalanced score is guaranteed, so mutual agreement to a 0.5 score makes no sense. I suppose with TD approval players could agree to a specific draw, for example if K plus rook pawn vs. K, agree to score it as stalemate.
I hope it is, the whole point of revising the draw rules, is to avoid most of the games to fall under the same category when the scoreboard gets filled. As it stands, only a handful of games still reflect in the outcome, the struggle that lead there and who it was that led the charge. Unfortunately, there are tournament situations where both sides may be happy with half a point and shake on that, well before it becomes necessary. I wasn't thinking that the TD should be bothered, shortening the length of games that are headed towards an unbalanced score, but rather to make sure that the games are playing in good "fighting spirits". To that end, either the TD keeps an eye on the games, or the draw offer is simply taken out of the equation. What do you think about that last option?
lkaufman wrote:
Tue Sep 17, 2019 4:11 pm
One further improvement: for all draw scores, Black gets 0.1 more and White gets 0.1 less than the stated amounts. This should make the game roughly fair.
You're probably right, but without hard data at hand, I don't know how the actual numbers would play out. I would rather see the rules in action and, after a sufficient number of games have been played, make the necessary adjustments. I will pass the proposal with a corollary.
Regarding repetitions, many are forced for at least one side, so awarding less than a full point is unfair to the side who has no choice. With the rule based on who repeats, at least each player will know who is to be blamed for the draw. Regarding "draws" by agreement, the only reason to allow it is to save time when the outcome is obvious, as in the stalemate example I gave. You could say that the players may ask the TD to call the game a draw by one of the specific rules if it is clear that this outcome is inevitable. There will be lots of games where it is obvious that a fifty move rule draw is inevitable, and having to play out fifty pointless moves is a bit much to ask of the players. As for the .1 offset by color, that was a proposal for normal OTB chess by Ed Epp which he promoted for decades; it is about fair for OTB chess. I agree it may not be fair for corr. chess with these other rules, we do need data. But some offset is needed, as White will obviously get at least the minimum 0.6 (or whatever the minumum draw score for the superior side is) with correct play, so it seems the offset should be enough to reduce that minimum to 0.5.
Komodo rules!

lkaufman
Posts: 3725
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Draws are getting cheaper, are 15W enough?

Post by lkaufman » Wed Sep 18, 2019 3:30 pm

Ovyron wrote:
Wed Sep 18, 2019 7:48 am
lkaufman wrote:
Mon Sep 16, 2019 2:27 pm
Does anyone run correspondence or freestyle events with rules designed to avoid or minimize draws, such as specified one-sided openings or startpositions, or modifications of the normal draw rules, or anything else?
Yes, Russian website crazy-chess.ru did this on 2017, they managed to find positions where it was unclear who had the advantage, and when people played them out with engine assistance, there were no draws, and white wins and black wins were close to 50%.

Here are the positions:













(credits to Programmist)

What was fun about some of these is that engines (of 2017) would give a huge advantage to some side. However, if the side managed to defend well then the advantage would jump out to the other side, and the chances of drawing would remain low.

If these positions are useful then there's probably a way to find other ones.
I suppose these were all played with normal chess rules rather than with chess960 castling, although each side looks like a 960 position in each diagram. Were there really "no" draws, or just not too many? Hard to imagine no draws at all. I could try any of them with the shootout feature of Fritz GUI with top engines to see.
Komodo rules!

User avatar
Ozymandias
Posts: 1134
Joined: Sun Oct 25, 2009 12:30 am

Re: Draws are getting cheaper, are 15W enough?

Post by Ozymandias » Wed Sep 18, 2019 4:18 pm

lkaufman wrote:
Wed Sep 18, 2019 3:25 pm
You could say that the players may ask the TD to call the game a draw by one of the specific rules if it is clear that this outcome is inevitable. There will be lots of games where it is obvious that a fifty move rule draw is inevitable, and having to play out fifty pointless moves is a bit much to ask of the players.
I thought about changing the resign button, so that it becomes a pull-down menu where one may choose to give up the fraction of the point that seems inevitable to loose. In case the other player doesn't agree, there's always the option to call the TD to arbitrate. We don't need an agreement, and even if we had one, it wouldn't really be about drawing, more like terms of surrender. These new rules would be a real game changer, let's see if they have the courage to try them out, over at InfinityChess.

lkaufman
Posts: 3725
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Draws are getting cheaper, are 15W enough?

Post by lkaufman » Wed Sep 18, 2019 4:55 pm

Ovyron wrote:
Wed Sep 18, 2019 7:48 am
lkaufman wrote:
Mon Sep 16, 2019 2:27 pm
Does anyone run correspondence or freestyle events with rules designed to avoid or minimize draws, such as specified one-sided openings or startpositions, or modifications of the normal draw rules, or anything else?
Yes, Russian website crazy-chess.ru did this on 2017, they managed to find positions where it was unclear who had the advantage, and when people played them out with engine assistance, there were no draws, and white wins and black wins were close to 50%.

Here are the positions:













(credits to Programmist)

What was fun about some of these is that engines (of 2017) would give a huge advantage to some side. However, if the side managed to defend well then the advantage would jump out to the other side, and the chances of drawing would remain low.

If these positions are useful then there's probably a way to find other ones.
I used the shootout feature to play out the first position with ten different engines at 2' + 1". Result: 7 out of 10 games drawn! So even this radical idea doesn't really help.
Komodo rules!

Post Reply