|
|
|
Author
| Message |
|---|
|
|
I want real elo system in QA! Please, Miguel! 
|
|
|
I do not know if I have a fide rating, but I have a USCF rating. Lately it has fluctuated between 1950 and 1980. The USCF ratings change very slowly. I had a performance rating of 2025 over 28 games but my uscf rating went from 1803 to 1977. QA ratings change very quickly even if you are over 2100. Lately my rating on QA has been 2200 give or take 50
Prior replies have discussed causes that make an indiviual's rating inaccurate. Here is another example. It is possible to postpone all your loses by not resigning and taking 6 days 23 hours per move in the games you are going to lose. But playing quickly in your other games. Even if you only play in Auto tournaments, ie constrained by 50 games in progress rule, you can score 50+ consecutive wins. That will inflate your rating enough to play in the next higher class of tournament. If you want to play in a lower class tournament without losing any game on purpose you need to play many 1 move per day games while you are stockpiling you 50 resignable games. The 1 move per day games will keep your rating near your true playing strength. Then suddenly resigning 50 games will deflate you rating by hundreds of points.
These tricks temporily move rating points from 1 player to the comunity or vice versa. When the trickster resumes normal play the points that he loned or borrowed from the community will be restowed. The larger concern is when a player gains or loses points and that are not balanced by his opponents change in rating. Prior replies have given examples of this.
If QA never addds points to the rating system, soon we will all be underated compared to fide or USCF or Fide. After my 50th game my QA rating was a near 1850. Playing 2000 games made me a better player. When I resummed USCF play after 2000+ QA games my rating went up by more than 150 points. Suppose that every point I gained here cuased my opponent to lose a point. Further suppose that my opponents stopped playing after my game with them. If they wish to regain the points the lost to me by playing me they will have to defeat a stronger player as indicated by my USCF rating. Playing a game makesboth players a tiny bit better. And this would cause the average QA rating to go down when compared to the averageof Fide.
Others have discussed the methods QA uses to offset this. The USCF has a rating floor. My USCF rating cannot go below 1800. When I beat a uscf player at his floor my rating goes up but his does not go down. There is also something called bonus points. someone who has an unusually strong tournament will get extra points.
more later
|
|
|
I have no problems playing a 2000 player here in QA... But on FICS a 2000 ELO player gives me serius troubles...
|
|
|
Miguel wrote:
| The implementation of the rating algorithm is correct and works as designed. Could be improved, yes, but it is not broken. |
Actually, the implementation is incorrect for provisionally rated players. The "special rule" that a provisional player's rating does not change if their opponent's rating is not within 400 points of them is not included in the USCF postal rating system and for very good reason: it has a disastrous effect on provisional ratings. The "special rule" means that, if you lose your first game (or even win it against a weak player), your rating is stuck at a low value, regardless of how well you perform against stronger players.
For example, if you lose your first game against a 1900 player, your initial rating is 1500 and will stay at 1500 if you only play against 1901+ rated players. So, even if your next 24 games are wins against 2000-rated players, your rating will stay at 1500 when, in reality, it should be about 2400. Indeed, if the results happen in a different order: 24 wins against 2000s followed by one loss against a 1900, the rating will be 2400. If implemented correctly, provisional ratings do not depend on the order of the results.
The special rule is wrong, wrong, wrong. The whole point of provisional ratings is that they should quickly adjust to about the right level.
Dave.
|
|
|
phystutordotcom wrote:
| If QA never addds points to the rating system, soon we will all be underated compared to fide or USCF or Fide... |
It's a mistake to think of ratings in this way. Ratings are a measure of your recent performance, not a traded commodity. Note that different rates of rating change (so-called "k-values") for players rated above and below 2100 mean that rating points are not conserved: if a 2100+ player plays a sub-2100, the former will gain or lose fewer rating points than the latter. The total rating of active players will also vary as new players come into the system and old players leave.
There's no reason to expect any particular relationship between ratings here and FIDE or USCF ratings. There's also no reason to expect any relationship to be constant over time.
| The USCF has a rating floor. |
This has nothing to do with keeping ratings constant over time. The point of rating floors is to prevent "sandbagging" -- deliberately losing a bunch of games so your rating falls, letting you enter and win prize money at a lower level tournament.
Dave.
|
| Previous 1 2 3 |
|
|
|
