My opinion of the new rating system.

At the risk of beating a dead horse, and reiterating Clarkey, ratings do not measure objective playing strength. They just don't.

It's not as thought being '1600' means some exact level of play, and rating systems differ in how accurate they are.

Rating systems capture relative strength, i.e., your strength relative to the other players in the system.

That's all they do; a given rating difference corresponds to a particular expected score in a match between those players (for example, a rating difference of 70 points corresponds to an expected score of about 60% for the higher rated player).

So, if two sites both use Glicko-2, AND use all the same starting parameters, and on one site your rating playing G/30 is 1600, and on the other it's 1900, that means exactly one thing.

That one thing is that the average strength of the players on the site where you're 1600 is higher than the average strength of the players on the site where you're 1900.

Neither rating is inaccurate, or inflated, or deflated. They both capture your strength relative to other players in that system based on your performances to date.

That's it. Ratings simply don't measure an independent, objective level of play.

Hopefully that's a helpful explanation :)

I really badly wanted that average opponent rating back and another subtle change that almost nobody noticed is that RD is now disappeared.

It always appeared below your rating in small fonts and gives a good idea of your accurate rating. Why it disappeared. You can keep it there as it was before and it is not even taking lot of space.

RD is no longer visible because there's no longer space for it. But now you can see how many games that rating was gained over which is a more human way to understand accuracy than some - to the average person - complicated 95% confidence interval.

Average opponent rating has also gone away because it would need to be calculated for 6 series of data points.

Also, interestingly enough, keep in mind that qualifier I used when talking about different ratings in systems that use the same system (like Glicko), namely that you also have to account for the parameters they use.

One such parameter? Initial rating, which will end up being about the average rating.

Here it's 1500.

At chess.com it's 1200. Difference of 300 points. :)

Oh, and just in case it wasn't obvious why I thought that was interesting, 300 points is the difference SeaTurtle noticed between his rating on chess.com and here.

That's interesting because assuming both use Glicko-2, that actually means the average strength of players is about the same on both sites.

Of course, that's a sample of one, so not all that reliable :)