RSIF as alternative to armageddon

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.
Post Reply
Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 12:28 am

Inspired from a post by Larry at viewtopic.php?f=2&t=71367
on finding a way to eliminate draw result in a chess game without too much rule and game changes mainly for humans and perhaps for engines too.

Introducing RSIF (Repetition Stalemate Insufficient Fifty) These are just short names for draw conditions. 3-fold repetitions, stalemate, insufficient mating material and the fifty move rule.

RS - white wins
IF - black wins

I modify Deuterium to optimize its play on this RSIF variant. I run couple of combinations on those conditions and eventually came up with RSIF combo with close result based from Deuterium RSIF.

Test results after 200 games, at TC 1m+0.1s, using noomen 3 move opening suite.

Image

I am surprised that white takes the lead here, because getting insufficient material and fifty move rule is seemed easy perhaps at least for humans. Example RvR, RvB, RvBN endings are a win for black, just try to avoid the repetition. There seems to be a secret on the repetition though where white will try to repeat positions to his advantage. And of course my implementation is not optimal yet to play each side with those conditions.

Download test games at https://drive.google.com/file/d/1Q4d3Wf ... sp=sharing

Funny ending.
Black promotes to a KNIGHT to get the insufficient and wins, there is no need for long checkmate :)

carldaman
Posts: 1708
Joined: Sat Jun 02, 2012 12:13 am

Re: RSIF as alternative to armageddon

Post by carldaman » Sun Jul 28, 2019 12:52 am

New and interesting, but rather significant/radical rule changes, imo.

Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 1:05 am

carldaman wrote:
Sun Jul 28, 2019 12:52 am
New and interesting, but rather significant/radical rule changes, imo.
This indeed need some changes to the engine, but for humans, they already know the rules. Perhaps a little bit of adjustment and practice.

Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 1:14 am

Ferdy wrote:
Sun Jul 28, 2019 12:28 am
Test results after 200 games, at TC 1m+0.1s, using noomen 3 move opening suite.
I got a mistake on the number of plies, noomen 3move is a 6-ply opening set. But in the game there are only 4 book moves it should have been 6 book moves.
[Event "RSIF game"]
[Site "?"]
[Date "2019.07.27"]
[Round "1"]
[White "Deuterium v2019.2.37.64 rsif_white"]
[Black "Deuterium v2019.2.37.64 rsif_black"]
[Result "1-0"]
[ECO "D00"]
[GameDuration "00:02:06"]
[GameEndTime "2019-07-27T21:31:33.834 China Standard Time"]
[GameStartTime "2019-07-27T21:29:27.061 China Standard Time"]
[GameVariant "RSIF"]
[Opening "Levitsky attack (Queen's Bishop attack)"]
[PlyCount "145"]
[TimeControl "60+0.1"]

1. d4 { book } 1... d5 { book } 2. Bg5 { book } 2... f6 { book } 3. Bd2 { -0.10/15 1.7s } 3... e5 { +0.15/15 1.2s } 4. dxe5 { -0.18/15 2.5s } 4... Nc6 { +0.29/16 2.2s } 5. e4 { -0.35/16 2.1s } 5... dxe4 { +0.25/15 1.8s } 6. Nc3 { -0.22/17 2.5s } 6... fxe5 { +0.35/15 0.83s } 7. Nxe4 { 0.00/17 0.98s } 7... Nf6 { +0.07/18 2.2s } 8. Bd3 { -0.08/17 1.0s } 8... Bf5 { -0.01/17 1.2s } 9. Qf3 { -0.17/18 1.4s } 9... Nxe4 { +0.09/17 0.96s } 10. Bxe4 { -0.09/18 0.99s }
My mistake is on cutechess-cli setting.
-openings file= ... plies=4
Forgot to remove that plies option.

So in the test games you will only see 4 book moves.

Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 3:05 am

Run another test but this time using Tc 5m + 1s.
Result after 54/100 games.

Image

carldaman
Posts: 1708
Joined: Sat Jun 02, 2012 12:13 am

Re: RSIF as alternative to armageddon

Post by carldaman » Sun Jul 28, 2019 3:36 am

Ferdy wrote:
Sun Jul 28, 2019 1:05 am
carldaman wrote:
Sun Jul 28, 2019 12:52 am
New and interesting, but rather significant/radical rule changes, imo.
This indeed need some changes to the engine, but for humans, they already know the rules. Perhaps a little bit of adjustment and practice.
Knowing the rules shouldn't be that hard, but altering and skewing all endgame theory?

lkaufman
Posts: 3684
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: RSIF as alternative to armageddon

Post by lkaufman » Sun Jul 28, 2019 4:17 am

Ferdy wrote:
Sun Jul 28, 2019 12:28 am
Inspired from a post by Larry at viewtopic.php?f=2&t=71367
on finding a way to eliminate draw result in a chess game without too much rule and game changes mainly for humans and perhaps for engines too.

Introducing RSIF (Repetition Stalemate Insufficient Fifty) These are just short names for draw conditions. 3-fold repetitions, stalemate, insufficient mating material and the fifty move rule.

RS - white wins
IF - black wins

I modify Deuterium to optimize its play on this RSIF variant. I run couple of combinations on those conditions and eventually came up with RSIF combo with close result based from Deuterium RSIF.

Test results after 200 games, at TC 1m+0.1s, using noomen 3 move opening suite.

Image

I am surprised that white takes the lead here, because getting insufficient material and fifty move rule is seemed easy perhaps at least for humans. Example RvR, RvB, RvBN endings are a win for black, just try to avoid the repetition. There seems to be a secret on the repetition though where white will try to repeat positions to his advantage. And of course my implementation is not optimal yet to play each side with those conditions.

Download test games at https://drive.google.com/file/d/1Q4d3Wf ... sp=sharing

Funny ending.
Black promotes to a KNIGHT to get the insufficient and wins, there is no need for long checkmate :)
Your result is indeed strange, since fifty move rule is the normal termination of a game where no one can make progress, and insufficient material is far more common than stalemate. Probably the problem is that the engine can't form a long-term plan to utilize the rules. For Black it would be to trade queens at every reasonable opportunity, since most forced repetitions are perpetual checks, usually from a queen. Also Black has to avoid K and pawn vs (Black) king, as White can always force stalemate unless he can't save the pawn. Maybe Black is often heading for the "drawn" KPk ending, only to see too late that it will end in stalemate.
Logically, if stalemate is to count as a win, it should be a win for the stalemating side. Similarly, if insufficient material is to count as a win, it should be for the side with a minor piece (ok, B vs B same color is insufficient and ambiguous). K vs K will never happen without one side having material while the other doesn't, so simply restoring the old rules about stalemate and bare king being losses will cover these two cases. Then giving 50 move rule to Black and rep to White would eliminate draws in a more logical way. But your results suggest that strangely White would still come out ahead, at least with current engines. I thought of this scheme a while ago but figured it would favor Black. Maybe not?
Komodo rules!

User avatar
Ovyron
Posts: 2524
Joined: Tue Jul 03, 2007 2:30 am

Re: RSIF as alternative to armageddon

Post by Ovyron » Sun Jul 28, 2019 10:07 am

carldaman wrote:
Sun Jul 28, 2019 12:52 am
rather significant/radical rule changes, imo.
This is just smart draw odds.

I like it, I have been seeking for an answer to incoming chess draw death, and this could be a chess variant more fun than chess.

I wonder if RSIF is optimal, though, what about these?

RISF

White Wins = Repetition / Insufficient
Black Wins = Stalemate / Fifty Moves

RFSI

White Wins = Repetition / Fifty Moves
Black Wins = Stalemate / Insufficient

And the ones with white and black's winning conditions reversed? (IFRS, SFRI and SIRF) is there a way to test what is the best?

My support would to to FRIS (Fifty moves / Repetition = White wins, Insufficient / Stalemate = Black wins), because it's the easiest one to pronounce (though, that's arbitrary.)
Great spirits have always encountered violent opposition from mediocre minds.

Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 1:28 pm

Ferdy wrote:
Sun Jul 28, 2019 3:05 am
Run another test but this time using Tc 5m + 1s.
Result after 54/100 games.

Image
Test on Tc 5m+1 is done. Final result.

Image

Download the test games at https://drive.google.com/file/d/1HHT6nn ... sp=sharing

White still leads. I tried to check a game and it showed that being equal in material is not easy at all for black, because white will repeat positions. Black will not repeat to avoid outright defeat and in doing so it is slowly brought to inferior positions which will lead to defeat. If white can check black and if the best reply would repeat position then black will eventually lose.

Sample game from the test.

Ferdy
Posts: 4079
Joined: Sun Aug 10, 2008 1:15 pm
Location: Philippines

Re: RSIF as alternative to armageddon

Post by Ferdy » Sun Jul 28, 2019 1:46 pm

lkaufman wrote:
Sun Jul 28, 2019 4:17 am
Your result is indeed strange, since fifty move rule is the normal termination of a game where no one can make progress, and insufficient material is far more common than stalemate. Probably the problem is that the engine can't form a long-term plan to utilize the rules. For Black it would be to trade queens at every reasonable opportunity, since most forced repetitions are perpetual checks, usually from a queen. Also Black has to avoid K and pawn vs (Black) king, as White can always force stalemate unless he can't save the pawn. Maybe Black is often heading for the "drawn" KPk ending, only to see too late that it will end in stalemate.
Logically, if stalemate is to count as a win, it should be a win for the stalemating side. Similarly, if insufficient material is to count as a win, it should be for the side with a minor piece (ok, B vs B same color is insufficient and ambiguous). K vs K will never happen without one side having material while the other doesn't, so simply restoring the old rules about stalemate and bare king being losses will cover these two cases. Then giving 50 move rule to Black and rep to White would eliminate draws in a more logical way. But your results suggest that strangely White would still come out ahead, at least with current engines. I thought of this scheme a while ago but figured it would favor Black. Maybe not?
I just finished with the TC 5m+1s, games are in https://drive.google.com/file/d/1HHT6nn ... sp=sharing

There is danger in repetition, see game 1 for example to see white's idea.


Post Reply