Assessing players' performance in a single match

Often times we hear these words after a match "Oh! That was a great match, there was strong play from both players...". Or "Last night's match was the worst I played at my club..".

These are definitely subjective comments. One may feel XYZ played brilliantly, another may say it was not so good performance.

Is there a standard way to guage players' performance in a single match? using any software? or guage? 

For example, a player rated 2000 USCF may perform at a 2300 in one single match. But, how do we know? 

Thanks...

Neat question.

I think people higher to or equal to those ratings can estimate if your game had qualities of a specific ratings range.

As a simple example; if you were playing a game against someone and had  no idea of their rating, and they played h2 on their second move and didn't castle, you would be able to estimate that their rating would probably be less than 1000.

A deeper example might be if you were playing against someone who understood how to get a rook to the 7th or 8th rank, you might be able to estimate that they have at least a 1300 rating.

I'm familiar with this type of thing. I am a native English speaker and teach English as a second language. I am able to quickly estimate someones English ability by having a conversation with them because my skill includes theirs and much more.

Kleelof, I think your reply is still a subjective one. I could play a game, and say "ooh that was tough, I am able to assess this high performance game..." = no good. I am looking for a standard guage. Thanks though.

A computer program could probably do that, by assigning an ELO estimate rating to each move and averaging over the game.  Perhaps someone has done it already?

I don't mean YOU . I mean someone who is looking at a game and is much more experienced than you or whoever is playing.

With the exception of some parts of the opening, and some well known mating patterns, chess IS subjective.

This is why opening lines have so many variations and 2 different GM's can look at a position and give two differnt assesments. If this were not true, then all GM's, or any 2 players of similar playing strength, would draw everytime.

You are not going to find a 'standard' guage that can be applied to a broad range of games.

Here is a recent thread that sort of addresses this idea:

http://www.chess.com/forum/view/general/how-accurate-is-this-test-if-at-all-elo-rating-test

This would not be possible.

Elo ratings are only relevant to the pool of players. And, as has been pointed out MANY MANY times in the forums here, the ratings do not always translate from one pool to another .

For example, lets say you have 2 groups of 100 people.

The first group of players have only been playing chess for 6 months and don't study. Because of the way Elo works, you would still see all the common ratings you always see; 1300, 1500 1800. This is because Elo only calculates based on who wins, and who loses. Not any standard like "1300's play this way and 1500's play that way".

Now, the second group is full of players who have been playing for years and study regularly. Again, you are going to see the same range of ratings. Again, because Elo only tracks who wins and loses, because it, in itself, is not a guage of specific skills.

If you take a 1500 from the second group, and put them in the first group, their rating would skyrocket compared to the players in that group. ANd if you took a 1500 from the first group and put them in the second group, their rating would drop because they are now playing stronger players.

Because of this, a computer program would not be able to assign any rating that has any real meaning to any group of real players.

I agree with kleelof. I think an experienced player's opinion is the best gauge.

sounds like a load of meaningless statistics to me.........

evaluating a players performance from a single game in isolation will yield fairly limited results , it needs to be compared to a players other games for a more accurate assessment.

sure , you could get a computer to number crunch all the moves from a single game but the data wont be of much use to evalute whether a player performed above or below their average capability

You are correct.

This is why a highly experienced player is a good source for gauging play based on a single game. They have the experience of playing and witnessing hundreds or thousands of games. They are able to build a reasonable context around an individual game.

And if you were refering to my message as a "load of meaningless statictics" you are not far off. The point of the post was to point out that Elo ratings are meaningless outside a specific pool of players and cannot be applied in an objective way to 1 or even a few games played by an individual player.

Elo ratings are only relevant to the pool of players. And, as has been pointed out MANY MANY times in the forums here, the ratings do not always translate from one pool to another .

But the pool of players a computer could compare with could be anyone with an ELO rating for example.  Agree with biff55 that it wouldn't be that meaningful, but it could be done.

Because of this, a computer program would not be able to assign any rating that has any real meaning to any group of real players.

Of course it could.

I'm sorry. Maybe I am not understanding you. But you said that you agree with biff that an Elo assigned by a computer would not be very meaningful, but disagreed with my statement that a computer could not assign an Elo that is meaningfull.

It appears to me, on this point, Biff and I were saying the same thing, but you agreed and then disagreed with it.

Interestingly enough, and I digress, I think the OP is wondering if there is a perfect hack to the game, to the board - an absolute, a precisely correct set of moves. Even at my level I know there's no such thing. So, maybe, the question needs to be thought through. No offense.

How do we know that there was strong play on both sides, or that someone played poorly by their lights? I don't think I understand your question. Strong play means that the moves found were effective. A bad match means that the moves played were not effective and the player has found better moves on other occasions.

If you want objectivity I don't really understand why. Lots of human judgments depend on evaluations which can't be reduced to arithmetic or are not capable of verification. But few, if any, think this undermines the value of those judgments.

Science and mathematics is another matter. (Although, come to think of it, 1+1=2 is pretty difficult to justify).

I have seen references in these forums to chess engine evaluations saying things like, "3 errors, 2 blunders" or the like. But I am doubtful such evaluations help anyone. Perhaps if I understood or used the engines I would understand or appreciate these "objective" valuations better.

Why would counting a players' performance in a single game be meaningless? I don't get it. There are instances in history where a single match was played between great opponents. Ex: Game of the century between Fischer & Byrne

I actually found a research article "Determining the strength of chess players based on actual play" 

http://web.tecnico.ulisboa.pt/diogo.ferreira/papers/ferreira12strength.pdf

but is there is a better method?

every one evaluates a postion differently .each is good at certain postion's like thier fav opening's so strength varies between postion to postion so as game to game :).

"There are instances in history where a single match was played between great opponents. Ex: Game of the Century between Fischer & Byrne"

But how easy is it to assess how well Byrne played there? He was not yet an IM if I recall correctly and got a lost position with white in the opening. That Fischer played well is one thing, but the opposition was so weak that it's difficult to say much about the level of play.

Theoretically, you could use big data analytics to get some sense of objectivity. Compare the game played to a whole lot of games and, based on moves and counter moves, peg both the guys at some "empirically observed" level. That would be an objective gauge. Me thinks.

That way you'll get a proxy assessment, similar to an experienced player's assessment. I think.

That also makes me wonder if chess.com can run big data analytics on their huge database of games played and publish kick ass stats on typical response moves from players at various levels. Or is that already something called the chess database?

Read the article (and more so page no. 14), understand it please and come back. They do give the rating of both players in that game. They give a method to calculate but I think it is too complicated for us to do for every game. So, now the question is: Is there an easier method/ better method? 

Just like passing a comment on a chess game is subjective, the replies on this thread have become subjective. Sorry.

While you couldn't get an accurate rating for a person from a single game you could get a broad indication of where that person is on the range.

Don't know how often people get their games analysed by computer, but I find, for example, that I get a broadly similar proportion of moves judged as "mistakes/questionable" per game, although there is some significant variations?! The game is probably more useful judge if its against someone with a similar ability, fought well (eg you dont win/loose via a big unusual blunder ) and a reasonably long one.  

Having said that wasn't the question rather to rate the performance in an individual game (rather than rating a player which would almost by definition require a number of games to be looked at)? Its an interesting question because some people may play some excellent games but also some bad ones, while others may be more consistent and end up with a similar rating.  It would be good to point at a game and say, oh that was played like a 2000 rated one even though I'm 1,800!

You could analyse a game with a high rated computer, which could indicate the proportion of "ideal","questionable", "mistakes"and "blunders" etc in particular game and assign a rating from that. However this in itself could give issues since some positions are much easier to play than others which are say much sharper. Also you have to get round the fact that computers often play differently to humans. The computer may not rate a move but if it is the basis of a reasonable plan that is difficult to refute it may be played by a very good player.

 It may be better for the rating to be done by top players. It would be interesting to see how consistent opinions would be if people had to vote!

You can only play against the opponent before you, fabelhaft. If Fischer demolished his opponent then surely it is possible to say he played well?

If some young person watched a player in one game and assured me on the basis of it that the player is a world beater and s/he intended to bet a good round sum on this person becoming world champion in a couple of years I might try to dissuade them. But I would also do that if the OP's standard method of assessment existed and confirmed my young friend's subjective assessment.