Jeremy Good wrote on Mon, Jun 5, 2006 03:15 PM UTC:
Gary, I want to thank you for all of your comments and input.
You have inspired me to continue probing this issue, but with a little more depth. Your application to the Fischer - Spassky championship is pretty interesting as it decreases the proportion of the overall winning percentage significantly for Fischer, and as you pointed out to me privately, would have had Spassky tied at a certain juncture when he was behind in that match (under the point system they were utilizing at the time). None of this is meant to take away any of Fischer's achievement in that match, but only to apply a different dimension and allow for another perspective.
Above all, you have made it clear that my initial proposal has one, maybe two, highly significant omissions. So you have significantly helped me to clarify my proposal.
One of the omissions concerns the type of tournament for which this point system is appropriate. It would be, exactly as Gary says, 'grossly unfair' to implement this point system in a swiss system tournament. Such a system is, as Gary says, appropriate to a 'round robin' and might also make world championship matches more exciting if applied to them.
The second omission may concern ratings and how they are calculated and whether this point system would impact that. I don't know anything about how ratings are calculated and shall defer to Gary who is probably much more knowledgeable on the subject, having been a tournament director. Joe Joyce gives me hope though when he suggests that what I've 'invented' is what he refers to as 'power ratings.' (I think that was the phrase used). An entirely separate category of rating, appropriate to the point system.
Measuring the relative worth of players can be done in different ways. Please see some of chessmetrician Jeff Sonas's work on this subject. My point system is not sophisticated enough to address this issue. Perhaps Mr. Sonas will want to investigate this (maybe he already has).
The separate debate about the relative merits and demerits of White's first move is something that I'm also inspired to delve into more deeply, so thank you for the intriguing reference to Adorjan's book. Do you happen to know what Adorjan's career average as Black vs. White was? As you say, the statistics can't disprove the thesis. Another interesting approach comes from a comment Kasparov made in his second-to-last CD on the Najdorf (part of an ongoing series) in which he describes the Najdorf as an opening which gives Black a chance, not just to equalize, but to win, from the start. It might possibly be interesting to compare the winning percentages Fischer and Kasparov enjoyed with the Black pieces while playing the Najdorf to the winning percentages they enjoyed playing against the Najdorf as White. Again, that could be misleading since both were without peer in their knowledge of that opening.
Bent Larsen, one of my favorite players, an adherent of Nimzovich and enthusiast of the Alekhine Defense, noted in his most famous collection of games, that he happened to have among them more wins with Black than White. I mentioned Bent's use of the Alekhine Defense because of the interesting hypermodern strategy that defense implies, which encourages a focus on the weakening effects of white's first move ('every pawn move exposes a weakness' is a fairly standard chess maxim). Bent Larsen suffered some notorious failures in chess, but he also enjoyed some wild successes.
Like Derek Nalls, my starting point for this discussion is the weight of statistical analysis leading me to pose an ideal standard model. I don't think it's a resolved controversy. How can we prove that White in FIDE Chess is not, in fact, in a state of zugzwang? I would like to hear more about this. We can not rely solely on the rich heritage of accumulated human knowledge because computers have recently shown that human knowledge of FIDE chess is relatively insubstantial. Computer scientists demonstrated this when they beat a world champion (Bent Larsen, by the way, was the first grandmaster to lose to a computer!) and began exceeding humans in ratings. We must now turn to the statistical analyses of computer vs. computer games, but that might be misleading too, since we have every reason to believe that even more superior computers will surpass the ones we currently have. What is needed is a mathematically precise theory of chess that proves decisively the relative winning chances of both sides and that is not quite here yet. I'm open to hearing more about this from Derek, if he has more evidence for me.
The point system I propose is not necessarily geared towards resolving this interesting debate!
It is primarily intended to encourage professional and serious chessplayers to be less willing to compromise the highest standards of gameplay for the sake of winning half a point here and there. It is meant to extract excellence, but it is a very imprecise attempt at helping us describe the actual situation relative to the first and second players. I admit its extreme inadequacy and welcome a more vigorous, precise and knowledgeable analysis of that situation. Such an analysis may lead to a differing proposal for a point system, based on our current knowledge of the game, something perhaps more likely to bring out the most excellent gameplay possible.
The alternate point system's attempt to provide greater balance to an apparently imbalanced system is a secondary effect for me and one that requires a much deeper investigation.
My initial suggestion to Gary was to make draws equivalent to losses, as a way to encourage both sides to always seek the most winning lines. Upon reflection, I decided that it would probably be more fair to award a draw a point, so that it is still much more desirable to win, but draws won't be so heavily penalized. If a draw were worth zero points, it would still not be equivalent to a loss, for the sake of tournament (or match) strategy, since draws would involve neither player gaining a point and losses would involve one player gaining three or four points and the other player gaining zero points. On the chessvariants yahoogroup this morning, John Kipling Lewis commented that he prefers draws be made equivalent to losses. 'Draw is the new loss' in this proposal. Another interesting experiment. I fear that quality of game play could suffer if the draw incentive were virtually removed in this way, but I welcome and encourage other people to weigh in on this aspect of the alternate point system.
There is something in particular about the proposal I made here which seems to me not quite right and that is the fact that Black would gain twice as much from a draw as White, giving Black greater incentive to draw (as Gary pointed out). Black would still have even more incentive to win, and you could argue, perhaps successfully, that Black already has incentive to draw, simply because of his presumptive disadvantage in the opening. Most theorists appear to believe (Kasparov's comment in his Najdorf series notwithstanding) that Black should strive to equalize in the opening and only after Black achieves equality can Black seek to win.
Under this theory, White has to fail to maintain the initiative for Black to have a chance at winning. The onus is on White to prove that he can't win. This is the crux, it seems to me, of Reuben Fine's thesis in his Ideas Behind the Chess Openings, a book former world postal chess champion Hans Berliner commends in Berliner's The System (as opposed to Nimzovich's more famous book My System). If this theory turns out to be the correct approach, then rewarding Black more than White for drawing shouldn't affect the gameplay negatively. It would still be questionable whether a draw for Black should be worth twice as much as for White, when a win for Black is not worth twice as much. A smaller increment of reward for Black, such as .25 points, would complicate this point system, but might be more appropriate. Especially since a win for Black should be much harder for Black to pull off (than obtaining a draw, so that the ratio expressed by the proportions of White to Black draws as opposed to White to Black wins seem off-balanced).
I'll admit: I added the extra point for a Black draw impulsively because it seemed to me at the time that if we are going to say that a Black win is worth one point more than a White win, it must be just to say that a Black draw is worth one point more than a White draw. It may be that we should just say that all draws are worth one point regardless of color while retaining the extra point for a Black win.
It would be interesting to conduct a poll: In this alternate point system awarding three points for a White win and four points for a Black win, should draws be worth one point for either player? Or no points for either player? Or two points for black and one point for white? Or some other, more, incremental difference?
Again, I welcome feedback, and reiterate that this is intended not just as discussion, but as a serious proposal (one that I'm willing to amend as the facts and opinions come in).
Gary, I want to thank you for all of your comments and input.
You have inspired me to continue probing this issue, but with a little more depth. Your application to the Fischer - Spassky championship is pretty interesting as it decreases the proportion of the overall winning percentage significantly for Fischer, and as you pointed out to me privately, would have had Spassky tied at a certain juncture when he was behind in that match (under the point system they were utilizing at the time). None of this is meant to take away any of Fischer's achievement in that match, but only to apply a different dimension and allow for another perspective.
Above all, you have made it clear that my initial proposal has one, maybe two, highly significant omissions. So you have significantly helped me to clarify my proposal.
One of the omissions concerns the type of tournament for which this point system is appropriate. It would be, exactly as Gary says, 'grossly unfair' to implement this point system in a swiss system tournament. Such a system is, as Gary says, appropriate to a 'round robin' and might also make world championship matches more exciting if applied to them.
The second omission may concern ratings and how they are calculated and whether this point system would impact that. I don't know anything about how ratings are calculated and shall defer to Gary who is probably much more knowledgeable on the subject, having been a tournament director. Joe Joyce gives me hope though when he suggests that what I've 'invented' is what he refers to as 'power ratings.' (I think that was the phrase used). An entirely separate category of rating, appropriate to the point system.
Measuring the relative worth of players can be done in different ways. Please see some of chessmetrician Jeff Sonas's work on this subject. My point system is not sophisticated enough to address this issue. Perhaps Mr. Sonas will want to investigate this (maybe he already has).
The separate debate about the relative merits and demerits of White's first move is something that I'm also inspired to delve into more deeply, so thank you for the intriguing reference to Adorjan's book. Do you happen to know what Adorjan's career average as Black vs. White was? As you say, the statistics can't disprove the thesis. Another interesting approach comes from a comment Kasparov made in his second-to-last CD on the Najdorf (part of an ongoing series) in which he describes the Najdorf as an opening which gives Black a chance, not just to equalize, but to win, from the start. It might possibly be interesting to compare the winning percentages Fischer and Kasparov enjoyed with the Black pieces while playing the Najdorf to the winning percentages they enjoyed playing against the Najdorf as White. Again, that could be misleading since both were without peer in their knowledge of that opening.
Bent Larsen, one of my favorite players, an adherent of Nimzovich and enthusiast of the Alekhine Defense, noted in his most famous collection of games, that he happened to have among them more wins with Black than White. I mentioned Bent's use of the Alekhine Defense because of the interesting hypermodern strategy that defense implies, which encourages a focus on the weakening effects of white's first move ('every pawn move exposes a weakness' is a fairly standard chess maxim). Bent Larsen suffered some notorious failures in chess, but he also enjoyed some wild successes.
Like Derek Nalls, my starting point for this discussion is the weight of statistical analysis leading me to pose an ideal standard model. I don't think it's a resolved controversy. How can we prove that White in FIDE Chess is not, in fact, in a state of zugzwang? I would like to hear more about this. We can not rely solely on the rich heritage of accumulated human knowledge because computers have recently shown that human knowledge of FIDE chess is relatively insubstantial. Computer scientists demonstrated this when they beat a world champion (Bent Larsen, by the way, was the first grandmaster to lose to a computer!) and began exceeding humans in ratings. We must now turn to the statistical analyses of computer vs. computer games, but that might be misleading too, since we have every reason to believe that even more superior computers will surpass the ones we currently have. What is needed is a mathematically precise theory of chess that proves decisively the relative winning chances of both sides and that is not quite here yet. I'm open to hearing more about this from Derek, if he has more evidence for me.
The point system I propose is not necessarily geared towards resolving this interesting debate!
It is primarily intended to encourage professional and serious chessplayers to be less willing to compromise the highest standards of gameplay for the sake of winning half a point here and there. It is meant to extract excellence, but it is a very imprecise attempt at helping us describe the actual situation relative to the first and second players. I admit its extreme inadequacy and welcome a more vigorous, precise and knowledgeable analysis of that situation. Such an analysis may lead to a differing proposal for a point system, based on our current knowledge of the game, something perhaps more likely to bring out the most excellent gameplay possible.
The alternate point system's attempt to provide greater balance to an apparently imbalanced system is a secondary effect for me and one that requires a much deeper investigation.
My initial suggestion to Gary was to make draws equivalent to losses, as a way to encourage both sides to always seek the most winning lines. Upon reflection, I decided that it would probably be more fair to award a draw a point, so that it is still much more desirable to win, but draws won't be so heavily penalized. If a draw were worth zero points, it would still not be equivalent to a loss, for the sake of tournament (or match) strategy, since draws would involve neither player gaining a point and losses would involve one player gaining three or four points and the other player gaining zero points. On the chessvariants yahoogroup this morning, John Kipling Lewis commented that he prefers draws be made equivalent to losses. 'Draw is the new loss' in this proposal. Another interesting experiment. I fear that quality of game play could suffer if the draw incentive were virtually removed in this way, but I welcome and encourage other people to weigh in on this aspect of the alternate point system.
There is something in particular about the proposal I made here which seems to me not quite right and that is the fact that Black would gain twice as much from a draw as White, giving Black greater incentive to draw (as Gary pointed out). Black would still have even more incentive to win, and you could argue, perhaps successfully, that Black already has incentive to draw, simply because of his presumptive disadvantage in the opening. Most theorists appear to believe (Kasparov's comment in his Najdorf series notwithstanding) that Black should strive to equalize in the opening and only after Black achieves equality can Black seek to win.
Under this theory, White has to fail to maintain the initiative for Black to have a chance at winning. The onus is on White to prove that he can't win. This is the crux, it seems to me, of Reuben Fine's thesis in his Ideas Behind the Chess Openings, a book former world postal chess champion Hans Berliner commends in Berliner's The System (as opposed to Nimzovich's more famous book My System). If this theory turns out to be the correct approach, then rewarding Black more than White for drawing shouldn't affect the gameplay negatively. It would still be questionable whether a draw for Black should be worth twice as much as for White, when a win for Black is not worth twice as much. A smaller increment of reward for Black, such as .25 points, would complicate this point system, but might be more appropriate. Especially since a win for Black should be much harder for Black to pull off (than obtaining a draw, so that the ratio expressed by the proportions of White to Black draws as opposed to White to Black wins seem off-balanced).
I'll admit: I added the extra point for a Black draw impulsively because it seemed to me at the time that if we are going to say that a Black win is worth one point more than a White win, it must be just to say that a Black draw is worth one point more than a White draw. It may be that we should just say that all draws are worth one point regardless of color while retaining the extra point for a Black win.
It would be interesting to conduct a poll: In this alternate point system awarding three points for a White win and four points for a Black win, should draws be worth one point for either player? Or no points for either player? Or two points for black and one point for white? Or some other, more, incremental difference?
Again, I welcome feedback, and reiterate that this is intended not just as discussion, but as a serious proposal (one that I'm willing to amend as the facts and opinions come in).