Armageddon Openings released

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: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sat Aug 24, 2019 3:50 am

Nordlandia wrote:
Fri Aug 23, 2019 11:17 pm
lkaufman wrote:
Fri Aug 23, 2019 7:37 pm
Nordlandia wrote:
Fri Aug 23, 2019 7:05 pm
Is it complicated to teach Komodo armageddon mode. If this is something to invest time into. Armageddon involve riskier play, which in turn can be fun to watch.
Simply setting Contempt to a high value for White should make Komodo play Armageddon pretty well. Of course we could do a lot better than that if Armageddon became all the rage, but we can't put it many hours of work on something that will only be used on rare occasions. I'm running some tests on my hybrid idea (NBSC) using shootout but with multiple versions of Lc0, Stockfish, Komodo, and Komodo MCTS to see how fair it is.
Viewing draw as defeat is a fundemental change to the engine.
Not as much as you might imagine. The side who is ahead (normally White in these openings) plays to win, and the engine doesn't really concern itself with whether the alternative is a draw or loss, it is trying to increase its positive score to infinity. Accepting a draw as White or not claiming one as Black even with the score saying otherwise is a mistake, but Contempt will cover that well enough. Simplifying towards a draw is also handled well by Contempt. It's not perfect, but with high setting (100 would be a first guess) it should cover most of what true knowledge of the rule would do.
Komodo rules!

lkaufman
Posts: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sat Aug 24, 2019 3:54 am

pohl4711 wrote:
Fri Aug 23, 2019 8:17 am
M ANSARI wrote:
Fri Aug 23, 2019 6:17 am
Why not simply allow only white to castle? This forces white to attack and black to defend. If black survives and draws he deserves to be given a win!
??? Perhaps you should take a look at the download or on my website... Level 2 is called NBC (=No Black Castling). White can castle in both directions and Black is not allowed to castle.
Based on my shootout tests so far of my NSBC (no short Black castling) version, I may have to endorse your level 2 NBC, because Black is holding the draw much more than White is winning (15 to 10). But this is not enough games, it could easily turn around. I'll try to get a hundred games.
Komodo rules!

lkaufman
Posts: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sat Aug 24, 2019 7:38 pm

lkaufman wrote:
Sat Aug 24, 2019 3:54 am
pohl4711 wrote:
Fri Aug 23, 2019 8:17 am
M ANSARI wrote:
Fri Aug 23, 2019 6:17 am
Why not simply allow only white to castle? This forces white to attack and black to defend. If black survives and draws he deserves to be given a win!
??? Perhaps you should take a look at the download or on my website... Level 2 is called NBC (=No Black Castling). White can castle in both directions and Black is not allowed to castle.
Based on my shootout tests so far of my NSBC (no short Black castling) version, I may have to endorse your level 2 NBC, because Black is holding the draw much more than White is winning (15 to 10). But this is not enough games, it could easily turn around. I'll try to get a hundred games.
After 60 games (at six different time limits, ten different engines), the score was 24 wins for White, 36 draws (which count as wins for Black), which seems pretty convincing, so even NSBC doesn't seem to give White equal chances in Armageddon. So I'm now running NBC (no Black castling) the same way. NBC does have the advantage that normal opening books are not so bad as they are for SALC since the bias towards preparing to castle short doesn't particularly hurt Black as it does with SALC or with NSBC. As a human GM, I would prefer to play White with NBC Armageddon, and I'm pretty sure most other GMs would also. Of course it's more pleasant to play White, but I mean strictly thinking about results, say a million dollars was riding on the game, I think most GMs would choose White. It's quite normal to sac a pawn to prevent castling, and many respected gambits seem to offer less compensation than NBC. But I'm rather expecting a close result this time. Maybe engines are better at defense (relative to other engines) than humans are (relative to other humans). What's fair between engines may not be fair between humans.
Komodo rules!

pohl4711
Posts: 1119
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Armageddon Openings released

Post by pohl4711 » Sun Aug 25, 2019 3:54 am

lkaufman wrote:
Sat Aug 24, 2019 7:38 pm
It's quite normal to sac a pawn to prevent castling, and many respected gambits seem to offer less compensation than NBC. But I'm rather expecting a close result this time. Maybe engines are better at defense (relative to other engines) than humans are (relative to other humans). What's fair between engines may not be fair between humans.
Looking forward to the results...
Of course, Armageddon openings were designed for cumpterchess = engines. For humans it seems much better to me, to use different thinking times for Armageddon, like it is done already today. Human playing strength is more affected by a reduced thinking-time, than engine playing strenth is.

lkaufman
Posts: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sun Aug 25, 2019 5:31 am

pohl4711 wrote:
Sun Aug 25, 2019 3:54 am
lkaufman wrote:
Sat Aug 24, 2019 7:38 pm
It's quite normal to sac a pawn to prevent castling, and many respected gambits seem to offer less compensation than NBC. But I'm rather expecting a close result this time. Maybe engines are better at defense (relative to other engines) than humans are (relative to other humans). What's fair between engines may not be fair between humans.
Looking forward to the results...
Of course, Armageddon openings were designed for cumpterchess = engines. For humans it seems much better to me, to use different thinking times for Armageddon, like it is done already today. Human playing strength is more affected by a reduced thinking-time, than engine playing strenth is.
So far the NBC result is 33 wins for White to 23 draws (which count as wins) for Black, roughly the opposite of what I got with NSBC. So subject to needing more games, it seems that the fair offset for Armageddon is about midway between NSBC and NBC. I'll get more games, perhaps it will clarify. Of course you are correct that human playing strength is more affected by reduced thinking time than engine playing strength. But we don't really know how much of a time handicap is needed for Armageddon in long-time limit games; there were just a few games played with a total of about one hour as U.S. Championship tiebreakers some years ago, which suggested that maybe 2 to 1 was about right. But it also depends on the level of the players. I'm pretty sure that the proper time handicap is much larger at 2800 level than at 2500, and much larger at 2500 than at 2200, etc. There is probably no simple formula for fair time-handicap Armageddon for human players. If you look at the percentage of draws in top level events like the current one in St. Louis, it is obvious that something must be done for top-level human chess to remain interesting. I don't blame the players, it is the rules that need revision. Maybe chess 960 will catch on, but if not we need to do something to enhance interest in top human chess.
Komodo rules!

leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 6:23 am

Re: Armageddon Openings released

Post by leavenfish » Sun Aug 25, 2019 7:58 am

lkaufman wrote:
Sun Aug 25, 2019 5:31 am

If you look at the percentage of draws in top level events like the current one in St. Louis, it is obvious that something must be done for top-level human chess to remain interesting. I don't blame the players, it is the rules that need revision. Maybe chess 960 will catch on, but if not we need to do something to enhance interest in top human chess.
"Something 'MUST' be done! Not so fast...most of the games are VERY interesting and complicated fights. I don't know why certain people keep moaning about the draws. YES, there are a lot of draws....generally very interesting draws. The last two days Anand and his opponents in particular has have fascinating games. Slightly faster time controls would be okay. 960...isn't chess, sorry.

But to just look at the results and think 'something needs to change'...iI don't think so.

Viewership is quite high - and there are more outlets than ever for viewers to watch the games and there are more top level tournaments than ever. There is NO lack of interest.

That said...I can live with some changes. For example: Let WINS ONLY count - let draws and losses amount to nothing in the end (that's right...a loss would not necessarily 'hurt' you, only lack of effort). Sure, you can still count a 1 for a win, a 1/2 for a draw and a 0 for a loss. But I rather like the idea of dividing the prize fund up by the number games won at the end of the event.

The idea goes something like this: Lets say there is[b] $100,000[/b] in prizes. At the end of they tourney lets say there are 5 wins: 2 for Anand, 2 for Carlsen and 1 for Nakamura. Anand and Carlsen get $40,000, Nakamua $20,000. Everybody else - $0.

Now to be realistic, they and all the others should maybe get $5,000 (just pulling a number, maybe less) for showing up and playing. You might even set aside 5-10% for whomever eventually scores the most points...but let the number of wins decide the vast majority of prize distribution.

The point is that even late in a tourney when someone is 'out of the running' - when they may not want to be pulling out their 'best stuff' in their openings -(wanting to save it for another tourney when they have they still have something real to play for)... or simply would normally have no real reason to exert themselves given that they are not going to finish well....they now would.

Naked Capitalism does tend to bring out the best in people. :wink:

User avatar
Nordlandia
Posts: 2477
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Armageddon Openings released

Post by Nordlandia » Sun Aug 25, 2019 12:14 pm

Normal chess is one of 960 possible starting arrangement of FRC. So it's wrong to say that Chess960 ain't chess. The fundamental differences is limited theory if not anything (too many starting postions to memorise) and castling rules. Other than that it's same rules as in normal chess.

lkaufman
Posts: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sun Aug 25, 2019 4:38 pm

leavenfish wrote:
Sun Aug 25, 2019 7:58 am
lkaufman wrote:
Sun Aug 25, 2019 5:31 am

If you look at the percentage of draws in top level events like the current one in St. Louis, it is obvious that something must be done for top-level human chess to remain interesting. I don't blame the players, it is the rules that need revision. Maybe chess 960 will catch on, but if not we need to do something to enhance interest in top human chess.
"Something 'MUST' be done! Not so fast...most of the games are VERY interesting and complicated fights. I don't know why certain people keep moaning about the draws. YES, there are a lot of draws....generally very interesting draws. The last two days Anand and his opponents in particular has have fascinating games. Slightly faster time controls would be okay. 960...isn't chess, sorry.

But to just look at the results and think 'something needs to change'...iI don't think so.

Viewership is quite high - and there are more outlets than ever for viewers to watch the games and there are more top level tournaments than ever. There is NO lack of interest.

That said...I can live with some changes. For example: Let WINS ONLY count - let draws and losses amount to nothing in the end (that's right...a loss would not necessarily 'hurt' you, only lack of effort). Sure, you can still count a 1 for a win, a 1/2 for a draw and a 0 for a loss. But I rather like the idea of dividing the prize fund up by the number games won at the end of the event.

The idea goes something like this: Lets say there is[b] $100,000[/b] in prizes. At the end of they tourney lets say there are 5 wins: 2 for Anand, 2 for Carlsen and 1 for Nakamura. Anand and Carlsen get $40,000, Nakamua $20,000. Everybody else - $0.

Now to be realistic, they and all the others should maybe get $5,000 (just pulling a number, maybe less) for showing up and playing. You might even set aside 5-10% for whomever eventually scores the most points...but let the number of wins decide the vast majority of prize distribution.

The point is that even late in a tourney when someone is 'out of the running' - when they may not want to be pulling out their 'best stuff' in their openings -(wanting to save it for another tourney when they have they still have something real to play for)... or simply would normally have no real reason to exert themselves given that they are not going to finish well....they now would.

Naked Capitalism does tend to bring out the best in people. :wink:
What you are proposing is more radical than most of the ideas we have discussed. Counting a draw as a loss (or near-loss) turns the game into a bluffing game, who can bluff the opponent into making an unsound move to avoid a draw? Even just pure Armageddon (Black wins draws) is better than this. But chess is too drawish for pure Armageddon to be reasonable. Armageddon with some restriction on Black to make things more equal seems the most likely to catch on and to be reasonably fair. It can be less time (but at top level at LTC it would probably have to be 3 to1 or so), it can be a castling restriction like we are talking about here, it could be a second White move before Black plays (not quite enough to equalize the game, but perhaps close enough for human play), it could be a bad move played (for example Black must start 1...f6?), but that seems inelegant. A modest idea (proposed by Ed Epp decades ago) is simply to give Black 0.6, White 0.4 in case of a draw, which is about fair and seems to me to have no downside, but it won't help very much. My personal favorite is to outlaw perpetual check (3rd repetition with check is illegal), but a rule change rather than a scoring change is much harder to get adopted.
Komodo rules!

leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 6:23 am

Re: Armageddon Openings released

Post by leavenfish » Sun Aug 25, 2019 7:42 pm

Again, I personally do not really think we need to change much at the top level – there is no lack of interest nor would it seem, lack of sponsorship…just some who moan at the number of draws although the content is really quite high. That is my point. However, speeding up the time controls slightly should always be considered as opening theory continues to advance. That only makes sense.

But…a clear way to get more exiting games - for those who crave closure/decisiveness in games - would indeed be to only count WINS when it comes to divvying out prize money at the end of the tournament. It would bring more of a Tal like quality to the play as opposed to an Ulf Anderson to go to the other side of the spectrum. That is not an easy task in a day and age of advanced theory where time controls are still quite 'slow' at the top level for standard chess.

Chess is not baseball with all its counting of statistics and records so we should not fear certain changes. Of course, any change over what currently exists would have 'some' downside.

But ‘encouraging players’ to risk more during a game (rather than create artificial time imbalances, add additional games and such) by maybe playing more non-standard openings (a very loose equivalent in the end to 360) or chose a sharper line when such become optional during a game is not a bad thing. It would certainly serve to encourage some which play incredibly safe (especially when they are ‘out of the top money’, don’t want to play novelties when they are cellar dwellers, not ‘motivated’, etc) during these top tournaments – some (remember Ulf Andersson?) are quick to draw/play it safe just to preserve their ratings. That is arguably more of a thing now as there is more money and a ‘circuit’ at the top level these days. When these same players are going to play each other again shortly (on the circuit), they simply are not going to ‘waste’ ideas when they are mid-pack or less…a change in the prize pay out would certainly change that.

Which live games would your average chess fan prefer to follow – those of Giri or Jobava? My guess would be Jobava. One thing this might well change is that someone not at all adverse to a draw like Giri would need to mix his game up more or risk sponsors opting to invite someone more ‘hungry’ like a Jobava.

In a way it is the system which keeps the same players playing each other which is at fault as well.

I’m watching right now, R8: Carlsen – Liren. Like those Anand games I mentioned, there is nothing at all ‘boring’ about the chess I am seeing. It is likely to end in a draw…but boy was this exciting! And at move 25 they both have about 1 1/2 hrs left. That is just...weird.

lkaufman
Posts: 3757
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Armageddon Openings released

Post by lkaufman » Sun Aug 25, 2019 8:18 pm

leavenfish wrote:
Sun Aug 25, 2019 7:42 pm
Again, I personally do not really think we need to change much at the top level – there is no lack of interest nor would it seem, lack of sponsorship…just some who moan at the number of draws although the content is really quite high. That is my point. However, speeding up the time controls slightly should always be considered as opening theory continues to advance. That only makes sense.

But…a clear way to get more exiting games - for those who crave closure/decisiveness in games - would indeed be to only count WINS when it comes to divvying out prize money at the end of the tournament. It would bring more of a Tal like quality to the play as opposed to an Ulf Anderson to go to the other side of the spectrum. That is not an easy task in a day and age of advanced theory where time controls are still quite 'slow' at the top level for standard chess.

Chess is not baseball with all its counting of statistics and records so we should not fear certain changes. Of course, any change over what currently exists would have 'some' downside.

But ‘encouraging players’ to risk more during a game (rather than create artificial time imbalances, add additional games and such) by maybe playing more non-standard openings (a very loose equivalent in the end to 360) or chose a sharper line when such become optional during a game is not a bad thing. It would certainly serve to encourage some which play incredibly safe (especially when they are ‘out of the top money’, don’t want to play novelties when they are cellar dwellers, not ‘motivated’, etc) during these top tournaments – some (remember Ulf Andersson?) are quick to draw/play it safe just to preserve their ratings. That is arguably more of a thing now as there is more money and a ‘circuit’ at the top level these days. When these same players are going to play each other again shortly (on the circuit), they simply are not going to ‘waste’ ideas when they are mid-pack or less…a change in the prize pay out would certainly change that.

Which live games would your average chess fan prefer to follow – those of Giri or Jobava? My guess would be Jobava. One thing this might well change is that someone not at all adverse to a draw like Giri would need to mix his game up more or risk sponsors opting to invite someone more ‘hungry’ like a Jobava.

In a way it is the system which keeps the same players playing each other which is at fault as well.

I’m watching right now, R8: Carlsen – Liren. Like those Anand games I mentioned, there is nothing at all ‘boring’ about the chess I am seeing. It is likely to end in a draw…but boy was this exciting! And at move 25 they both have about 1 1/2 hrs left. That is just...weird.
Well, we seem to agree that something should be done to encourage more enterprising, play-for-the-win chess. I agree that we have some interesting high-level games, maybe 1 out of 3 or so, but there would be a lot more interest if nearly every game were interesting. But your comments ignore the fact that chess is not even close to being a fair game, White is much better, just compare wins at top level. Anything that is to be done about draws needs to put the onus on WHITE to do something about them. It is quite unreasonable to punish Black in any way for a draw, he should be playing for a draw until White errs. Randomly selecting openings from some set of openings that are representative of decisive games at top level without favoring White too much appeals to me even more than 960, it avoids prep for individual games and cuts way down on draws. But if you want the players to have free choice from move one, then some special rules or conditions are needed to reduce draws, and they must be asymmetrical, somehow both encouraging and enabling White to play for the win. Some have proposed just banning drawish defenses like the Berlin, but that seems pretty artificial. I'd like to hear what the Elite players themselves suggest. I know Kramnik proposed the above opening lottery idea. Some liked the Armageddon playoff games in Norway, but this doesn't make the initial games more interesting.
Komodo rules!

Post Reply