Thanks, rooperi, How many think that a player with strength of 2400 is going to play 8 slow paced games and win all 8 and come out with a rating of 1800 and then that player will be satisfied?
This is a terrible rating system. And yes one solution or partial solution would be to give players a provisional rating for the first X number of games.
Thanks, rooperi, How many think that a player with strength of 2400 is going to play 8 slow paced games and win all 8 and come out with a rating of 1800 and then that player will be satisfied?
This is a terrible rating system. And yes one solution or partial solution would be to give players a provisional rating for the first X number of games.
Well, it would be partially his fault, after the frst few games he should be choosing stronger opposition.
A player of that strength will surely know other strong players here and can play against them. If s/he does have to play patzers like me for a while it would not be more of a strain than a simultaneous.
It is a bit of a shock to bump into a player playing way better than their rating but I don't really think that is so terrible a thing as to undermine the whole system.
A worse defect is that while you are still finding your level you can enter tournaments with a rating limit and then, while the tournament progresses the true strength is revealed as the rating climbs. If the final round of a tournament meant to be for 1600 to 1800 players takes place between a 1690 player who has played really well (for him) and a player by now rated 1900 then I can see the 1690 player feeling a bit aggrieved.
For a sound comparison between the two you should get atleast 100 people that play on chess.com and USCF or FIDE or whatever rating system you'd like to compare chess.com to. I don't think it's meaningless or bull to try and compare the two, since with enough data from both pools you can make a good and accurate analysis.
The idea of starting at 1200 is that 1200 is the average chess rating...
If you truly are rated 1600, you will have a 1600 rating after you play enough games.
Another way the initial ating is given is after the play ~5-10 games. Based off the final score and average ating of the other players, one can estimate the player's rating. Of course, the more games, the more accurate.
One thing that helps new players get to their "real" rating is their high initial Rd. A player's rd determines how accurate one's rating is (it also determines how much that rating can change in a single game). The more games you play, the smaller your Rd gets. If you don't play games for a while, your Rd becomes bigger.
For example, a new player on this site will win several hundred points for beating a high rated player.
Johny Really? Do you think a player rated 1200 but who has strength of 2400 will be able to find other players rated above 2200 who will play him at his 1200 rating?
I am a little amazed that some do not see the problems with this rating system and some of the ideas they have...
What is a "Rd"?
Perhaps I'm wrong about that. A high rated player would lose a lot of chess.com rating points losing to a 1200 player. I guess I was thinking that anyone rated 2200 and above would be bound to have national and/or fide ratings and play here primarily for fun or practice. But people are generally proud of ratings which, of course, is why they exist. So maybe a strong player would have to play a dozen or two dozen games against lower rated opponents on first coming here.
The Rd, I'm pretty sure, relates to a rating idea used here and invented by a Mr Glicko. When you first start - or upon returning after a long break - your rating is made more volatile. That speeds up the process by which you find your correct level. The 1200 player who could, if he found opponents - hold his own against 2200 strength players would see a very fast rating rise. Within a dozen games s/he would get to at least 2000.
The more games played, though, the less the significance of this factor. My rating here rose from 1200 to 1600 very quickly but now (mid to high 1700s) changes at the slow rate you expect.
Saying the two groups cant be compared seems odd to me. Ofcourse they can, they both play chess. You just have to state properly and clear in what you compare them. If you dont do that we all discuss something else.
In my example of the soccerplayers the two groups play a different type of soccer. But when the two groups are merged they are getting compared just by playing in the same competition.
The Glicko rating system is fine as long as you do not pick an arbitary rating that all players have to start with.
Is 0 less arbitrary than 1200?
Both zero and 1200 are arbitary when you can have something better.
However zero would be even more arbitary! [as everyone would stay at zero]
The Glicko rating system is fine as long as you do not pick an arbitary rating that all players have to start with.
Well, in the USCF, you'd get a "provisional" rating based on your opposition. But since your provisional rating can't be more than 400 points above the average rating of your opponents, you'll run in to the same thing. I mean, you play 4 games against 1200 rated players and win all of the games, and your rating is then 1600. As opposed to the chess.com system, where after winning my first 4 games against opponents with an average post-game rating of 1202, my rating was 1599... is one really so much better than the other?
A person who hasn't played any games should not have a 1200 rating, a 1500 or a 0 rating. A player without a rating should be unrated. David Gliksman, the inventor of the Glicko system, agrees with this, and his system starts everyone out as unrated.
Chess.com gives players a default rating, which is a BAD change from the Glicko system. I've heard that the reason for this is to keep people from having "inflated" ratings at the beginning. That logic doesn't make any sense to me.
SmyslovFan--what you say makes sense to me.
You know, you can call an unrated player unrated, or you can call them 1200. But if 2 unrated players play each other and draw, you still have to give them a rating. So, "Unrated" has a rating even if you don't show it.
You know, you can call an unrated player unrated, or you can call them 1200. But if 2 unrated players play each other and draw, you still have to give them a rating. So, "Unrated" has a rating even if you don't show it.
David, once there's a pool of players with an established rating, you do not have to rate a game between two unrated players. Just wait for them to play a few rated games against players with ratings. You may want to research "Glicko rating system" on your favorite search engine to see how it works.
At chess.com you start out with a rating of 1200 Gleicko (sp?) and go from there. The 1200 starting number is purely arbitrary. If the folks who started chess.com had chosen 1500 to start with instead of 1200, then every rating in the Gleicko system (blitz or online) would be 300 points higher than it is today. Given the arbitrary nature of the absolute value of the scale, discussing rating inflation or deflation relative to another rating system with a different pool of players is just silly.
Well, its not entirely silly. Someone might wish to know what rating they would achieve in a different pool. By comparing people who are active in both pools, you can derive an average approximation. Its not a big deal, I don't know why some people are so obsessive about denying the fact that this can be done.
The formula is simple: standard@USCF - 250 ~= blitz@chess.com
I'm aware that there are differences in the rating systems, and the player pools, but none of that is really relevant. The fact of the matter is that standard USCF ratings are 2-300 higher than chess.com blitz ratings.
I merely made this thread, because every thread I saw was claiming that chess.com ratings were inflated. This may have been true several years ago, but as one individual noted previously, this does seem to have changed in recent years.
The only certain way to know your rating in a particular pool is to join that pool and start playing rated games. I understand that among your circle of friends there is a correlation of:
standard@USCF - 250 ~= blitz@chess.com
Over the years here, I have heard other people make similar assertions and the number isn't always ~250 or even all that close to it. I have no reason to suspect anyone's anecdotal observations are not true, yet these differences exist. I suspect that the reason is sample size and composition. While you can extrapolate from a small sample to a large group, you really have to be careful about how you take the sample and what you do with it.
Your survey of chess club friends is not a random sample. You know each other and you play chess with each other a lot -- possibly more than you all play outside the group if you meet frequently. Over time, who you play will affect how you play. It might be that for someone who joins your group for a time and acquires the same "tribal wisdom" (for lack of a better term, meaning common exposure to the popular subset of the openings in your group, tactical/strategic styles, preferred time controls, etc.) will correlate similarly in their USCF standard and chess.com blitz ratings as the rest of your group. A different group with different tribal wisdom content would likely correlate differently.
I'm sorry if "silly" was too harsh a word to use in my earlier post. Perhaps I should simply observe that extrapolating from a non-random sample of 20 to two different groups of 50,000+ is probably not something that a professional statistician would consider a sound methodology and let it go at that.
Peace.
Saying the two groups cant be compared seems odd to me. Ofcourse they can, they both play chess. You just have to state properly and clear in what you compare them. If you dont do that we all discuss something else.
In my example of the soccerplayers the two groups play a different type of soccer. But when the two groups are merged they are getting compared just by playing in the same competition.
Exactly. More precisely, what is the purpose of the comparison? In this case, its simple: if a USCF standard player wants to guess what their rating would be on chess.com blitz, they can consult this formula:
USCF(standard) - 250 ~= chess.com(blitz)
By inference, we might assume the following, assuming that the player is capable of concentrating fully for the entire length of a standard game:
chess.com(blitz) + 250 ~= USCF(standard)
Of course, exceptions are inevitable, as these are average estimates.
Your survey of chess club friends is not a random sample. You know each other and you play chess with each other a lot -- possibly more than you all play outside the group if you meet frequently.
Oh, no, you are just making assumptions. This was not a survey of my chess club 'friends', it was a survey of random people for whom I could verify both an active chess.com and active USCF rating. Most of them were people I have never played against.
Instead of just stating I'm wrong, why don't you check the numbers for yourself? I'm pretty sure you'll find a similar average result. Don't tell me that 'others' have come up with different numbers, because those people either did the calculations wrong, or the averages have changed. Just try it for yourself.
Why not give players provisional ratings for their first X number of games. That way if you're playing a new very good player, his 1200 will only be provisional and won't change YOUR rating, while it will only be used to help establish HIS rating