Ten months ago, I started a topic here talking about rating inflation. I had the perception that a rating inflation was happening here. At that time, it was a perception. Now it is certain.
When, last year, I wrote about the possibility of a rating inflation, my rating was 1912 and according to the stats I was "better than 87.8%" of lichess players, that month. Today I'm 1844 and, according to the stats, I am "better than 87.5%".
Also, the average rating at the 50th percentile was about 1612, now it is about 1550. And I think it's still decreasing, going down to 1500.
This is just "the numbers". I didn't make any consideration about the strenght of the pool. I also don't play live and I don't know how much difference there is between a traditional ELO rating and Lichess'.
If the average rating on a site decreases this hardly corresponds to a rating inflation taking place - quite contrarly this may be a proof of rating deflation - or am I wrong?
I think this just means that there are many new players (more weak players and few players stronger than you).
Logic.
I agree with Rembrandt16. This is due to new members.
The numbers going down would indicate deflation.
I didn't play on Lichess for one year. My rating stayed of course the same, but I gained several percentiles in the distribution and I am now better than 91% of classical players.
So by doing nothing, in a few years I will be the crème de la crème, if everything goes according to plan.
I myself think the Geko2 ratings are harder then elo, elo I am rated over 2100 and gaining but here only 1480, time controls make for the ratings problem here, some of the high rated players here only play bullet 3 min time against players over 500 pt lower then themselves, there is a great difference between 40 moves in two hrs and 8 mins 5 sec intervals. I see the higher players dropping out not new lower players, but it
is much easier to find a game against lower players. I get the same amount off my rating as on it for one 100 pt above or below. This isn't the case in elo.
2100 elo and lichess classical <1500? Then I would be Magnus' range with a lichess classical 2100 ...
@Itzal Deflation: To calculate Inflation you have to be able to let at least one variable stay constant over time. For example your real player strength. Assume your player strength against Stockfish is konstant from t=0 to t=1. Then you can calculate if your Rating is "deflating". As long as your real strength (against a benchmark) is fluctuating. You are not able to do that. Another possibility is to establish rated games against engines. If we fix a rating of 1500 to level 2 stockfish and let it play against real players it should on average perform ~1500. If it performs higher one could argue that some variables in the rating system should be adjusted. However, I do however have the impression that the rating gain and loss is slightly increased. The argument that more lower rated players decreased the average rating is imho not true. Since rating loss should somehow be related to rating gain. ??
@Zara90 In a system, the average rating is 1800. We add 100 new players at 1500. The system average has deflated.
@jonesmh Assume you start off with 2 players (A and B). If Player A wins, player 2 loses vice versa. Lets add another player, player C at 1500. What if A and B win against player C? Does the rating inflate? No. The average is predetermined within the parameters of the glicko system. Now with a sample of 71,498 weekly classical players I assume that small fluctuations caused by new players doe not affect the average. Rating systems are closed systems, internal comparison is possible. If you add 10000 new players with 1500 and the current average is at 1550 for 71500 players. The new average is merely 1543. The rating does not "deflate" heavily. Although you might be in a higher percentile. But even then if you double the amount of players lower then you and you were 90% now yoú are 95%.