About ratings

I've been observing 4p chess for more than 4 years now.

I see that a lot of inflation happened to the ratings. For Rapid FFA for instance, 2000 used to be high level 4 years ago, now it's for noobs. Still 2 years ago 2300 players were very good, today they play the 2nd stage FFA terribly, throw games, don't have a clue about the balance (!), etc. And as of today, only from 2550-2600 real high level begins (it's about 150 players only).

The leaders of the leaderboard had max. 2700 4 years ago, now they reach 3100 (!).

(The main reason of this inflation is that new players with virginal 1500 rating arrive, offer easy points to more experienced players and quite often leave forever, it feeds the system with the points. As for the leaders, they very often benefit from voluntary or involuntary farming and in fact are not THAT strong).

I believe it is our collective interest that the ratings reflect the levels accurately. In the same time, the formula behind proved to be good "in se".

(As already mentioned earlier, the only one big problem of the current formula is that it takes into account the average level of the players involved and not at all the configuration of the board: if you win in a game where you had 2700 in front and 2300 + 2300 as neighbours it's the same rating variation as for the case you have a 2300 in front and 2300 + 2700 as neigbours, whereas of course it is MUCH more difficult to win in the 2nd case).

What I suggest is very simple:

1. The games where the difference in ratings between the best and the worst players is higher than 200 pts, should be discriminated by 10, i.e. the ratings' variations will be divided by 10. As such games are much less relevant than the games with players of approximately same level.

2. The maximum variation (either positive or negative) should be limited to 10 points for all players. (I often see +25, -40... It is of course impossible that the level of a player changes that drastically in just 20 minutes...). The ratings shouldn't be that volatile.

i agree. even if im a noob, i also see this happening since years. what i find also disturbing is people on the leaderboard abuse the system (e.g. brain or queen, for rapid), pls at leat the top players!!!

2. definitely not. caps will distort the changes. if anything, changes could be scaled (think of it as +1.5 -0.25 -0.25 -1 rather than +3 -0.5 -0.5 -2 but we cant just scale some games and not others, that just distorts the overall system. also remember changes depend heavily on glicko RD. frankly i dont think this is a problem.
ideas to reduce ratings are:
perhaps, one time, reduce all ratings by 30% or so. eg newrating = 1500 + 0.7 x (rating - 1500) do this one time. q: what about blitz and bullet? same?
another option is to just introduce slight deflation, eg rather than 3 -0.5 -0.5 -1, use 2.9 -0.55 -0.55 -1.05, which will bring ratings down slowly. (and can be adjuste anytime)
fwiw i dont think it's too bad. in hindsight we should have just left the starting ratings as 1200 (we added 300 to everyone 2-3 years ago), but at that time we had no idea ratings would get this large. at the time top ffa was 2400, top teams 2100.
we could do -300 for everyone and start newbies at 1200 again, that seems like the best option. but: players dont like it. and the q is also, for which variants? atomic etc too ?? probably not. this is also a site wide and also "political" issue. i think lichess starts people at 1500.
fwiw i dont think toppest players at 3200 is that bad, magnus is 3200 too. the awkward part is more that 2000 rated 4players are terrible, unlike 2pc

Space, it's not a goal in itself to reduce the ratings (otherwise it won't solve the issue at all). The real problem is the fact that the current ratings do not reflect accurately the levels of the players, i.e. do not play their role.

I do remember when 300 were added (it was more than 3 years ago, before the COVID!). But such changes in fact do not influence anything.

Also, you compare with 2p chess, for me it is not important, 4p chess is totally different from 2p, it's another game, definitely.

The real problem is not the level that this or that category of players has reached, but the fact that currently the difference between the highest decile and the next one for instance does not correspond to their real difference of level.

That's why I believe that the changes for any game should be limited (10 pts max.) and that the games which clearly are statistically irrelevant don't generate the same rating changes. A game where you have 1900, 2200, 2500 and 2900 cannot be but pure BS, and the rating changes (for any winner) should be heavily discriminated (hence the idea of division by 10).

[..] please stay on topic

A very clear indication of the ratings' inflation is this. In FFA you currently need to be on Top 20: 2808 in Rapid, 2600 in Biltz, 2406 in Bullet. I.e. 200 and 400 points more in Blitz and Rapid than in Bullet. Why, whereas you always begin with 1500? Because there are much more players in Rapid (8786 active players currently) than in Blitz (562) and Bullet (271), so the feeding effect I mentioned above is fully proved.

Bbbbbbbbbbbbbbbbbrrrrrrrrrrrrrrrrrrraaaaaaaaaaaaaaaaaaaaaaaaaaaa

2900 isn't even enough to make the FFA Rapid leaderboard now

Most of the players and administrators were in favor of the FFA scoring system instead of SOLO - I warned you all what would happen, one of the most visible consequences is this.

Other disastrous consequences for the game are yet to be seen (change in tactics - avoid fourth place at all costs, small loss of points for second and third which allows players who have "grown up" in the FFA scoring system to still play for second place with revenge on those they don't like. In a SOLO system such players would quickly go under the radar - now they still keep their ratings spoiling the game for serious players; etc )

But I lost my battle for SOLO, so now let's enjoy the consequences of the smarter ones who advocated such a change.

@neoserbian, you are 100% right; BUT the battle is not over yet, by no means.

It is clear that 95% of the players are simply too weak to understand the issue. And most players under 2200 Rapid will always prefer a system where the 2nd gets some points, and will always consider (erroneously) that the 4th was the one who played the worst and has to be punished the most, etc.

So the morphing system was good, but not good enough. Basically it doesn't go far enough. I definitely believe that the system should be 3 1 -1 -3 under 2000, and from 2000 to 2600 progressively move to 3 -1 -1 -1. Why? Because in this case 97% of them games being under 2300, will be as currently more or less, whereas the top games will become almost solo, or pure solo.