Rating Calculation

Why are ratings on this site calculated using the players ratings at the end of the game?

Do other CC sites use end ratings also?

I understand that with CC, games can last months and a players rating can change a lot.

When you start a game, you look at your opponents rating and he see's your rating. If you okay with that then you guys start playing. Lets say he is 1600 and you are 1600 so if you win you get 16 points, you lose you lose 16 points.

You play you game and 2 months after your done. Your rating is say 1605 and his is 1305. So you get 4 points for your win over him. You expected 16 points though. But if his rating is 1905 you get 28 point of him. But if you lose to the 1305 person you lose, 28 points and if you lose vs the 1905 person you lose 4 points.

My point is this, when you started the game you wanted to play against a 1600 person or higher, not some 1305 person. Therefore you expected a minumum of 16 points if you win and lose a minumum of 16 points if you lose to him.

Now if this person is 1305, if you win you get 4 points, you lose to him you lose 28 points. If he's 1905, you win you get 28 points, you lose to him you lose 4 points.

Therefore, you rank them this way, 1.he's 1905 2.same as the start rating 1600 3.1305

1.he gains points might happen but not that likely.
3.very probable, you know which users i'm talking about

Therefore isn't case 2 the best? Use the start ratings which you both agreed with at the beginning of the game?

My guess is that it's because a recent rating is more accurate than an older one, and by the time you win, you get what you deserved, and not what you thought you deserved two months ago.

Just a guess only.

Originally posted by RahimK
Why are ratings on this site calculated using the players ratings at the end of the game?

Do other CC sites use end ratings also?

I understand that with CC, games can last months and a players rating can change a lot.

When you start a game, you look at your opponents rating and he see's your rating. If you okay with that then you guys start playing. ...[text shortened]... 2 the best? Use the start ratings which you both agreed with at the beginning of the game?

I'm rated at 1200 after loads of timeouts. I know I'm a 2000 player. I start 100 games against 1800 players. Each time I win, I get 32 points. After 100 games, my new rating is now 4500.

End of subject.

D

Originally posted by Ragnorak
I'm rated at 1200 after loads of timeouts. I know I'm a 2000 player. I start 100 games against 1800 players. Each time I win, I get 32 points. After 100 games, my new rating is now 4500.

End of subject.

D

Yes, then a sandbagger's advantage would start to diminish as soon as he benefits from it.

Makes sense.

Originally posted by Ragnorak
I'm rated at 1200 after loads of timeouts. I know I'm a 2000 player. I start 100 games against 1800 players. Each time I win, I get 32 points. After 100 games, my new rating is now 4500.

End of subject.

D

Nice, I like that answer and it makes perfect sense. Thanks.

I wouldn't want to play against you because of how your rating dropped so much. Reminds me of Dustn.

Do all CC sites use the player ratings at the end of the game then?

I can't answer for all sites ofcourse, but I have seen it at an other site.

Originally posted by RahimK
Why are ratings on this site calculated using the players ratings at the end of the game?

Do other CC sites use end ratings also?

I understand that with CC, games can last months and a players rating can change a lot.

When you start a game, you look at your opponents rating and he see's your rating. If you okay with that then you guys start playing. ...[text shortened]... 2 the best? Use the start ratings which you both agreed with at the beginning of the game?

End ratings are used everywhere I've played correspondence, including the USCF and IECC.

Of course, it can be bad when a player stronger than you times out most of her games, and then returns just in time to spank you. It also can lead to perverse strategies, such as playing slowly against a sandbagger as you wait for his rating to improve.

Originally posted by Wulebgr
End ratings are used everywhere I've played correspondence, including the USCF and IECC.

Of course, it can be bad when a player stronger than you times out most of her games, and then returns just in time to spank you. It also can lead to perverse strategies, such as playing slowly against a sandbagger as you wait for his rating to improve.

USCF CC site right, or USCF otb?

We use start ratings for OTB play which makes sense and I think all the OTB ratings systems use that also.

I was thinking about waiting for my opponents rating to increase before beating them to get more points but they aren't sandbagging. Might as well use the system to my advantage. But then again, he's rating could drop lower and with a quick timer, you can't wait forever.

Therefore, I'm going to pass on that idea and take what i can right now.

Originally posted by Ragnorak
I'm rated at 1200 after loads of timeouts. I know I'm a 2000 player. I start 100 games against 1800 players. Each time I win, I get 32 points. After 100 games, my new rating is now 4500.

End of subject.

D

A bit simplistic, don't you think? If it were that simple, why in the consulting vote did 33 out of 110 still vote for the start-of-game rating application? And that included me. Answer: there are pro's and con's. The real issue in your example is that a player of 2000 strength would be able to start 100 games (before concluding any) against 1200 players.

An alternative would be to calculate the ratings on fixed periodical points in time, taking account of the games that have been concluded in that period. Similarly to the FIDE ratings. I know it is not 100% comparable because in the FIDE system all games are started AND fiinished in the same period.

Originally posted by Mephisto2
A bit simplistic, don't you think? If it were that simple, why in the consulting vote did 33 out of 110 still vote for the start-of-game rating application? And that included me. Answer: there are pro's and con's. The real issue in your example is that a player of 2000 strength would be able to start 100 games (before concluding any) against 1200 players. ...[text shortened]... comparable because in the FIDE system all games are started AND fiinished in the same period.

He's explanition is nice and simple and makes sense. Getting 100 game isn't hard, and who says you gotta get them vs 1200?

Start games with long timebanks, Start using your timebanks and keep getting more and more games vs 1600,1700 etc.. after you have 100 or whatever.

Even if you win and your rating goes up to 2005 or whatever, big deal keep getting more games. However, at 2000 or 2200 they use different rating systems in OTB. I'm not sure if they do that here.

After 2200 you get 4 points for win vs 2200 instead of 16 or maybe its 8 but ya its not the same anymore.

Originally posted by RahimK
He's explanition is nice and simple and makes sense. Getting 100 game isn't hard, and who says you gotta get them vs 1200?

Start games with long timebanks, Start using your timebanks and keep getting more and more games vs 1600,1700 etc.. after you have 100 or whatever.

Even if you win and your rating goes up to 2005 or whatever, big deal keep getting m ...[text shortened]... u get 4 points for win vs 2200 instead of 16 or maybe its 8 but ya its not the same anymore.

I am not saying it is difficult to do this, only I was stating that this is a fault in the whole system. A rating system works only in a relatively stable environment. In real life, this is done through leagues, ranking tournaments, etc. ... You want to have people playing mostly games in the range of their real rating. No system will work properly if this condidtion is not met. That is why Russ has made an attempt to get new players as fast as possible in the range of their real strength by having a different method of calculating than for the established players. So, the real issue is players being far away from their real rating. If that were solved, applying the end-of-game rating would be the better choice (reflecting the reality), because in that 'ideal' world, the latest rating would be the closest to the real rating.

Sandbaggers or whatever you want to call them, and people who keep timing out and fall 400+ points are the problem to the stability.

I got my answer already about why the starting ratings can't be used so i'm good now with this thread.

Originally posted by RahimK
Why are ratings on this site calculated using the players ratings at the end of the game?
[Text cut]
Therefore isn't case 2 the best? Use the start ratings which you both agreed with at the beginning of the game?

There was a vote on this a while ago, if you dig around in the forums you'll find an explanation from David Tebb. The ratings system is flawed in that it was designed with the assumption that a player would only play one game at a time, it works well enough despite that. The problem with using ratings calculated at the start of the game is that it is unstable, as Ragnorak pointed out. The point is that our 4,500 player wouldn't risk losing ratings points on any game started when rated 1,200...

Originally posted by RahimK
Sandbaggers or whatever you want to call them, and people who keep timing out and fall 400+ points are the problem to the stability..

using a rd-factor like in glicko-rating would elegantly solve the convergence problem, and a lot better than the current system. for those who are not familiar with the idea, glicko uses both rating AND a relative rd-factor which gives an estimate for how accurate the rating is. the lower the rd the more accurate the rating. you start with a big rd, which also means rating chances are very big, so the rating will converge to your 'real' strength quickly. a bit like during the provisional-phase here. the more games you play, the smaller the rd and the smaller the overshooting of estimation. the rd also increases over time, so after long non-active periods (when rating typically becomes erratic due mass-resignation, time-outs, or change in skills) the rating would jump fast to the 'real' level again.

the big idea here is you could also modify the rd with every TO or resignation, so even if a player was active, but decided to drop a lot of games, his rating would shoot right back up during the next couple of games. so instead of maybe a 100 (?) games to get back from 1200 to 2000 it would take 5-10 to be there.

you'd also quickly see if a player really is active or not.

Originally posted by RahimK
Why are ratings on this site calculated using the players ratings at the end of the game?

Do other CC sites use end ratings also?

I understand that with CC, games can last months and a players rating can change a lot.

When you start a game, you look at your opponents rating and he see's your rating. If you okay with that then you guys start playing. ...[text shortened]... 2 the best? Use the start ratings which you both agreed with at the beginning of the game?

This exact thing happens to Rahim! That 1700s player I was playing dropped down to 1300s in a single game!!! Now, how is this possible??? What happened? Talk about POINT ROBBERY!