Rating higher than "best win"?

ChessNetwork declines all my challenges in live chess 1 0 - he aborts the games on move 1.

Good points. 

My appologies then gentlemen, I misunderstood the use of the term 'capped'. Now that this is clear to me, I do agree with you Rolandwood

I think our current system is fine as it is.

I don't really care.  I thought the idea sought to eliminate stronger players gaining a lot of rating points by playing weaker players and encouraged them to play people at their level.

But good points were made against the idea.

Rolandwood,

You completely MISUNDERSTAND (and it looks like a lot of people are in the same boat as you, so don't worry too much) the purpose of ratings. The idea behind ratings is that it is a scientific attempt to measure the odds of what will result when 2 random opponents play each other.

To use a practical example, let's say we pick 20 unrated chess players at random, and conduct an experiment. Player #1 plays the other 19 players once, and he beats them all for a record of 19-0. Clearly, if he can be 19-straight 1200 rated players, his playing strength (and thus his rating) is clearly higher than 1200. However, according to your idea, he would not have gained a SINGLE RATINGS POINT because all of his best wins came at the expense of other 1200s (let alone the fact, that Chess.com would consider your win to be somewhere around 1175 if you beat a 1200). This makes NO SENSE whatsoever. If Player #1 were then to be matched with another random 1200 player, the computer odds would say they both have a roughly equal chance of winning the game because their ratings are the same. However, based on the previous results, it's blatantly obvious that Player #1 is a considerable favorite to win the game.

If you consistently beat 1400s, it's clear that your ratings should be 1500-1700+ to indicate this factor. This would tell other 1400s that if they played you, there is a strong probability that you will win the game. That's basically the point of ratings.

Question: Because I understand all the points against this idea....

Is there any way to incorporate something like this idea to make sure an online rating is closer to a "real life" rating?

Yes, only consider CORRESPONDENCE CHESS ratings and not OTB ratings.  Then the regular chess.com ratings will be similar.  Also, limit the range of ratings you can play based upon current rating, which is what many (High School) individual tournaments do, I do not know about other organizations.

immortalgamer,

Probably no, for two reasons:

1. People have a ton of extra resources at their disposal online: books, databases, TIME, friends, articles, choice of environment, and so forth (not even considering the possibility of chess engines). A player who is using some or all of these resources will have an edge against a player who is not. It's not unnatural for an online player to mull over a position for a long time, checking some databases, and using that god-awful analysis board. I would look at an OTB player strange if he was taking over 10 minutes/move if it wasn't a tournament game.

Also, in addition to analysis board, some people disable the touch-move rule, which gives you another huge advantage.

The vast majority of the population will have an online ratings that's far higher than OTB because they do in fact play better online (due to all the accumulated advantages).

2. There are so many opponents online, and a lot of them are looking for rated games. So if you are challenging tough competition and getting rated games, your ratings will reflect these conditions.

However, this is not so for most OTB settings. There's not a huge population of chess players out there. And the majority of games are casual, unrated games to begin with at the local chess club or coffee shop. Even in tournaments, most people probably play once or twice a year if they're lucky, so even if their skills improve a lot, it's not guaranteed that their ratings will as well. Especially if you live in a smaller town, where the same 10-20 people show up for every tournament. They all will constantly improve, but they will basically and continuously trash each other's ratings to no end. They were rated 1300s and 1400s but then they all got better and slowly became 1800s but because they only play 1300s and 1400s and constantly lose to each other, virtually no one there will have a chance to improve beyond their current ratings.

In summary, because online chess players have more tools at their disposal, and have more "official" rated games in which to measure their skill, the ratings will invariably be higher online. I'm not sure much could be done to off-set these fundamental differences.

Basically, if every single game of chess you played, online and OTB, HAD TO BE RATED, then you would get more accurate results. But that will never happen. And still, if you were playing online and had access to books and databases and extra time, and you played only at the level you would normally without these tools, then I would look at you funny as well.

if this was so, all of us would be at 1200, no one would rise because we would be caped at 1200. a 1200 betting a 1200 would be caped at 1200.

Everyone's still trying to compare their rating to their opponent's rating.  This is simple elitism, and devalues rating as an indicator of your true strength.  Winning percentage is a much greater overriding factor.  There are plenty of highly-rated players with terrible winning percentages who decline challenges simply because they don't want to lose.

Well, I guess you could just wait until your rating is as high as you want it to be, then either stop playing or play unrated games from that point on...

The current system is mathematically correct.  If you don't like it, please don't play on Chess.com.  There are plenty of other Chess sites that use inferior ratings systems like the one you propose.

I thought the site used the Gliko forumula for ratings?  If so, the forumula as written up by Glickman in his oft-cited article has a built in cap at 350 points.  i.e.  you don't increase your rating if you beat someone rated more than 350 points beneath you.

I don't know whether you interpreted his comments correctly, but it sure gave me a good laugh

 I concede the mathmatical neccesity for a higher rating than your best win. I WANT TO THANK EVERYONE FOR THE LESSON.

I feel that it is poor sportsmanship to only play lower rated players just to increase your rating.

Rude much

So you're saying that if you're the highest rated chess player in all of chess.com that you will have nowhere to go in rating but down? (No one is a higher rating than you so your rating is capped at the best win you've had up to this point?)

One possible option would be to cap the maximum rating based on the average opponent rating + 500. So if my average opponent score is 1600 then my rating would be capped at 2100. If I want to get a higher rating I have to play more people around 2100 to raise my average opponent rating.

The only downside is that higher rated players wouldn't be so keen to play lower rated players, but others might not see this as a disadvantage at all. It would also stop someone flying off into the distance (such as chessnetwork) because they could never have a rating more than 500 points higher than the rest of the pack.

Ah, the real reason for the suggestion...

Everyone wants to increase their rating.  But the truth is, it doesn't really matter who you play.  You're better off playing those people that you know you can beat, for sure, but I have beaten lots of opponents rated higher than I am.

Rating is a personal thing.  Use it to determine your own strengths and weaknesses and stop worrying about what other people use it for.  If they want to have a nice, high, shiny rating, then they could always pay a bunch of 3rd-world workers to open accounts and play losing games.  What would it really matter?  Sure, I'd love to be rated 1700, but I'm not.  And the point of that is that I really need to improve - I believe that this is what ratings are all about in the first place.