ichabod801 wrote:
Why would you choose the earlier rating based on less information and less recent information over the later rating?
+1
Highest rating achieved
Sort:
Bishop_2_Move wrote:
I think it should be the pregame rating,that's what they were until you beat them.
This is the key misconception. The pregame rating is not what they were. It's the estimate of what they were.
You can take your Mathematical Statistics and put them where the sun don't shine in the Cumberland Gap. The consensus is when you start a game whatever the rating is; is what you expect to receive or lose when the game is finished. This is simple, direct, straight forward, linear and logical. This is concrete not some esoteric mumbo jumbo about whether the numbers are positive negative or neither. You also will note that the number of games you play also affects your results. Play slower and fewer games and you will garner mega more points from any wins from any rated opponent then if you play more games against any rated opponent. Tell me that makes equal sense? If you play 5 games against any rating 1500-1800-2000, whatever, you will expect to gain 150-250-400 ratings points and even more per game if rating is higher whether they time out or are banned. Play 20 games against the same opposition and you will be lucky to gain 50-100 points if you beat all of them without time outs!!
Jan 7, 2010
0
#25
Schachgeek wrote:
LAGER wrote:
Hate this myself. Worse is when the player is banned or leaves and times out and you get the scraps of his/her rating. Time outs, game wins should be calculated at the rating the game started with, period!
I neglected to consider that.
Players who are banned should have their rating "frozen" for the purpose of rewarding his or her victims, before ratings changes are computed. If you allow the games of banned players to just time out and rate in that order, those at the bottom of the list will receive crumbs if anything.
this is the only thing i have issue with as well. i completely think that post game ratings are more accurate and should count in rating changes. however, in the case of banned players or players who simply leave and let all their games time out it would be great to have their ratings drop but their opponents' ratings be based on a "frozen" rating of the player. much easier said then done i would imagine though.
I think that is what is done. After some number of consecutive time-outs (for some reason 5 sticks in my mind) all further rating changes are based on the rating at that time even though the losing player still loses rating points.
modernchess wrote:
Tricklev wrote:
The pregame such as when the game first started, or the pregame such as just before you snatch the points away from him?
As when the game started, in my opinion.
Not when the games started, because of the fluctuation as other games the player finishes. I would say the rating he had just as you made the move that mate him.
LAGER wrote:
You can take your Mathematical Statistics and put them where the sun don't shine in the Cumberland Gap. The consensus is when you start a game whatever the rating is; is what you expect to receive or lose when the game is finished. This is simple, direct, straight forward, linear and logical.
And completely and utterly wrong. Don't expect me to go along with a consensus that is willfully ignorant of the truth because it can't handle something that isn't simple.
Loomis wrote:
I think that is what is done. After some number of consecutive time-outs (for some reason 5 sticks in my mind) all further rating changes are based on the rating at that time even though the losing player still loses rating points.
The policies on the site are ever-changing. The "lordbobbetti rule - post 95" of an actual ratings floor after five consecutive time outs appears to have been dropped.
Here lordbobbetti has a floor of 1439 but the rating of EpicPwnage drifts lower and lower. Makes it difficult to determine what rating is being used for EpicPwnage.
ichabod801 wrote:
marvellosity wrote:
It's nonsensical that it takes the post-game rating.
If you are playing someone rated 2000 with 10 wins, 5 draws and 5 losses, then you either win, lose, or draw with someone rated 2000 with 10 wins, 5 draws and 5 losses.
The highest win should clearly be 2000, as they weren't 1970 with 10w, 5d and 6l until after you'd beaten him. Saying it was 1970 is like saying you just beat someone with 10w, 5d, 6l, but that simply isn't how it went down. He only had 5l when you beat him.
It seems so self-evident, I can't see how people can disagree :/
You're comparing apples and oranges. A rating is not a win/loss record, it's an estimate of performance. The post game estimate is the most accurate one at the point in time which you beat him.
And rating is made up entirely of your win/draw/loss record against people of various ratings.
When a 1500 beats a 1600, the Glicko system uses 1600 as the number it derives it's +/- score from. So should the highest rating stat.
To the 1500, Glicko says "so you beat a 1600, I estimate your new current level is 1534" and to the 1600 it says "so you lost to a 1500, I estimate your new current level is 1577".
I am quite astonished how upset some people can get about such stuff...
Okay lets try to explain it not too scientifically... Those so called "ratings", are mathematical guesses. approximations. I'll try to give an example: you went to the supermarket to buy milk today. when you entered the market, you guessed the milk would probably cost 50 cent (pregame "rating"). you took the milk from the shelf, and went to pay it, you paid 45 cent (aftergame "rating"). You want to know now, which price was printed on the shelf when you entered the market? You have these two approximations: 50 cent, or 45 cent. When you made the approximation of 50 cent, it was indeed the correct time. Still, you will go and say "the sign on the shelf said probably 45 cent", because your last information, even if behind in time, is more likely to be correct.
I think the mistake some people here make is that they think this rating is actually a real fact, like a Win/Loss record would be. but it is not, it is a mathematical guess about a number that no one knows.
With other words: the after game rating is probably closer to the pregame rating than the pregame rating itself was.
I think it should be the pregame rating,that's what they were until you beat them.