Originally posted by BishopcrwNo quibbles. Just answer this...
Anything else you want to quibble about?
A genuine 1300 rated player plays a genuine 1000 rated player a match consisting of one million games. What do you estimate the end ratings to be?
I'm saying 1300 and 1000 respectively. Your answer?
There is a high probability that a 1300 player will lose to the 1000 player often enough to offset any long term gains.
For example, if the win expectancy is 84.9% then the 1300 will gain 4.83 points for a win, and the 1300 player will lose 27.17 points for a loss.
So the expected value of long term gains is 84.9%(4.83) + 15.1%(-27.17)=0.
In the long run, the expected rating increase is 0 and the ratings will stay the same (this assumes no player's ability changes in that time).
Originally posted by chesskid001Surely if a player does this there is always a risk they will lose and their long awaited higher rating would crash. Along the way they are also giving many lower rated players the chance to play a higher rated player and learn from the experience. If it is, as you suggest, a way of creating an "artificially" high rating then it is a long slow way of earning this rating and I'd be happy to leave them to it expecting that one day they will want to know what it's like to play a stonger player.
I think a nice statistic that should be added to the player profile on RHP is the average opponent rating (at the time of starting a game). This would help reveal who is highly rated for beating low rated guys, vice-versa, and who is genuinely at their rating.
You could also argue that a player might acheive an artificially low rating by only playing higher rated players and this may give them some unfair advantage say in a ranked tournament. It's very difficult to legislate against this sort of thing and probably rare enough not to be a concern.
Originally posted by Varenkayou're right about the effects of playing carefully of course. but do you think someone at 1600 by playing only 1200's is as strong as someone at 1600 by playing only 1600's? I don't. my subjective feeling from playing both kind of players is that the one playing only 1200's feels around 100-200 points weaker than his rating suggests.
But if you're playing more carefully, you're playing better. And yes, it's easier to increase you're rating if you play better against any level of opponent.
in banded tournaments you see the effect of this clearly. -if all players were say, 1600-1699, the groups would end up with everybody having about equal amount of wins. but it doesn't go that way. there are BIG differences in the results, some even losing almost all of their games. I think it's because some players have been playing against weaker players than others, on average.
I don't think it's a flaw, but an inherent trait in the system because of it's nature. rating system does not give an absolute indicator of a player's strength, it's only a statistical approximation of a player pool. on average, it's quite accurate. but it breaks down a bit when describing players that play differently from other players of their strength on average. in effect, the formula assumes everybody is an average player of his particular strength.
This debate doesn’t rest on subjective opinion - it’s based on concrete equations.
If a genuine 1500 player wants to try inflating his rating to 1600 by only playing opponents rated 1200, how does he do that if his win ratio corresponds to a difference of 300 points? He can only increase his rating to 1600 if he maintains a win ratio corresponding to 400 points difference. But then, if he does that, he genuinely is 400 above 1200 and therefore deserves to be 1600.
FIDE awards the Master title to players who’s FIDE rating gets to 2300 or above (given certain criteria). It would not be cheating to advise players aiming for such a title to only play FIDE rated players around 1900 to 2000. Ever heard such advice? No. Because it’s no easier than playing opponents rated 2300, 2400 or 2500. The strength of play may vary a lot, but so will the points being gained or lost in each game.
There are plenty of reputable web sites that explain the rating system. Just show me one site that mentions the “inherit trait” that allows inflated ratings by consistently playing weaker opposition. I can find references to rating inflations due to players entering or leaving the pool of players, but nothing relating to what you believe is possible.
Originally posted by chesskid001In theory, it should not matter.
I think a nice statistic that should be added to the player profile on RHP is the average opponent rating (at the time of starting a game). This would help reveal who is highly rated for beating low rated guys, vice-versa, and who is genuinely at their rating.
If you play 100 games against someone whom you can beat 99% of the time, chances are you'll win 99 and lose 1. If your rating is accurate, your one loss should knock you down so far that it eliminates the gains from your 99 wins.
I don't know if there are any studies done on ELO ratings to check how reliable they can be throughout the range of rating differences.
Originally posted by VarenkaReally! Really!
This debate doesn’t rest on subjective opinion - it’s based on concrete equations.
Concrete equations. That is pretty solid no variabillity to move or shift those equations at all.
WOW! I guess you have got me there!
Can you feel the sarcasm?
The equation is based on probablilty which is an ESTIMATE of predicted outcome! It is niether precise and only partially reliable.
Thanks to Space and Tech for their chip ins to this debate.
When playing however many games (100-1,000,000) and a players likely hood of winning is 99%(example only). Although in theory it shouldn't make a difference, it means another average has to be take of his high and low score, due to the wild fluctuation of his many wins and rare losses, to get anything reasonbly close to what level at which he is actually playing. That means his history is only occaissionally accurate. And if you look at the rating further you will notice that one of the first pieces of the entire equation is your previous or historical score.
And if you don't think that probability estimate can't be further manipulated, particularly on online sites, you are incorrect again.
Because you are forgetting about a players color preferrence. Which they have significant influence over when setting up challenges and invites. And if you do know so much about the rating system on the web then surely you know that the ACP (Association of Chess Professionals)released a suggestion improving ELO includes this concept and is as reasonable for other methods as well.
I asked a simple and direct question, see "Your answer?" above. Just provide an answer.
You can't comprehend what a "concrete equation" is. A "concrete equation" means we don't have to guess or be subjective about how the rating system works. There are concrete equations that specify exactly how it works (see the FAQ). I wasn't referring to the values output from these equations. Understand now?
You acknowledge two other posters, but you fail to highlight that both disagree with you.
You try to introduce something else that may skew ratings, i.e. colour allocation. If this is part of the problem being discussed here, highlight which part of the initial post addresses it. There's no mention of colour allocation; just a myth that ratings can be inflated by consistently playing lower rated players.
If nothing else, just answer the simple rating question I put to you. Then everyone can judge for themselves whether you understand the rating system. I was able to provide an answer.
Originally posted by chesskid001A lot more work for Russ! 😳
I think a nice statistic that should be added to the player profile on RHP is the average opponent rating (at the time of starting a game). This would help reveal who is highly rated for beating low rated guys, vice-versa, and who is genuinely at their rating.
Originally posted by VarenkaLets say, that you are rated 1500 and play 1000's and become 1600. That doesn't mean you play 1100's, you just keep playing 1000's and winning, albeit slowly.
Supposing I only play people 500 points below me. According to yourself this should allow me to increase my rating. So my rating increases and I once again subtract 500 from it, and I only play players at this new lower limit. I repeat again and again, and my rating keeps increasing and increasing...
Or does it? Of course not. This notion of "it's ea ...[text shortened]... sense. If you think there's truth in it, you don't understand how the rating system works.
To go back to the original point : what would be more interesting, and useful, for me would be to see how many games a player had won against players rated higher than himself.
I would look at a 1900 player differently if I know that he's beaten a bunch of 2000+ rated players in the past.
Originally posted by chesskid001What is the win expectency of a genuine 1500 versus a genuine 1000? What are the points gained or lossed or a win, loss or draw?
Lets say, that you are rated 1500 and play 1000's and become 1600. That doesn't mean you play 1100's, you just keep playing 1000's and winning, albeit slowly.
I'm not asking for an opinion here; I'm asking you to look at the actual rating system described in the FAQ. Your hypothesis doesn't match up with the real equations.
Originally posted by chesskid001Here's a statistic:
I think a nice statistic that should be added to the player profile on RHP is the average opponent rating (at the time of starting a game). This would help reveal who is highly rated for beating low rated guys, vice-versa, and who is genuinely at their rating.
SUBSCRIBE.