Arimaa Forum (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
Arimaa >> General Discussion >> Article on chessbase about not resigning
(Message started by: omar on Dec 21st, 2008, 7:21pm)

Title: Article on chessbase about not resigning
Post by omar on Dec 21st, 2008, 7:21pm
I came across this interesting article on chessbase.com about not allowing resignations. Really, I had nothing to do with it :-)

http://chessbase.com/newsdetail.asp?newsid=5057

As most of you know I also discourage resigning (but I don't forbid it). When I watch live games or replay them I like to see games come to a natural conclusion. I appreciate players who don't abandon their army and put up a good fight to the very end. Of course in games without any element of change or hidden information it is tough to recover and one may feel very inclined to resign. And there are times when you just have to leave; so the resign button should always be there. But even in losing positions I like to think that maybe my opponent will blunder or the connection will go bad and I might still have a chance :-)

There is also an article with feedback on this proposal:

http://chessbase.com/newsdetail.asp?newsid=5060


Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 22nd, 2008, 7:13am
This proposal to forbid resignation in chess is justified almost entirely by reference to the spectators.  Yes, there is some feeble attempt to claim that playing every win out to checkmate would also be good for the players themselves, but that's a fairly weak argument: although I'm a mediocre player, I can force checkmate with a king and rook against bare king 100 times out of 100.  It's completely mechanical, and no grandmaster is going to get any benefit out of going through the motions.  It would be like telling Stephen King to recite the ABC's because it's good for him.  And as for the notion that the winning player might blunder, that is true of some positions, but not true of all positions, so a blanket rule will always cover some positions where the chance of error is zero for all practical purposes.  Again, think of the insult of asking a novelist to recite the alphabet to prove that he can do it right.

No, the only real argument behind the proposal is the enjoyment and benefit of the spectators.  The financial support of chess derives almost entirely from spectators, who are in turn almost entirely chess players to some degree.  People who know nothing whatsoever about chess can't enjoy watching chess; it's people like me who know a little who want to see what happens when the very best go at it.  We are the ones who are sometimes perplexed by resignations and would like to see a few more moves of "explanation" on the board.  Considering the fan base is a powerful argument.

Although spectator enjoyment is a strong justification for not resigning, a non-resignation rule is a horrible idea, because it is totally not enforceable.  As many commentators have pointed out, it is possible to lose on purpose in chess.  It is even more possible to lose on purpose in Arimaa, either by killing all of one's own rabbits or by pulling an opposing rabbit to goal.  Not only is suicidal behavior not preventable by any rules, it serves exactly the opposite purpose that a "no-resignation" rule would intend.  Suicide is even more confusing to the beginner, and makes the game even less enjoyable to a spectator than resignation would, because it is sometimes unclear whether or not the bad move was intentional or not.  Even if it is ultimately obvious that one side is losing on purpose, any suicidal moves raise the question of where the suicide began, e.g. was losing the elephant a blunder, followed by suicide, or was losing the elephant the first purposely bad move?

I think the policy of discouraging resignation is a good one.  Next time I feel like resigning, I will try to remember that I am not just playing for myself, but potentially also playing for spectators.  Sometimes I can bear a little personal boredom or humiliation for the common good.  But I believe it is counter-productive to write non-resignation into the tournament rules, as has been done for the World Championship and the Postal Mixer.  The result of the rules has been many suicides in hopeless positions.  We simply can't prevent people from giving up, and the attempt to prevent it results in absurdity.  I strongly recommend that any penalty for resignation be removed, and the doomed attempt at enforcing this behavior be replaced with nothing more than appeals to players to play on for the good of the sport.  I believe that most of the games that have been played out to the finish have been done so on the good will of the players, not because the players feared losing their entry fees, and that if the counter-productive rule were replaced by exhortation only, we would see just as many games fought to the death as we do now.  The only difference would be that suicidal moves that skirt the rules would be replaced with honest resignations.

Title: Re: Article on chessbase about not resigning
Post by Adanac on Dec 22nd, 2008, 8:33am
     
The decision of whether or not to resign should depend upon the skill level of the two players.  New players should always play to the checkmate since anything can happen and the winning player might accidentally walk into a stalemate or give away the queen.  On the other hand, asking a grandmaster to play to the end in a hopeless position can become ridiculous:

Here’s a legendary game between Friedrich Samisch and Aron Nimzovitsch.
http://www.chessgames.com/perl/chessgame?gid=1102400

The first time I saw this game, I was just learning chess and I didn’t understand the resignation at all.  It was very illuminating for me to play through all of White’s available moves and realize that his position was going to rapidly fall apart no matter what he did.  By resigning early, Samisch was able to walk away from the board with his dignity intact and (millions of?) chess students had a great game to study.  On the other hand, had he been forced to continue Samisch would have endured a long, drawn-out crushing defeat with no hope of building any active counter-play.  Many spectators would have observed the debacle and felt pity for Samisch without fully appreciating the strength of Nimzowitsch’s first 26 moves.

The guidelines I would prefer are these:
1.  If an Arimaa goal (or Chess checkmate) is inevitable within the next few moves, play out the game to its natural conclusion.  Here, I agree with Omar.  There’s no reason for chess masters to resign so many mate-in-one and mate-in-two positions.
2.  If the position you’re position is close to hopeless, but there is no forced loss, try to snatch victory from the jaws of defeat using all of your ingenuity and every available resource.
3.  Having exhausted all attempts at #2 and now facing an even worse material deficit and an utterly hopeless position, (and assuming you’re not in situation #1) simply resign rather than waste the time of both players.  I don’t think the Arimaa community is good enough for this situation to occur very often – more on that below.

     To reiterate a point that I made a few months ago, however, the Arimaa community has a very small player pool and we have a very limited knowledge of the game.  When Fritzlein publishes his book, we’ll be at the chess equivalent of the year 1512 when Damiano published his famous book.  Once we add hundreds of thousands of new players, publish dozens more books and learn lots more theory then we can catch up to where chess was in the 19th century (I doubt it will require 300 years though, things evolve much faster these days)!  I mentioned the 19th century-era of chess because that’s when early resignations seemed to have become much more commonplace and the best players no longer felt they had anything to gain by dragging out the games.  So to keep things in perspective:  we’re not even close to being seasoned Grandmasters with nothing more to learn from endgames.  If your opponent is annoying you by placing all the pieces in the back rows with no attempted counter-attack, it’s still a good learning experience for both sides.  Learning how to force a goal can be very instructive, and learning how to stonewall your opponent in a lopsided game will certainly help build the goal defence skills.

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 22nd, 2008, 12:22pm
I think we need to make a distinction about the appropriateness of resigning, and the appropriateness of having a rule that prevents resignation.  I guess I didn't disentangle the two clearly enough in my former post.

First, I basically agree with Adanac about when one should resign.  I agree that the weaker the players are, the more inclined they should be to fight to the end.  I agree that Arimaa theory is at a very nascent stage; we don't yet have the equivalent of chess Grandmasters, or even chess International Masters.  Because we collectively aren't very good, we should be less inclined to resign, especially in live games where the time pressure makes us play worse.  I have been swindled many times when I have had a huge material advantage, and every swindle was a proof that my opponent was correct not to resign.  Moreover, I usually play on even when I am way behind in material, as in this game from the 2008 World Championship (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=70761).  And even when loss is inevitable, if the end will be quick it is no great sacrifice to play the few moves until the finale.

That said, I also agree with Adanac that there are some times when resignation is appropriate.  In that context it does not make sense to forbid all resignations.  

Quite apart from resignation being fitting, however, neither Adanac nor Omar nor anyone on the chessbase forum has addressed the futility of having a rule that can be trivially circumvented by losing on purpose.  Even Mehrdad Pahlevanzadeh, who proposed the rule and is pushing it, has no better answer than, "When players are faced with actual game or tournament rules they adapt to them and follow them correctly."  He is only speculating about that for chess, since the no-resignation rule has never been tried there, but for Arimaa we don't have to speculate.  We know that people lose on purpose to skirt the no-resignation rule, because it has already happened.  Do I need to name names and post example games?  And when people lose on purpose, it is bad for the spectators, i.e. the very people whom the rule is supposed to benefit.

I'm sure anti-resignation folks can produce plenty of Arimaa games in which resignation was premature.  Indeed, my book includes an actual game position where the player who resigned had a forced goal!  My purpose in including that example in my book is precisely to discourage people from resigning.  But that changes nothing about the futility of trying to put fighting to the death into the rules of a tournament.  I would like to hear even one argument on the other side, not claiming that resignation is bad, but claiming that having a rule to prevent resignation will actually have its intended effect.  To my mind we already have ample proof that it doesn't work.

Title: Re: Article on chessbase about not resigning
Post by Adanac on Dec 22nd, 2008, 12:40pm
Good point, Fritzlein.  If players are just going to open a path for the opposing rabbits to score, we may as well encourage them to resign instead.  I admire a hard-fought battle to the very end, but we can't artificially force a great battle if one of the players has already given up psychologically.  I've never agreed with the "Never Resign" rule but it was nice of Omar to include a "Resign" button anyway despite his personal preference otherwise.  

And I definitely agree that too many players give up too quickly.  It's amazing that bold, daring moves can often turn an Arimaa game around, yet often players will simply "turtle" and accept their fate... ???

Title: Re: Article on chessbase about not resigning
Post by clauchau on Dec 23rd, 2008, 5:06am
It might be profitable to tournament results, players ranking, evaluation heuristics and research in game theory to take the length of games into account when they are played out. How about evaluating any game state as a number +-q^n between -1 and 1, where q is below and very close to 1 and n is the number of plies it takes to get the game really over - or averaged estimates of such values in case of uncertainties.

In real games, players who want to resign might agree on an estimate of the length n, although I haven't yet figured out any nice protocol in case they can't agree but don't want to fully play the game out either.

q=0.999 looks good to me for Arimaa. For example :

- winning as Gold by forfeit yields the standard +1 score as a result of the game.
- winning blazingly fast as Gold on move 5 (n=9) gives Gold about +0.99 points.
- winning as Gold on move 50 (n=99) on a well fought game gives Gold about +0.91 points.
- winning as Gold on move 500 (n=999) yields about +0.37 points.
- hypothetically winning as Gold on move 5000 (n=9999) practically amounts to drawing because it yields +0.00005 points.

For the MinMax algorithm, it simply means the value of a node is now q times the max value of its children.

Title: Re: Article on chessbase about not resigning
Post by omar on Dec 27th, 2008, 8:48am
Thanks for sharing your views on this topic guys.

For the World Championship and Postal Mixer I really would like to see all the games played out to a natural finish since these are some of our best games with more viewership than a casual game. But Karl's comments have made me think that penalizing a player for resigning is not the right way to achieve this.

I will get rid of the rules which penalize a player for resigning in these events. So a player is welcome to resign at anytime and there is no penalty. But I will introduce the following rules which I think will make the games interesting to the end for the players and spectators.

In the WC, if it can be shown that the winning player missed a forced win in 1, 2, or 3 moves at any point in the game then the losing player gets 50% of the prize money allocated for the winner of that game. The missed win solution must be posted as a game comment within 24 hours of the game end time. The solution can be posted by anyone, not just the losing player. If a player resigns they forfeit the ability to win the prize money if a solution is found.

In the Postal Mixer, if it can be shown that the winning player missed a forced win in 1, 2, or 3 moves at any point in the game then the losing player gets a portion of the winning players registration fee. The amount of this portion is the registration fee divided by the number of games the player signed up for. The missed win solution must be posted as a game comment within 24 hours of the game end time. The solution can be posted by anyone, not just the losing player. If a player resigns they forfeit the ability to win the prize money if a solution is found. To make this work though we may also need the rule that if a missed win solution is not found then the winning player is awarded a portion of the losing players registration fee equal to the registration fee divided by the number of games the losing player signed up for.

These rules will complicates the prize distribution work for me, but I think it is worth it because it makes the games more interesting for players and spectators while encouraging games to be played to a natural finish. Also it can generate more analysis and discussions of our best games and maybe even some puzzles that can be added to the puzzles page.

Title: Re: Article on chessbase about not resigning
Post by Forum Admin on Dec 27th, 2008, 8:56am
Moved this from the Off Topic section to General Discussion.

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 27th, 2008, 1:01pm
Once again, Omar, you prove how open-minded you are.  I really appreciate the way you listen.  That provides a big incentive for me to keep on blathering.  :)  That brings me to your latest proposal: it is intriguing but I am pretty sure the incentives do not line up the way you intend them to.

Suppose that in a Postal Mixer game, my opponent misses a forced goal in three, handing me the win.  (This almost happened to me against RonWeasley, although it was a goal in five he missed, allowing me a goal in three.)  If I go ahead and take the win, I get no reward, but if I pull his rabbit to goal, then I get part of his prize money since he was the winner and he missed a forced goal.  Assuming that the prize money is a powerful incentive, then you are using it to reward losing.

Indeed, someone who wins all of his games in the Postal Mixer can get at most all of his entry fee back, whereas someone who intentionally loses all of his Postal Mixer games will get at least his entry fee back, and probably more.  I might purposely play to set up remote goal threats for my opponent, hoping that he will win in four when he could have won in three; the more successful I am at losing like this, the more prize money I get, but even when my opponents always find the fastest goals I still get all my money back.

What you are suggesting is a way for the losing player to take money from the winning player.  Why did you create this way of hurting the winner?  Because you wanted some way to penalize resignation.  First you had to give the trailing player something to play for, so that you could take it away again if he resigns.  But the crowning irony is that you can't even take it away!  People can still suicide their own pieces instead of playing to the death.  In fact, the right suicide might even create a goal in three the other player would have trouble spotting.  There is nothing in your proposal that makes accelerating a loss any worse than losing slowly, so you are back to the same futility that you had before.

I agree with you on one point: it is worth some bookkeeping overhead to encourage fighting to the death.  Indeed, you already created that overhead for the Postal Mixer, i.e. the scoring system.  Every additional move that a player made instead of giving up earned them additional prize money.  If you are willing to have complexity of prize distribution, why not bring back the old scoring?

Also for the World Championship, there used to be an incentive for the leading player to win as quickly as possible and for the trailing player to survive as long as possible: points in the Spectator Contest.  But if you really like the new Spectator Contest format better, and you don't want to go back, then you could instead reward the loser with a penny per move instead of potentially taking half a dollar away from the winner for missing a goal in three that might take a puzzle solver fifteen minutes to find, and which was found only after the fact and only with the help of a computer.

In short, I think you have means to encourage fighting to the death, and we should explore those possibilities, but I don't think you have quite hit on the ideal means yet.

Title: Re: Article on chessbase about not resigning
Post by omar on Dec 27th, 2008, 5:02pm

on 12/27/08 at 13:01:17, Fritzlein wrote:
Suppose that in a Postal Mixer game, my opponent misses a forced goal in three, handing me the win.  (This almost happened to me against RonWeasley, although it was a goal in five he missed, allowing me a goal in three.)  If I go ahead and take the win, I get no reward, but if I pull his rabbit to goal, then I get part of his prize money since he was the winner and he missed a forced goal.  Assuming that the prize money is a powerful incentive, then you are using it to reward losing.


No I did think about this, you would get rewarded for winning. See my quote below.


Quote:
To make this work though we may also need the rule that if a missed win solution is not found then the winning player is awarded a portion of the losing players registration fee equal to the registration fee divided by the number of games the losing player signed up for.


But I guess I should also add that if both players are shown to have missed a win in 3 or less than the winning player gets the prize.


Quote:
I might purposely play to set up remote goal threats for my opponent, hoping that he will win in four when he could have won in three; the more successful I am at losing like this, the more prize money I get, but even when my opponents always find the fastest goals I still get all my money back.


But you would be running the risk that you opponent might see the win in three and get the prize. I think you missed that there is a prize for the winning player.

I am sure that with this rule in place the players and spectators will analyze the end games very closely and I think that will help us get better.


Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 27th, 2008, 5:41pm

on 12/27/08 at 17:02:28, omar wrote:
But you would be running the risk that you opponent might see the win in three and get the prize. I think you missed that there is a prize for the winning player.

The way you stated it, unless I misunderstood, there is no prize for the winning player finding a goal in three, so I am not running any risk within the game.  I can only get hurt if my opponent misses the goal during the game, then I fail to find the goal within 24 hours of game end, then the winning player finds it after the 24 hours are up.  That risk is miniscule, but in order to reduce it to zero, I can make sure before I submit my suiciding move that I have calculated out a forced win for him.  Then either he misses the forced win and I make money for sure, or he finds it, and I have managed to effectively resign, which is what I wanted to do in the first place.

But let's say I misunderstood your rule.  Let's say you have worked out a way that the winner and loser have equal chances to take money from each other after the game, so there is no incentive to be the loser so as to gain a post-game advantage.  Let's assume the post-game game is fair between two players of equal goal-finding ability.  Unfortunately even then the money-taking aspect will be tipped to the advantage of whoever is better at finding forced goals.  In that case I would be highly motivated to resign all games which I am way behind on material, because I am bad at finding goals, and I want to make sure the game never gets close enough that a forced goal in three exists.  A fair goal-finding game tacked on to the end of a regular game will not be in my favor, so I had better resign to avoid it.

The notion of having prizes for pointing out goals at the end of the game would indeed encourage post-game analysis, as you point out.  It would certainly make us better at finding goals if we practiced analyzing in this way.  But this has nothing to do with discouraging resignation, does it?

Title: Re: Article on chessbase about not resigning
Post by Adanac on Dec 30th, 2008, 11:07am

on 12/27/08 at 17:41:17, Fritzlein wrote:
The notion of having prizes for pointing out goals at the end of the game would indeed encourage post-game analysis, as you point out.  It would certainly make us better at finding goals if we practiced analyzing in this way.  But this has nothing to do with discouraging resignation, does it?


My first thought, when I read Omar's new proposal, was that we would have more incentive to hone our goal-finding skills.  But won't people just use Bomb to find forced goals, thus defeating the entire purpose, not to mention favouring players with commerical programs?

Like many players here I just play for fun, and it won't bother me at all if a new prize distribution system is introduced.  But if a losing player finds a goal-in-three where a goal-in-four was actually played by the winning player, there could be some considerable whining and complaining in the forum and game comments!  Finding a goal-in-four in a live game means that you'd better search long and hard for a goal-in-three instead before pressing "Send"  :-/

Title: Re: Article on chessbase about not resigning
Post by omar on Dec 30th, 2008, 7:27pm

on 12/27/08 at 17:41:17, Fritzlein wrote:
The way you stated it, unless I misunderstood, there is no prize for the winning player finding a goal in three, so I am not running any risk within the game.
Yes, there is no prize for the winning player for finding a goal in there, but there is a prize for the winning player for winning the game.


Quote:
I can only get hurt if my opponent misses the goal during the game, then I fail to find the goal within 24 hours of game end, then the winning player finds it after the 24 hours are up.
There is no need for the winning player to find the goal after the game. The winning player gets the prize for having won the game. Maybe I need to reword the rules more clearly.

The basic idea is this. The winning player gets a prize for winning the game. The losing player gets a chance to take away part or all of that prize by showing that the winning player missed a 1, 2, or 3 move forced win. No matter how bad the imbalance is the losing player has a reason to continue to the end because now he is not playing for a chance to win, but a chance for the winning player to miss a forced win in 3 or less. No matter how bad the imbalance is the winning player has no reason to say "just resign, you are wasting my time, you are insulting my intelligence", he has to focus on making sure he does not miss a win in 3 or less.

This kind of rule does not discourage resigning, it encourages not resigning. There is no penalty to discourage me from resigning. But, why should I resign just because I am losing the game, I will test the winning players ability to find forced wins to the end. I don't lose anything for doing that, in fact there is a chance I might gain something. I am encouraged to not resign. The encouragement of post-game analysis is a positive side effect.

Also this rule protects a losing player from a humiliating defeat from a winning player who has decided to keep picking away at pieces instead of finishing of the game.

Also if the losing player commits suicide instead of directly resigning it will be because the losing player knows the winning player missed a win in 3 or less and the losing player will post it after quickly finishing off the game. So suicides will be explained otherwise they won't occur.

Consider the recent Continuous Tournament game between arimaa_master and naveed. Though naveed won the game, he missed a goal in 1 on his way to winning. Had this been a WC finals games naveed would have won $5 if he did not slip up while finishing off the game, but since he did half of that prize would go to arimaa_master. Had this been a Postal Mixer game $5 of naveeds registration fee would have been won by arimaa_master instead of the other way around (assuming both players signed up for 4 games and paid $20 registration fees).

Title: Re: Article on chessbase about not resigning
Post by omar on Dec 31st, 2008, 4:27am

on 12/30/08 at 11:07:49, Adanac wrote:
My first thought, when I read Omar's new proposal, was that we would have more incentive to hone our goal-finding skills.  But won't people just use Bomb to find forced goals, thus defeating the entire purpose, not to mention favouring players with commerical programs?
Yes, you can use Bomb, you can consult a friend, or even the whole Arimaa community. I think what we will see is everyone looking at these games to see who will be the first to find a missed win. But during the game you won't have any help, so the players do have to hone their own goal finding skills.


Quote:
Like many players here I just play for fun, and it won't bother me at all if a new prize distribution system is introduced.  But if a losing player finds a goal-in-three where a goal-in-four was actually played by the winning player, there could be some considerable whining and complaining in the forum and game comments!  Finding a goal-in-four in a live game means that you'd better search long and hard for a goal-in-three instead before pressing "Send"  :-/
I don't think there would be any complaints. If a position had a line that leads to a forced win in 3 and a different line that leads to a forced win in 4 then it is clear the winning player missed a win in 3; this is precisely the kind of chance the losing player was looking for and the losing player should be rewarded for continuing to play and bringing the game to such a position. Depending on the type of game; postal, regular, blitz, etc. the "win in 3 or less" could be changed to some other value or the amount of reward the losing player takes away from the winning player could be varied.

Title: Re: Article on chessbase about not resigning
Post by RonWeasley on Dec 31st, 2008, 6:03am
I agree with Adanac here.  In theory I don't mind giving rewards to the losing player for continuing to play well.  However, I would like to avoid penalizing the winning player for playing "carefully".  Specifically I worry about seeing a complex possibility of a win in three which turns out to be a blunder.  I understand that this makes things exciting, but it changes the nature of the game.

Imagine Fritzlein avoiding strategic goal pressure because he wants to avoid potential goals in three.  He could turn his games into long piece hunts if he wanted to and this would have the opposite effect of what this rule intends.

For my part, I would tend to ignore this and play my usual plodding style.  If I'm winning, I wouldn't want to risk losing by going after an unclear tactical win in 3.  If I'm losing, I am content with dragging the game on, even without a prize, just for the sport of it.  In the Owl Mixer, if I lose, I like to break the 60 move mark, and that's enough motivation for me without taking something away from the winner.

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 31st, 2008, 6:17am

on 12/30/08 at 19:27:30, omar wrote:
Yes, there is no prize for the winning player for finding a goal in there, but there is a prize for the winning player for winning the game.

OK, that's what I didn't understand.  In order to be clear about the Postal Mixer, you would need to explain what amount of money, if any, a player gets for simply finishing his games as opposed to timing out.  If I see where you are headed, there is no reward whatsoever for fighting to the death per se, and instead there will be a reward if and only if a forced goal is missed by the winning player.  That proposal is at least theoretically aligned with what you are trying to achieve.

I do think that Adanac is right that everyone will just feed endgame positions to Bomb or another bot engine.  Indeed, without a bot, it might not be clear whether a goal in three has been missed.  Otherwise, what if somebody posts after the game that move XXXX is a goal in three that the winning player missed, and the winning player disputes the claim?

Also, what if the winning player has a goal in three, but overlooks it.  Suppose he instead plays a move that he thinks is goal in two, but is really goal in four.  But then the losing player overlooks the one defense that prolongs it to goal in four, and loses in two instead, just as the winning player had foreseen.   After the game it is discovered that the losing player played sub-optimally.  Does the winning player, despite having won in two moves, still get penalized for having missed a forced win in three?

If the rules are clarified a bit better, I am not averse to trying out your proposal, although it will probably have unintended consequences.

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Dec 31st, 2008, 6:37am

on 12/31/08 at 06:03:08, RonWeasley wrote:
I agree with Adanac here.  In theory I don't mind giving rewards to the losing player for continuing to play well.  However, I would like to avoid penalizing the winning player for playing "carefully".  Specifically I worry about seeing a complex possibility of a win in three which turns out to be a blunder.

Yes, I have definitely been in situations where I have had a lead and was starting a goal attack, when my opponent counter-attacks instead of defending.  I might think that my attack is going to be faster, but sometimes I am not sure, so I look for a good defensive move.  If I see a move that totally shuts down the counter-attack, why not do that first and then finish winning later?

Under Omar's proposal, I can imagine being on the verge of victory when the other player suicidally advances a rabbit.  I could just capture the rabbit and continue with my attack next turn.  But wait, taking the free rabbit might cost me half my prize money!  Instead of being allowed to play a totally obvious defense, I must determine whether my goal attack is faster, and immediately force a goal if I have one.  I must play offense instead of defense.

But offense is only a good idea if I am going to win the race.  Perhaps I will misjudge the race and feel forced to play for an immediate win when I don't truly have a forced goal, and perhaps this will cost me a win.  Because I know I will be compared to a standard of perfect play after the game, I might make a huge blunder within the game by trying to be perfect.  In other words, the perfect is the enemy of the good.

If Omar implements his proposal, I will certainly look for forced goals more than I do now.  I will try to improve an aspect of my game that needs improving in any case.  But I reserve the right to be upset when (A) my slow, sure win is penalized because there happened to be a risky win that was faster, or (B) I feel pressured to play for an immediate, risky win, and lose the game because of it.

Title: Re: Article on chessbase about not resigning
Post by Adanac on Dec 31st, 2008, 8:41am

on 12/31/08 at 06:37:06, Fritzlein wrote:
But offense is only a good idea if I am going to win the race.  Perhaps I will misjudge the race and feel forced to play for an immediate win when I don't truly have a forced goal, and perhaps this will cost me a win.  Because I know I will be compared to a standard of perfect play after the game, I might make a huge blunder within the game by trying to be perfect.  In other words, the perfect is the enemy of the good.

If Omar implements his proposal, I will certainly look for forced goals more than I do now.  I will try to improve an aspect of my game that needs improving in any case.  But I reserve the right to be upset when (A) my slow, sure win is penalized because there happened to be a risky win that was faster, or (B) I feel pressured to play for an immediate, risky win, and lose the game because of it.


Nicely said - I was thinking along the same lines last night but you wrote it more eloquently.  And I shouldn't have used the word "whining" because RonWeasley and Fritzlein have come up with scenarios where the winning player would be justifiably miffed.

Just for the fun of it, I'll throw out another idea:  Every time a player loses a game by resigning or any blatant attempt to throw the game (such as piece suicide to clear a path for the enemy rabbit) results in a $0.25 penalty for the next World Championship entry fee.  Every loss played to a natural conclusion credits the player with a $0.25 discount toward the next World Championship.

Objectives:
1.  A loss is worth less than a win but...
2.  Players that fight to the end receive a discount next year
3.  Quitters pay a penalty next year

It's extra administration work, but not too much.

Title: Re: Article on chessbase about not resigning
Post by Janzert on Dec 31st, 2008, 10:23am
If I'm understanding the proposal correctly I think it's a bad idea for the WC, but a complete disaster for the postal mixer.

Currently the way to maximize your monetary reward in the mixer is to simply play every game without timing out or resigning. Winning or losing doesn't matter. With this proposal you need to lose every game in such a way that the opponent misses a forced goal for a move. Winning a game at best gives you an equal payout to losing and may lower your payout. I think the effect is also magnified since by winning every game at best you can only break even.

Janzert

Title: Re: Article on chessbase about not resigning
Post by omar on Jan 1st, 2009, 11:11pm
This is starting to feel a little bit like a One vs TheMob game :-)

First let me just say that I am not trying to hold up my side out of stubbornness. Many of you know me well enough to know that I carefully consider other views and can be persuaded with a sound argument.

One of the things that attracted me to this proposal was that it would generate greater analysis of the end games. Other methods could be used to encourage not resigning, but they would not have the positive side effect of also encouraging better end game play during the game and more analysis after the game. Others have not been as excited about this as I had hoped. Though this is not what we initially set out to solve, I think this side effect is really great and is something worth getting excited about. I am excited enough about it that I am willing to put in more effort with the distribution calculations.

So the thing that this proposal does solve are:

It encourages not resigning as opposed to discouraging resigning. I think that in itself is a step in the right direction.

It avoids the problem with the losing player purposely trying to lose quickly to finish the game; the suicide problem. Most solutions that encourage not resigning avoid this problem and solutions that try to discourage resigning will encounter this problem.

It avoids the problem with the winning player purposely taking longer to finish the game; the torture problem. It should be noted that most solutions whether they encourage not resigning or discourage resigning will not fix this problem. It is an added benefit that this proposal does.

It also provides some free parameters which can be tweaked to adjust the solution based on the type of game and how much encouragement we want to provide for not resigning. The free parameters are: level of perfect play, time limit to show a solution, and the prize for finding a solution.

The solution is pretty simple. The winning player would receive a prize for winning the game. If the losing player did not resign he is given a certain amount of time after the game ends to show that the winning player could have finished the game sooner by not missing a forced win line with less than a certain number of moves. The solution can be posted by anyone and not just the losing player. Computers can be used in trying to find the solution. If a solution is found and posted as comments to the game within the allotted time, the losing player receives a portion of the winning players prize. If it is also shown within the allotted time that the losing player missed a forced win line with less than a certain number of moves, the losing player does not get any prize.

In applying this to the WC lets say that the forced win line must be 3 moves or less; the allotted time to find the forced win line is 24 hours after the game end time; and the prize for finding a solution is 50% of the winning players prize for this game. The prize for the winning player is preset for each game.

In applying this to the Postal Mixer lets say that the forced win line must be 3 moves or less; the allotted time to find the forced win line is 24 hours after the game end time; and the prize for finding a solution is 100% of winning players prize for this game. The prize for the winning player for this game is the losing players registration fee divided by the number of games the losing player signed up for.

In the WC I expect the winning players may very well miss forced wins in 3 (or less) and it would be foolish for a player in a winning position to focus too much on this and blunder the game. Winning the game is still the primary goal and winning it without missing a forced line is a secondary goal if you are in a dominating position.

In the Postal Mixer players have much more time and there should be less missed forced wins in 3 (or less). But if it happens it gives the losing player a chance to recover the prize which the winning player won from them.



Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Jan 2nd, 2009, 9:17am

on 01/01/09 at 23:11:07, omar wrote:
This is starting to feel a little bit like a One vs TheMob game :-)

First let me just say that I am not trying to hold up my side out of stubbornness. Many of you know me well enough to know that I carefully consider other views and can be persuaded with a sound argument.

I certainly know you well enough to know that you listen to other people and take their opinions into account, and I appreciate this aspect of your personality.  Forgive me for saying that you are also stubborn.  I happen to think it is a rare virtue to be stubborn when you are right and everyone else is wrong, so I believe it is wise of you not to be convinced merely by the number of people who disagree with you, and to focus on the substance of their arguments instead.


Quote:
One of the things that attracted me to this proposal was that it would generate greater analysis of the end games.

I grant you this point unequivocally.  If you change the prize structure the way you are suggesting, there will be more analysis of endgames.  People will get better at goal attack and goal defense.  This is worth promoting.


Quote:
It avoids the problem with the losing player purposely trying to lose quickly to finish the game; the suicide problem.

I think you are dead wrong on this point.  Your proposed prize structure, although it will discourage resignation, does very little to discourage intentionally suicidal moves, unless a forced goal is already in the offing.

Consider the last resignation I faced (actually the last game I played), this game against aaaa. (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=92570)  Not a single piece had been captured, but aaaa resigned in light of the fact that he couldn't stop me from capturing a horse on my next move, with a dominating position to boot.  This seems like a prime example of a situation in which you would want the losing player to struggle on, but also a prime example of the type of position which fans of resignation would feel is not worth playing out.

Suppose that that game had been a World Championship game with the rule in place that the winner gets 100% of the prize money, but the loser can get half of that if he doesn't resign and the winner overlooks a goal in three.  Suppose aaaa had been motivated by the prize money to change his behavior.  What would his most rational move have been?  I submit that instead of resigning, his most logical choice would have been 17s ee3e ef3x md3e hc3x me3e mf3x, suiciding his elephant, camel, and horse.  His heavy pieces are not doing anything to increase his chances of winning half the prize money: on the contrary they may delay the secondary-game-within-the-game for another twenty moves or so.  By getting his big boys off the board, aaaa could remove perhaps an hour of pointless struggle, and get much more directly to the part of the game where he has a chance to win the consolation prize.

You have already seen with the gameroom ratings that, no matter how you set up the reward system, at least some people will play to maximize the stated reward, regardless of how pointless that behavior would otherwise be.  One could not otherwise explain ArifSyed winning hundreds of games against Bomb2005P1 by formula, but stopping and moving on to other bots as soon as he no longer gains one rating point per game.

I think you need to look hard at exactly what you are proposing to reward, because that is exactly the behavior you will get, at least some of the time.  You are not proposing to reward people avoiding intentionally bad moves.  You are proposing to reward them for a goal-finding error by their opponent, and by extension for getting into positions where their opponents might make goal-finding errors.  This is not the same as rewarding a player for fighting to the death, or for trying to snatch victory from the jaws of defeat.  It's not even the same as rewarding a player for forcing his opponent work as hard as possible to convert a won position.


Quote:
It avoids the problem with the winning player purposely taking longer to finish the game; the torture problem.

You might well be wrong about this too.  I don't think the "torture problem" is a very big problem at present, but such as it is, you might make it worse.  At the moment when I have a decisive advantage, I am somewhat motivated to make a goal attack in hopes of finishing up the game sooner, and only deterred from attacking if there is some risk that it will lose me the game.  After you change the prize structure, though, it will be dangerous for me to launch a goal attack for fear of missing the fastest ending in a messy situation.  Perhaps it will turn out that it is safer for me to get the situation under control in some other way, trying to simplify before advancing a rabbit, because in a simpler situation I might be less likely to overlook the fastest winning route.  We won't know until we try.

I can imagine a situation where the losing player is unmotivated by the prize money, and is fighting to the death merely because he feels that is the way the game should be played.  Meanwhile the winning player is motivated by nothing but the prize money, and starts playing bizarre moves which are not directed at winning as fast as possible, but rather are directed at making absolutely sure he doesn't miss a win in three.  He might, for example, suicide most of his rabbits so as to have fewer chances of overlooking a forced goal on the side away from the action.  This behavior by the winning player seems like a poor way to treat someone who is a good sport in defeat and who only wanted to play out a game to its natural conclusion, but that unsporting behavior might be what you are proposing to reward.

The original goals of preventing resignation, as laid out in the linked article, were to increase spectator enjoyment and education.  If you change the rules so that there are two prizes (one for winning, and one for getting the opponent to overlook a forced goal), then the players will have two objectives, which may well increase spectator bafflement.  For some moves spectators will have to guess whether the player who made that move was trying to win, or trying to avoid missing a goal in three, or trying to induce the opponent to miss a goal in three.

To summarize: I do expect your prize-distribution proposal will do a good job of achieving some other objectives (namely increasing understanding of goal attack and defense, as well as encouraging post-game analysis), but I don't think it does anything toward the original objectives.  Furthermore, insofar as people play for the prizes, I think it will probably have unpleasant side effects.

Title: Re: Article on chessbase about not resigning
Post by Adanac on Jan 2nd, 2009, 9:18am
Omar, out of deference to the game’s inventor, I’m sure we can all agree to implement a system that encourages fighting to the end.  But can we maybe give a small “grace period” to the winning player as a modification to your original proposal?  My biggest concerns are these two scenarios, basically the same as the ones described earlier by Fritzlein and RonWeasley:

a.   Gold has a clear path to victory but Silver has 1 mobile rabbit that *might* score first.  Gold prudently uses four steps to halt Silver and then discovers after the game that half his prize money is gone because a “goal-in-three” has been missed due to his “unnecessary” cautious play.

b.   Gold might be able to force goal in 3 in (suppose there’s too little time to calculate precisely) a game with perfect material equality, but he can de facto guarantee victory by capturing the Silver Camel in the south half of the board and then comfortably score the goal later.

Alternate proposal: Once the winning player has a proven forced goal (or immobilization, etc.) in “X” moves then we give him a “2X” move grace period.  So, for example, if Silver sees a forced goal-in-four for Gold after Silver’s 39th move (X=4, 2X=8 ) then Gold can win by move 43 with optimal play and must win by move 47 in order to keep the full prize money.

Advantages:

1.   Similar to Omar’s proposal, we still minimize the losing player’s suffering, though it may extend for a few additional moves.
2.   The winning player has the luxury of making a cautious move or two without the pressure of losing half the prize money (or worse yet, losing the game after chasing an imaginary - but perceived necessary - goal-in-three move).
3.   This “X” factor can be any large number so Silver can search for any goal-in-four, five, six, etc.

Disadvantages:
1.   This doesn’t encourage perfect play in the endgame, as Omar’s does, just “good enough” to win.
2.   Still puts live-game pressure on careful players, but a bit less.
3.   Less incentive for the losing player, since spotting a forced goal and then surviving for “2X+1” moves is difficult to do.
4.   Still gives a large advantage to players that own a copy of Bomb (or any other commercial software).

It’s easy to modify to something like “2X-1” to eliminate the grace period for a goal-in-one, if we want to do that (seems reasonable).  

There's a general feeling on this board that the penalty is too harsh under the original system, and perfect play too difficult to accomplish in live games.  I'd prefer a more "forgiving" prize distribution system.

Title: Re: Article on chessbase about not resigning
Post by omar on Jan 6th, 2009, 2:07am

on 01/02/09 at 09:17:30, Fritzlein wrote:
Suppose aaaa had been motivated by the prize money to change his behavior.  What would his most rational move have been?  I submit that instead of resigning, his most logical choice would have been 17s ee3e ef3x md3e hc3x me3e mf3x, suiciding his elephant, camel, and horse.  His heavy pieces are not doing anything to increase his chances of winning half the prize money: on the contrary they may delay the secondary-game-within-the-game for another twenty moves or so.  By getting his big boys off the board, aaaa could remove perhaps an hour of pointless struggle, and get much more directly to the part of the game where he has a chance to win the consolation prize.

It would certainly speed up the game, but wouldn't that also simplify the game and put you in total control. Then it would be much easier for you to make sure that you are not missing a win in 3 and only advance your rabbits when you have a clear path. It might be better for aaaa to keep the strong pieces and try to complicate the position as much as possible; perhaps by using the strong pieces to pull your rabbits closer to goal and hope that you miss a win in 3 which he can hunt for after the game.

But even if he does choose to suicide his pieces, it is not to circumvent a "do not resign" rule; it is just his approach to winning the consolation prize. The game would still be explainable in that context and it would still be interesting to see if his approach works.


Quote:
After you change the prize structure, though, it will be dangerous for me to launch a goal attack for fear of missing the fastest ending in a messy situation.  Perhaps it will turn out that it is safer for me to get the situation under control in some other way, trying to simplify before advancing a rabbit, because in a simpler situation I might be less likely to overlook the fastest winning route.
What you are describing is a situation where you are playing it safe. I think that is perfectly fine. What I had in mind as torture is best described by this quote from the chessbase feedback article:


Quote:
We could reach instances when players might jump and shout to the Arbiter: Oh my god! He doesn't want to mate me! He wants to capture all my pieces and pawns... please put an end to this game... please!


If we assume that the winning player is motivated by not losing some of the prize then we shouldn't see this kind of behavior.


Quote:
If you change the rules so that there are two prizes (one for winning, and one for getting the opponent to overlook a forced goal), then the players will have two objectives, which may well increase spectator bafflement. .....
For some moves spectators will have to guess whether the player who made that move was trying to win, or trying to avoid missing a goal in three, or trying to induce the opponent to miss a goal in three.


I think your point here is that this rule we introduced to encourage not resigning will slightly contaminate the game itself such that the moves are not explainable purely on the basis of best Arimaa play. We are in effect not playing the original game, but a slightly modified version. No doubt odd behavior could be explained in the context of all the rules present, but you want all moves to be explainable in the context of only the original Arimaa rules. OK, now that is a sound argument and reason enough to be careful of using this rule.

In that case if we want to encourage not resigning and also want to make sure that players moves are explainable purely on the basis of best Arimaa play then I think the only alternative is to reward the losing player for extending the game as long as possible while taking away some of the reward of the winning player as the game gets longer. Though this may not be as spectacular to watch or thought provoking for the audience it will be explainable purely on the basis of best Arimaa play.

Though this continues to pressure the winning player to win as quickly as possible, we will lose the post game discussion and analysis of missed forced wins that would have occurred with the other rule. This is regrettable since, these will be some of our most well played and highly viewed games.

Perhaps we can make a spectator contest out of hunting for missed forced wins.


Title: Re: Article on chessbase about not resigning
Post by 99of9 on Jan 6th, 2009, 3:03am
I would be willing to play with this proposal as long as the percentage of the pot that the loser could win was (significantly?) less than 50% (I suggest more like 10%).

However, I think Fritz is right that it will induce some strange behaviour in some situations.

I still prefer the reward sets that incentivize losing slowly.

If you care so much about endgame analysis (a separate issue to resignation), I think it would be better to just add a prize for people who discover shorter forced wins.  We cannot expect anyone to always find goals in 3 under WC time controls.  I agree with others that most of the goal finding will be done by computers anyway...

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Jan 6th, 2009, 7:29am

on 01/06/09 at 03:03:24, 99of9 wrote:
I would be willing to play with this proposal as long as the percentage of the pot that the loser could win was (significantly?) less than 50% (I suggest more like 10%).

This reminds me to clarify that, although I have argued vigorously against giving away half the winner's prize for missing a forced goal in three, I would be willing to play under that rule.  I expect it to subtract from my enjoyment rather than adding, and furthermore I don't think Omar will be happy with the changes in behavior he induces, but who knows?  We can always try it and see what happens.  Worst case, I can just ignore the prizes and be happy I have a place to play Arimaa for free.

Title: Re: Article on chessbase about not resigning
Post by The_Jeh on Jan 6th, 2009, 9:20am
As far as avoiding resignations go, I have always thought that the social pressure we have created in our Arimaa culture was enough to discourage them, without the need for real pressure. Resigning, it seems to me, has always been most concerned with preserving the dignity of the losing player and honoring the winner. It is akin to saying, "I may have been outplayed, but I am talented enough to recognize that I have been outplayed, and I also recognize that the talent of my opponent is sufficient to defeat me from here." So to not resign in a hopeless situation in chess becomes an insult to both players. Chess has always had a chivalric flavor about it, and so resignation fits well with it aesthetically. But since in Arimaa we have created a different culture, where the action that in chess is honorable is now the opposite, I think people will be inclined to play on.

With regards to imposing a penalty on a player for missing a forced goal, I am definitely opposed. Any tournament changes should seek to increase the players' liberty to play with the styles and strategies they wish, not put constraints on their decisions. Isn't the desire to win itself the best motivation for a player to play the best moves - to play the opening, middle, and end games better? Then it is only the end result that should be rewarded. On the other hand, if other players take this change to heart, and I ignore it, I expect my winning percentage to increase, so maybe it's not all that bad....

Title: Re: Article on chessbase about not resigning
Post by aaaa on Jan 12th, 2009, 4:31pm
I would vehemently oppose anything that could to any degree whatsoever compromise a player's goal of maximizing his or her chances of a win. In fact, I'm even somewhat troubled by the fact that one gets additional prize points in the CT for maximizing the length of a lost game, since, sportsmanship aside, that could make someone choose a guaranteed slow loss over a shot in the dark.
BTW, this is yet another advantage of having no draws; it eliminates any meta-issues that could be considered during the game (e.g. "a draw is sufficient for me to win the tournament").

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Jan 14th, 2009, 6:51am

on 01/12/09 at 16:31:29, aaaa wrote:
In fact, I'm even somewhat troubled by the fact that one gets additional prize points in the CT for maximizing the length of a lost game, since, sportsmanship aside, that could make someone choose a guaranteed slow loss over a shot in the dark.

Good point, aaaa.  If there is a slim shot at a counter-attacking swindle, I don't want people to pass it up because there are more sure points in a slow, defensive loss.  People should be encouraged to maximize their winning chances.  I'll have to reconsider the scoring rules before I restart the Continuous Tournament.

Title: Re: Article on chessbase about not resigning
Post by clauchau on Jan 15th, 2009, 5:11am

on 01/14/09 at 06:51:59, Fritzlein wrote:
If there is a slim shot at a counter-attacking swindle, I don't want people to pass it up because there are more sure points in a slow, defensive loss.  People should be encouraged to maximize their winning chances.

Getting additional prize points in the CT for maximizing the length of a lost game is not so wrong. It's only a matter of setting the additional points small enough.

Suppose I hesitate between a counter-attacking swindle and a slow, defensive loss. If I somehow estimate my gains as

- a probable 1/100 for the counter-attacking swindle
- or a sure 1/1000 for the slow, defensive loss

then I'll choose the swindle.


Title: Re: Article on chessbase about not resigning
Post by omar on Feb 12th, 2009, 9:14pm
First chess match using the play-to-mate match rule.

http://chessbase.com/newsdetail.asp?newsid=5195

The match even had a prediction contest with prizes for the spectators :-)

Title: Re: Article on chessbase about not resigning
Post by rozencrantz on Mar 24th, 2009, 1:00pm
I am a fairly new player, and very new to the community.

I'm a little confused as to how this will affect the Postal Mixer.

Right now, it seems like as long as I play all of my games to conclusion, the $20 fee is a sincerity deposit, and I can expect to get it back.

Under the new rule, because I am still not a very good player, and especially weak tactically, it looks like I could easily be penalized for winning. The odds of that are slim, but I am reluctant to enter if there is a chance that I could lose part of my deposit by winning.

Am I misunderstanding this somehow?

Title: Re: Article on chessbase about not resigning
Post by Fritzlein on Mar 24th, 2009, 1:54pm
I believe that, in face of strong opposition from the playing community, Omar is not implementing his proposed distribution of prize money.  The Postal Mixer rules appear unchanged from last year.  Is this correct, Omar?

Title: Re: Article on chessbase about not resigning
Post by omar on Mar 29th, 2009, 9:43am
Yes, there is no change. It is using the same rules as last year.



Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.