chess online
| From | Message | ||
|---|---|---|---|
|
 Harpov's floor / rating system flawThis is an obvious flaw with the current rating system. My suggestion is that if a player remains at or near his floor for 50 games then his floor setting should be recalculated downwards. If this policy had been in place then harpov's floor would now be at 1300 and his continued losses would have negligible effect. |
||
|
Most importantly, the rating floor solves 2 big issues: someone's rating being reduced down to next to nothing because they were unable to play chess for a period of time and timed out in all of their games, which is very hard to recover from, both psychologically and time-wise. And the second issue it helps eliminate is when the player who timed-out in a lot of games was returning to playing, their rating being significantly lower than their real playing strength, which lead to a lot of unpleasant surprises, frustration and unnecessary accusations of "sandbagging" from their opponents. We are however considering automatically reducing player's rating floor by 100 points once a year (i.e. on January 1st, for example), if their current rating is also within 100 points of the floor, but it hasn't been set in stone yet. |
||
|
|
||
|
For example, let's imagine a player rated 2000, who plays 20 games against other players rated equally (2000). Let's say they lose half of the games and win the other half (as expected). They will end up with roughly the same rating (assuming their opponents' ratings will stay the same as well). That's because they'll be losing 10 points after each defeat, and gaining 10 points after each win, so the wins and losses will cancel each other out. Now, let's give that same player 400 rating points more out of the kindness of our hearts. Since they are still playing at the same level (2000, even though they are rated 2400 now), if they engage the same opponents as before (rated 2000), they'll still lose half and win half of the games. But in this case they'll only be gaining +1.8 points after each win, and losing -18 points after each defeat (remember, it's now rating 2400 vs 2000, or +400 point difference). So in the end, their rating will be lowered by +1.8*10 - 18*10 = -162. Which is over a third of the distance back down to 2000, their true rating. Eventually, after enough games are played, their rating will return back to 2000. Now, let's assume that the same player, plus 400 free rating points, decides to only play equally rated opponents (i.e. 2400 vs. 2400). The math says that because they are still playing at the 2000 level, out of 20 games, they are only expected to win 2 games and lose the other 18. In the end, their rating will still be lowered by +10*2 - 10*18 = -160 points (remember, for equally rated opponents, it's +10 points for a win, and -10 points for a loss). So they are still heading back down to rating 2000, no matter who they play, and the 400 free rating points we "handed out" to them didn't make any difference in the long run. By the way, the fact that the amount of lost rating points (-162 vs -160) is roughly the same in both examples (vs. 2000 and vs. 2400 players) is no coincidence. It is yet another excellent property of the Elo formula that is used as the basis of GameKnot's rating system — in the long run, it doesn't really matter who you play against, lower rated players or higher rated ones, your rating will still end up roughly the same. You'll just have to win a lot more games vs lower rated players (i.e. invest more time), or be prepared to face a lot more defeats and not so many triumphs vs higher rated players. But in the end, the only thing that will affect your rating is how well you play chess, nothing else. Although you'll probably learn more from playing higher rated opponents... |
||
|
prohibition |
||
|
donfernando |
||
|
Is there any data on this? Some of the "seniors" of GK (mateintwo, cyrano, fmgajin, for example) have roughly kept their ratings, however, indicating that either there is no inflation (or that they're simply playing worse these days... ). So, have you guys looked into this (just for curiosities sake) in the past? |
||
|
The top ratings are not a good indication of a rating inflation for a number of reasons. For someone at the top who doesn't play a lot of games at the same time, it can take a while to reach their true rating, simply because there aren't that many equally matched opponents near the top. There are also new players coming up all the time, who's never played on GameKnot before. There is also the so-called Flynn effect at play, because chess has a direct correlation to person's IQ level: en.wikipedia.org |
||
|
|
||
|
|
||
|
Now, back to something that matters? |
||
|
tobiasverhulst 27-Mar-13, 03:39 |
|
||
|
Also, that a system seems ready-made for abuse does not mean that it IS being (or will be) abused. |
||
|
|
||
|
Abuse of any system is likely to happen For example, let's say a newly rated person plays enough games to establish a rating floor of 1100. That person can play at that level for as long as they want although their "real" rating is 1700 at home. So that person can play in a a 1400 rating or less tourney and have an easy time of getting a winning prize. @ Gameknot I think even maximum security prisons can't 100% guarantee all rules are followed nor any other system for that matter. - Jeremy |
||
|
Then there are penalties that the CC board will bring down on you with a continued pattern of sandbagging....as in #18 below. Ratings: 17. Once play begins, games are rated whether they conclude normally or by forfeit. If your rating decreases by one or more rating classes as a result of forfeits or withdrawals, you will be required to enter prize tournaments in the rating class you occupied before the forfeits. Penalties 18. The CCD may assess penalties for violations of these rules. Penalties include, but are not limited to, informal reprimands, warnings, reflection time reduction, forfeitures, or withdrawal. Warnings are usually issued before more severe penalties but the CCD may skip this step. Smooth and timely completion of games is the main consideration. Penalties will be assessed as necessary to accomplish this purpose. TA |
Log in 
