|
< Earlier Kibitzing · PAGE 19 OF 20 ·
Later Kibitzing> |
| Feb-13-12 | | MORPHYEUS: <nimh>. May we request a comparison of Carlsen with Fischer and Capablanca? (accuracy, etc.) Thank you. |
 |
Feb-13-12
 | | nimh: <Thanks for the reply. Is there hope of seeing some studies with a larger sample size?> Maybe in the farther future, currently I'm focusing on developing a reliable methodology. <May we request a comparison of Carlsen with Fischer and Capablanca? (accuracy, etc.)> Page 8
http://web.zone.ee/chessanalysis/su...
Btw, a while ago you let know that you'd include the difficulty of positions in your research. How are you going to measure the factor? |
|
| Feb-13-12 | | MORPHYEUS: I was hoping i can find Carlsen in your "average error by complexity" graph side by side with Fischer, Capablanca, Deep Blue, etc. <Btw, a while ago you let know that you'd include the difficulty of positions in your research. How are you going to measure the factor?> I heard you use stockfish to measure the complexity of the position, i never saw that before. I can probably find the answer if i study your work in closer detail. However, i'd appreciate it if you can find the time to explain it to me here. I'd probably modify my methods after yours, but i don't expect it will be 100%. |
|
Feb-14-12
| | nimh: Oops, I messed up!
Carlsen's stats were in the other paper, on page 10 http://web.zone.ee/chessanalysis/st...I won't include other players, except Kasparov in Linares 1999, Karpov in Linares 1994 and Fischer vs Taimanov & Larsen there.
In the latest study I focus on time periods instead of individuals. It's easier this way since the rate of responsibility is lower. It doesn't matter whether the results you get are
either
1860 0,150
1880 0,145
1900 0,140
1920 0,135
or 1860 0,144
1890 0,151
1920 0,146
1950 0,129
What's important to notice is the trend. In both cases the regression lines are virtually identical. In my latest study, I've used 4 of measures of difficulty of position: complexity, difference between two best moves, material, and evaluation of the best move. The most important of them is the difference between 2 best moves.
Calculating complexity is time-consuming and not easy. The amount of material on the board is the weakest indicator. And as for the evaluation, it actually isn't a difficulty indicator per se, but rather affects all engine-related measurements. I think you should take only the <difference between 2 best moves> and the <evaluation of the best moves> from my study, and try to find other indicators yourself. Here are some ideas:
1) material imbalance, the amount of material that differs for both sides. Knight vs bishop is the imbalance of 3.25 points; two rooks vs one rook and 5 pawns is 5 points. 2) evaluation vs material imbalance, shows how large the engine evaluation is compared to material for one side. If you're down a pawn, yet the eval is around 0.5 in your favour, then the imbalance is 1.5; but if you're up a whole rook with equal eval, the imbalance is 5. 3) the number of interactions between hostile pieces. An interaction occurs when your opponent's piece is in the range of attack of one of your pieces. 4) time spent to reach a certain depth. In the start position on my comp, R4 gets depth 18 in 23 secs, but in the following position she requires 4 seconds only. 
click for larger view5) the symmetry of positions. The symmetry of start position is equal to the amount of material for one side, in the following position, the only pieces placed in symmetry are one rook and the e&f-pawns, so the symmetry value is 7. 
click for larger view |
|
| Feb-17-12 | | MORPHYEUS: Thanks nimh. |
|
Feb-23-12
| | nimh: I have an interesting suggestion to you, Morphyeus. Forget about the analysis work you're currently conducting. Researches attempting to prove that a player rated X at time control Y performs as well as a players rated Z at time control N have no scientifical value at all. Instead, how about conducting a comparative analysis on various methods of measurements of difficulty of positions? You would do something that has never been attempted before and you might even discover useful information for future researches. How good relatively a certain indicator of difficulty of positions is determined in two ways. 1) its influence on the accuracy of play. This is measured by plotting a difficulty indicator against average error on a graph and looking at the angle of the linear regression line. The steeper the line, the more influence it has on the level of play. 2) its independence on other indicators of difficulty. Correlation with other indicators should be minimal in order to cover as many aspects of difficulty as possible and avoid wasting time and effort. So, what do you say? You interested? |
|
May-04-12
 | | Boomie: <nimh: <Boomie>
Second question - how big percentage of population did take chess seriously at that time? Isn't it at least as rare as the occurrence excellently talented humans?
How probable, then, is that the maxed out natural ability meets deep interest in chess?> Good questions. Cultural trends create spikes in the arts. For example, by the end of the 18th century, just about everyone in Europe either played a musical instrument or sang. But I have no idea how popular chess was at various times. Minds like Morphy or Capa rarely end up in chess. That's why I guessed that one comes along in chess every 50 to 100 years. If we include Kasparov, Karpov, and Fischer, that makes only 5 such minds in the history of chess. I suspect La Bourdennais was another and Carlsen but in any case, they appear quite rarely. |
|
May-04-12
| | nimh: You see, the probability is very small that Morphy and Capa had maxed-out natural talent for chess. And actually I don't believe in the idea of talent having a kind of limit. In any case it makes more sense to presuppose that, for example, Capa or Morphy were less talented than Carlen and Kramnik, based on the argumentation I wrote earlier. |
|
May-05-12
 | | playground player: <nimh> Your material is inherently fascinating. I have no understanding whatsoever of computer analysis, but your conclusions are always intriguing. |
|
May-05-12
| | nimh: Thank you! Which of the conclusions are the most intriguing? |
|
May-06-12
| | playground player: <nimh> A lot of the kibitzing around here seems to involve belittling the great players of the 19th century and the early 20th: and yet it seems that you have applied modern methods to confirm the greatness of Paul Morphy and Jose Capablanca. Almost everybody else seems to be saying, "Oh, those players were just the best of a bad lot--even I could beat them, if they played today!" Again, I have absolutely no understanding of your methods: but I can see you've been conscientiously working at this material for a long time, and it just blows me away that you affirm the greatness of my favorite players. |
|
May-06-12
| | nimh: Actually the purpose wasn't to claim anything about greatness. Do we agree that the notion of 'greatness' means the level of dominance and impact on subsequent generations? What's being measured is the absolute level of play. No matter how much weaker Morphy and Capablanca were compared to today's super-GMs, their greatness and significance in chess history is beyond doubt. |
|
May-07-12
| | playground player: <nimh> I would agree with your definition of chess greatness. Again, I have no understanding whatsoever of your methods. I wouldn't even understand the explanations. But it fascinates me that you would rate old-timers like Morphy, Lasker, and Capablanca among "the five most remarkable players in chess history." I suppose it's similar to comparing baseball players of different eras ("If he got to hit against today's pitchers, Willie Mays would have hit 1,200 home runs...") In the long run, there's no way to do it except abstractly. |
|
May-07-12
| | nimh: I live in Europe. It's the first time ever I hear about someone called Willie Mays... Had you used Jordan and his 37.1 points in 86/87 season as an example, I'd have understood perfectly. :) I'm not even sure if his abilities at that time would have been sufficient to break the 30 pts barrier today. The level of defense and players have improved greatly. 25 years is an awfully long time. |
|
May-07-12
 | | Shams: <nimh> I don't know much about basketball but my friends who do usually say that Jordan would put up even better numbers today due to rule changes limiting defensive players. |
|
May-08-12
| | nimh: The actual effect of the rule changes should be clear if we compare the average number of points after and prior. And we shouldn't count out other factors either. |
|
May-08-12
| | playground player: <nimh> Oops! Sorry about that. Willie Mays was the greatest baseball player of my generation (some would say it was Mickey Mantle, or Hank Aaron--but ignore them). His place in baseball was similar to Michael Jordan's place in basketball. But I don't suppose American baseball of the 1950s and 60s holds much interest for Europeans. In any sport or game, it's always hard to compare players of different eras because so much changes. At least chess hasn't changed the rules, as has been done in baseball and football. Meanwhile, I enjoy your chess research, even though I can't understand your methodology. Computers, I'm afraid, will always be a mystery to me. |
|
May-08-12
| | nimh: Thanks once more for kind words!
At least I can proudly declare that I know at least a few things about someone called Babe Rush... :) I think he was quite good player, wasn't he? :) |
|
May-09-12
| | playground player: <nimh> George "Babe" Ruth was, quite simply, the best baseball player ever. Had he lived in ancient times, the Greeks or Romans would have made him a god, like Hercules. But he died a long time before I was born, so of course I never saw him play. All I have seen are the grainy old photos and newsreel clips... and the incredible statistics. If you ever get interested in baseball, <Phony Benoni's> chessforum is a good place to be. (No, he didn't pay me for that plug.) |
|
May-09-12
| | nimh: I've watched some baseball games several times on TV and played a computer game. And before that a NES baseball game. But unfortunately they haven't been enough to stimulate the interest of the game. |
|
May-11-12
| | playground player: <nimh> At one time, not so long ago, I was virtually a walking baseball encyclopedia. But then they made changes in the game, and the seats that my wife and I used to occupy in Yankee Stadium went from $15 a game to $272 a game. Yowch! What with all the changes, and mining the public for money, I don't pay much attention to baseball anymore. I don't think anyone can just "get into" baseball without having been brought up on it. Knowing the history and folklore and tradition of the game, the personal abilities and histories of the current players--I don't see how baseball can be enjoyed without that. But, like chess, and much more so than most other sports, baseball lends itself to computer analysis. In fact, contemporary baseball is intimately involved with computer analysis--no team manager would take a step without it. |
|
May-11-12
| | nimh: When was that? It's pretty hard to believe, a change of ownership, I suppose? It surprises me that games involving physical activity can be subjected to computer analysis. What methodology do they use and what does such analysis describe exactly? |
|
May-12-12
| | playground player: <nimh> If you're interested in this subject, rather than have me give you an inadequate answer, I suggest you talk to <Jim Bartle> or <Phony Benoni.> They understand "Sabermetrics" (the term invented for the study of advanced baseball statistics) a lot better than I do. The thing about baseball that lends itself so beautifully to computer analysis is that the data base is very large, and the activities on the field can be easily described by statistics. Some of these statistics go back 100 years or more. A lot of things happen in a basketball game that are not measured statistically--picks, box-outs, good passes that don't immediately produce a score, most defense, etc. But in baseball, hitting, pitching, and fielding are beautifully translated into numbers. So if you know, for instance, that over the last five years, Joe Gesundheit has batted .200 against pitcher Leo Cafone--100 hits in 500 at-bats: 70 singles, 20 doubles, 1 triple, and 9 home runs--you can make a pretty accurate prediction of what will happen the next time Joe bats against Leo. Plus you have the figures on exactly what kind of outs he made, the 400 times he didn't get a hit... and so on. I hope this helps. |
|
May-13-12
| | nimh: I read the sabermetrics article on wiki. It's a very interesting subject, and I think I'll devote some time to study it more closely.
Let me ask one more question, is that useful for making cross-era comparisons as well? Would sabermetrics be useful for predicting e.g. what anyone's stats with Babe Ruth's abilities and build were today? |
|
May-14-12
| | playground player: <nimh> That's a good question, and I'm not sure I understand the subject well enough to give you a good answer--so please don't consider anything I say authoritative. But... Cross-era comparisons in baseball should be statistically valid, provided the data is abundant and accurate. Happily, in baseball, that is very much the case. We have much the same data base for the 1927 season as we have for 2011. Up to a certain point, baseball history has been meticulously kept. Thus it was possible, for instance, for the Strat-O-Matic Game Co. to reconstruct the 1927 season in minute detail. (This is not an ad for Strat-O-Matic. I'm just using it as an example.) It's the nature of the Strat-O-Matic game that you could, if you wanted to, take Babe Ruth from the 1927 Yankees and make him a member of, say, the 1968 Yankees. Suddenly he has to hit in an era notable for overwhelming pitching and better than usual defense. If you play him for a full 1968 season, his final statistics will differ from his 1927 stats. I haven't actually done this particular experiment, so that's about all I can say. I think you'll find that baseball has the data available to allow you to make valid cross-era comparisons. |
|
|
|
< Earlier Kibitzing · PAGE 19 OF 20 ·
Later Kibitzing> |
|
 |
Deep Fritz 13
|
|
|
