Might this hueristic analysis give a better approximation of a club player’s actual playing strength? When playing an opponent who is much stronger, the club player will likely lose, perhaps by a late middle game blunder. Thus, the loss as interpreted by ELO rating change may not adequately reflect the true strength of the losing player in the opening or middle game.
chessbase.com/newsdetail.asp?newsid=7621
In my opinion, your “true strength” is your ability to win games. If you play good moves in the opening and middle game and then blunder in the late middle game and lose, you lost. It does not make you “stronger” than someone else who has the same results as you.
Alekhine said that to beat him in a game you had to beat him three times, once each in the opening, middle game and end game.
For those focused primarily, if not exclusively, on elo ratings, wins are the only focus. However, this study could provide alternative foci for determining ones strengths and weaknesses with an eye to improvement. Hence the reason the word “suppliment” is in the thread title.
Agreed. I’ve won enough lost games, even against stronger players (and lost enough won ones, even against weaker ones) to know this. I’ve even won tournaments in which I was losing at some point in every game. Some people have thought I didn’t deserve those wins, but you don’t win until you win, and you don’t lose until you lose.
The analysis these fellows have come up with is based on looking at every move in a player’s game. That makes it a potentially marvelous tool for analyzing a player’s strengths and weaknesses both historically and presently. It would be interesting to know (excluding outright blunders) how ones strength varies throughout a game by converting the heuristic analysis of each move to an ELO rating for that move. A player might be playing through a middle game at 1800, but losing games because his endgame play is at a 1300 level. So, while the player’s rating may be 1300 or less overall, the player’s real strength in a significant part of the game is actually above the end ELO rating. Not only will that help to pin point weaknesses in general, it might be used to determine which endgames a player is playing well and where study is needed.
OK, I do agree that this could potentially be a useful diagnostic tool. I was just taking exception to the phrase “actual playing strength”.
If a player plays 1800 middle games and 1300 endgames, no way is his “actual strength” anywhere near 1800.
If, say, 80% of games get to the endgame (or get to a point where endgame considerations kick in, such as whether or not to trade queens), then what you might call “actual playing strength” (the weighted average) would be around 1400.
If, on the other hand, that percentage is 20%, then the weighted average would be more like 1700.
Of the two, I believe 80% is a far more reasonable assumption.
Players who play endgames poorly are often in denial. They don’t even realize the extent to which their poor endgame play affects their final score. Sometimes, when faced with the choice between a losing middlegame and a winning endgame, they choose the former, because they are afraid of endgames.
Bill Smythe
Moderator Mode: Off
This seems to be too much trouble for the knowledge gained.
A player is only as strong as his weakest link in a given game. If he is particularly bad at certain endgame situations, but only sees those situations 1 out of 7 games, his results will be better. The rating is a reflection of the final result of the game performances.
To use the rating system as a learning tool, to find what parts of his game are weak, is cumbersome and quite inefficient.
All a player needs do is to look at his losses and see what happened in each. If a particular weakness shows itself consistently in a certain part of the game or a certain situation, then that is a weakness to be worked on, obviously.
Considering how many people are totally turned off by playing rated chess, an alternative strength rating might keep them interested. It might be an alternative “semi-rated” system to introduce people to chess without having to worry about a rating. And if you do away with rating floors, it will prove a nice alternative for those aging tournament players who don’t want to see their “real” rating drop.
Questions. How many people regularly play chess in the US? How many of them are USCF members? Now–what ya gona do to build membership. And if this rating class becomes available outside of the USCF, what will that do to growing the USCF?
Moderator Mode: Off
I talked about using the rating system as a learning tool to identify part or parts of one’s play that needs work. I still say the rating system is too wrong for this and also would be too cumbersome.
The problem I see with using a partial game or part-of-game rating system is that it would be too near impossible to rate parts of games. How would you do that, use a computer program? The problem with using a computer program for this would be that the computer programs don’t always have a correct assessment of the position and game.
Well, I’d like to see this rating system and how it works before adopting it.
Have you read the linked article and looked at the links there to the actual papers. It seems they have a program to do much of this.
Moderator Mode: Off
They say in that article that it still needs a lot of work.
They are also analyzing World Champions and their play.
If you want to dream, that’s fine. I certainly don’t see any software on the near horizon that we could use to give a player’s rating through the different moves of each game.
The logistics would also be difficult at best for the average player in the USCF. This would require that each game score of all rated games be submitted to the USCF for rating. Then the official USCF rating “machine” would need to analyze all those games to give each player his/her rating for that game. That is a lot of games and a whole lot of computing power needed.
It can just as easily be done for any player’s games and the USCF doesn’t have to be involved at all. It could eventually be marketed as a stand alone program or incorporated in existing analysis products like Fritz, ChessAssistant and Aquarium. I would be curious to know how this new approach is different from the friend and sparring modes of Fritz that adjust handicap as you play multiple games against the computer. It seems to me that given the direction being taken with Fritz 13, such a program would be a useful addition.
So, the question is, how might the USCF use this potential rating tool to promote its mission and bring new people to the USCF who are already active chess players. Why should one bother playing in rated tournaments if a player can have his strength measured by running his games played at the club, coffee house, school or anywhere through such a program?
Moderator Mode: Off
That’s a simple one. The answer is competition. Playing rated tournaments is real competition. The rating isn’t the thing without the real and standardized competition that coffee house, school or anywhere doesn’t provide.
A person can analyze and play better moves when relaxed and not in a tournament style competition, while not being able to perform as well under the rigors and pressures of a rated game under tournament conditions.
This is why the rating is so venerated by tournament chess players.
Look ICC, FICS and the correspondence servers have their own rating systems. When talking with a tournament chess player and asking them about their rating, they inevitably use their USCF rating as their real rating with the others being not so real.
I talked with a player that has a low USCF rating yet has a significantly higher correspondence rating on one of the email correspondence or server correspondence systems. This player puts his value and energy into his correspondence games, giving little to no real energy to his OTB rating. For him, the correspondence rating is a reflection of his chess strength.
If you want to attract people to chess, then it behooves us all to help identify which format is best for him/her at a given time. The rating of that format will be what he/she values. Remember, the rating is nothing more (and nothing less) than a reflection of past performance in a particular chess format.
The use of computer programs to help assess a player’s ability and strength, as well as it being a tool to help learning and training, is fine and well. There are programs, like Chessmaster and Shredder that keep track of an individual’s rating as they play against that program. In fact they alter their play quality to be more like the individual’s as play goes on, game after game.
The moral of the story is that the rating a program says one is versus their USCF or FIDE one will always fall short because those programs cannot duplicate the actual conditions of a rated game.
A question. How on earth did chess survive and even grow before the advent of ELO ratings?? Rather shocking, eh? 
ELO performance ratings won’t help your game get better, they just tell you how you’re doing, not what went right or wrong.
Attempting to derive any useful meaning from statements like “you play king and pawn endings like a 1500 player”, no matter how the statements are generated, is foredoomed. What does it mean to play pawn endings like a 1500 player? Anything specific? How does “you play king and pawn endings like a 1500 player” give any more concrete information than “your ending play needs a lot of work”?
What it means is that the program these gentlemen have devised might be used to make that determination. But, lets say the program determines that you play the middlegame like a 1500 player and play the endgame like a 950 player. An exaggeration probably, but that should tell you something about your play and what to focus on. Likewise, if you play endgames at 1800 but, play middlegames at 1300, you might need to focus on the middlegame to reach more end games. Whether they can statistically come up with an overall composite rating form this that is not merely a win-loss-draw function will be interesting. I guess it’s really an attempt to find an objective measurement rather than rely on ELO which can be skewed by human frailty (blunders, fatigue, cheating like the Soviets did to people).
Igor Khmelnitsky’s Chess Exam and Training Guide offers a self-diagnostic of one’s strength in various areas of the game. Strongly recommended!
Available here, ugly HTML and all:
YIKES. More detailed directions may be required.