the rating system is totally screwed up .
what do you mean by this, you have to be more specific to makes sense?
I agree that rating system isn't very good here. There are several fundamental problems.
1. The rating is comparable to FIDE and USCF ratings.
2. Lichess doesn't use the elo rating system which makes it very hard to understand the ratings.
3. There isn't much stability with the ratings. I have wild fluctations in my rating in a single day. I don't think this is just do to variability in my playing or who I play. During periods, I can easily beat players in the 1700's and then during other periods I struggle to beat them. Basically my rating fluctuates in the span of a week between mid 1700's and mid 1800's. It feels like the ratings as a whole fluctuate wildly up and down.
1. The rating is comparable to FIDE and USCF ratings.
2. Lichess doesn't use the elo rating system which makes it very hard to understand the ratings.
3. There isn't much stability with the ratings. I have wild fluctations in my rating in a single day. I don't think this is just do to variability in my playing or who I play. During periods, I can easily beat players in the 1700's and then during other periods I struggle to beat them. Basically my rating fluctuates in the span of a week between mid 1700's and mid 1800's. It feels like the ratings as a whole fluctuate wildly up and down.
1. The ratings are not comparable to USCF or FIDE ratings. Ratings are only comparable to others in their rating system.
2. Lichess ratings are easily understood by comparing your percentile and not the actual rated number given if that troubles you.
3. The Glicko-2 system that Lichess utilizes is one of the best rating systems available to date and is stable- much more so than the very dated ELO system.
In summary, whatever flaws you perceived are either biases, ignorance, or perhaps inherent issues in all rating systems universally, nothing specific to Lichess’s rating system itself.
-Jordan
2. Lichess ratings are easily understood by comparing your percentile and not the actual rated number given if that troubles you.
3. The Glicko-2 system that Lichess utilizes is one of the best rating systems available to date and is stable- much more so than the very dated ELO system.
In summary, whatever flaws you perceived are either biases, ignorance, or perhaps inherent issues in all rating systems universally, nothing specific to Lichess’s rating system itself.
-Jordan
Presumably online ratings have a good deal more volatility primarily because people can play so many games (compared to OTB).
Lichess uses the Glicko 2 system for ratings.
Read more :- lichess.org/page/rating-systems
Read more :- lichess.org/page/rating-systems
1. no rating system is comparable to another. basic feature on any rating system with different pool of people
2. Elo-system and Glicko-2 produce ratings that mean totally same thing there is nothing more to understand. i.e 200 points difference means stonger player get 3/4 points int the long run . Glicko-2 is just faster in finding correct level.
3. variation are normal on short time limits as the results wildly vary. Given you strength @morphista if were to attend a OTB tournament you woudl be playing with K=30 meaning in normal weekend tournament with 5 rounds you could have swing of 100 points assuming you more thatn 30 rating games and 200 points if you had less.
Elo rating system is badly dated as was designed at time when calculations had to done in paper using a table of expectancies. Glicko-2 is also dated but commonly used. It is lot better than Elo but nowhere close to being goodin modern stadards.. some maximunlikelihood estimator over player history woudl converge faster and be berhaps tidy bit more stable if the perforance of the player is stable.
For better - no source code available - could be http://universalrating.com/ .. it is maximum likelihood estimator over history of games and one developer is Mr. Glickman developer of Glicko-2. there are plenty of similar systeml like the bayes-elo
2. Elo-system and Glicko-2 produce ratings that mean totally same thing there is nothing more to understand. i.e 200 points difference means stonger player get 3/4 points int the long run . Glicko-2 is just faster in finding correct level.
3. variation are normal on short time limits as the results wildly vary. Given you strength @morphista if were to attend a OTB tournament you woudl be playing with K=30 meaning in normal weekend tournament with 5 rounds you could have swing of 100 points assuming you more thatn 30 rating games and 200 points if you had less.
Elo rating system is badly dated as was designed at time when calculations had to done in paper using a table of expectancies. Glicko-2 is also dated but commonly used. It is lot better than Elo but nowhere close to being goodin modern stadards.. some maximunlikelihood estimator over player history woudl converge faster and be berhaps tidy bit more stable if the perforance of the player is stable.
For better - no source code available - could be http://universalrating.com/ .. it is maximum likelihood estimator over history of games and one developer is Mr. Glickman developer of Glicko-2. there are plenty of similar systeml like the bayes-elo
I think it is very difficult to keep such a system at bay. It‘s like high-frequency trading. People come and play until they reach their cap, make a new account, play until they reach their cap… a lot of new players come, how to assess them precisely and so on.
It has to be relative simple mathematics coz you cannot treat every exception individually.
It has to be relative simple mathematics coz you cannot treat every exception individually.
If players were more interested in the pleasure of playing than in their level, this kind of topic wouldn't even exist!
Yes, totally, a rating system used by a site with 50 million players (i think) is scuffed.
(sarcasm)
(sarcasm)
This topic has been archived and can no longer be replied to.