Online Ratings System Adjustments

Beginning at 6 AM CDT on Tuesday, June 22, and continuing through until 6 AM Thursday, June 24, US Chess will be conducting a necessary update to its ratings server and system. Any events submitted during this period will be rated after the system is tested and restored.

What’s Going On?

There are two issues in play. One is a programming bug that has been in the online ratings system since it was launched. The K factor is computed incorrectly (1/4 K rather than full K). This essentially means all online ratings are wrong. Most players online ratings will go up, but some will go down.

The second issue comes from the Ratings Committee, and was discovered after reviewing data on how ratings are initialized from other ratings information. They came up with a blended / weighted / time sensitive approach where more recent ratings are given more weight than older ratings. It will look at the other five ratings types plus FIDE and Canadian/CFC ratings. (CFC ratings still have to be set up manually. We do not have cross-links between CFC IDs and US Chess IDs like we do between FIDE IDs and US Chess IDs, so there is a new manual adjustment record type for entering CFC ratings that can then be used for blended initializations.)

The online ratings need to be completely recomputed back to 2014 due to the programming bug. The Ratings Committee wants to use June 1, 2020, as the start of the new blended initialization formula, because that way any (new) players who have an online regular rating but not an OTB regular rating would have their first OTB regular rating based at least in part on their online regular rating.


I don't know if I like retro fitting based on online play history, I have never really been convienced when playing at chess sites that I am not playing a 'bot from the website.

I'm not sure what you mean by "retro fitting". The online ratings themselves are being recomputed. OTB ratings are being recalculated from last year but only to allow for the (very rare) case where someone had an online rating that was used to initialize their OTB rating (which was being done anyway---it's just a better online rating being used now).

So far I’m kind of concerned with this. My online rating in blitz looks to have gone up. However my OTB quick rating has gone down and is also reporting a number of games played above what I have actually played in my provisional process. Previous to a rated tournament last night I had only played 15 games. My rating is showing that I have played 19. It is also showing almost a 150-200 point drop in the provisional rating that was previously published and the one currently showing in the cross tables whilst other people's ratings remain the same. I know the system needs time to update, but that definitely shouldn’t have happened.

I am very happy about this. I noticed a while ago that players kind of get stuck in the middle and that there's not really people rated above 2000 and have been wondering for a long time how good i really am because people's online ratings were more often then not so much lower compared to their otb ratings which didn't make much sense. Like know/have played a national master that's got an 1800 online blitz rating and a international master that's got a 2000 online blitz rating and they arguably don't play like that because they have those titles and significantly higher ratings on and have played 1800s that are rated 1400s in blitz so very happy about this :D because it finally puts the online ratings in equalibrium with over the board ones and has them going by the same k factor. Thank you guys for double checking things/inspiring to play even more :D Recently renewed and haven't been playing rated because I didn't see the point I think largely because of this whereas now I'm inspired to see how good I am again because online is going to match up close to otb that i don't have much of in my area :D Many other players i think feel the same way/many thanks :D

Thank You. I am an average B-Class player. Now due to life and job commitments I don't play as much. But I feel I am decent player. LOL.. Now that I see my new rating albeit currently in the C-Class (top end). I am sure after a few more games my rating will go up. I really appreciate your technical staff for correcting the ratings. I hope after this pandemic calms down I would like to play in OTB one day but it is still a little bit too soon for that. So for now I really appreciate we can play online in the comforts of our home. Again thank you for correcting this because I know I play much stronger than a 1399 rating. Not to discount other players who are rated there (because I was there about 20-25 years ago) .

Bad to idea to initialize someone's OTB rating based on online play. Many of us have not been functioning at 100% over the past year due to anxiety and other issues from the pandemic and would normally avoid competitive chess under those circumstances. But many of us still played online rated anyway, due to the promise that online rated was completely separate from OTB rated. Now this latest change seems like a bait and switch from USChess. Thankfully this change has not affected me personally, since I already have an established OTB rating, but this could be unfair to others.

Agree. It did not affect me because my OTB ratings were established before June 2020, but it is unfair to other people. It is also a bad idea to think that the level your OTB Rating is is based on your online play as most of us are stressed out from the COVID-19 pandemic and as a result, do worse in competitive chess compared to if it was in person.

I am pretty sure this only directly affects people with provisional ratings, and only for so long as their rating remains provisional, based on my understanding of the formula. There might be some re-rating of games played against provisional players, but this should be a minimal effect.

Honestly, I do not know what will happen to the future ratings and I an confused. My online quick rating literally increased by almost 200 points, but my online regular rating decreased by about 50 points.
I am also worried because my first online regular chess tournament says I played 14 games when in fact, I played 4 games. Same thing with my first online quick chess tournament, but it was 13 games when I played 3 games.

Regarding the 14 and 13 games... when you play online it initializes your online rating as though you have 10 "units" of your OTB rating if you have one. So it is 10 repetitions of your OTB rating plus the four games in the one case or the three games in the other.

Too many FIDE and sandbaggers
Should be allowed to compete with other provisional players not with long time established rated players. Its cheating at tournaments and USCF just wants its $$

Game counts aren't intended to represent the exact number of games you have played in that ratings system, they're intended to tell us which ratings formula to use (we use a different one if the count is less than 9) and also to decide when a rating has become 'established' (generally, game count > 25)

When we initialize someone's rating, we assign a game count to it.  For players who came to us with an existing FIDE rating, that game count was either 5 or 10, but most other ratings systems didn't see an assigned game count  Now we assign a game count to every rating as it is initialized, based on the other ratings information we have, using a blending formula that takes into account how recent those games are and the player's age.  That game count is usually 10 or less. 


Mike Nolan, Ratings Consultant

Dan, I think you're referring to some events that were run on an online chess server under rules that existed back then, before we had any ratings systems specifically for online events.

Those would have been treated as OTB quick events, because that's what they were coded as, and would be part of your OTB quick ratings history.

The ratings system needs to take into account the acceleration of ratings with today's computer programs and access to training materials, etc. My contention is that ALL players ratings should be adjusted upward especially in OTB where today's 2000 player is yesteryear's 2200. IMHO. Floors should be raised 100 points.

I'm not sure the new algorithms are correct for calculating rating changes. In my most recently rated online tournament, I played two games then dropped out. One was a win, the second a loss.

I expected my rating to go down (since both player were rated lower than myself), but I ended up losing 22 points. This seems excessive for one loss.

These online ratings have been messed up from the start. The new changes might be a step in the right direction but the ratings of my entire online club are still wrong. In OTB play I became a Candidate Master around 1979. I was soon nearly 2100 around 2050 until 2011 until I reached my 60's. Now at age 71, I hover around 1850. Since the pandemic I've had many  online events with my students. They are getting better but I play about the same, as evidence by my gain of a few points at the US Open. I was then shocked to see that my quick rating is below 1500! I was above that when I first joined the USCF over 50 years ago!  And my students all have ratings that are ridiculously low. This system in it's present form is worthless.

Be it hereby decreed that ALL chessplayers are stronger than their ratings indicate. The USCF should be willing to send certificates attesting to this fact to anyone who requires it.

The rating-algorithm document says that up to 10 games of a regular rating are used to initialize a player's first rating at a faster and/or online time control, and that up to 5 games at the faster and/or online controls are used to initialize a player's over-the-board regular rating, if the player gets the ratings in that order.

But that's not what's happening.

Instead, only up to 5 games of a regular rating are used to initialize a player's first rating at a faster and/or online time control, while up to 10 games at a faster and/or online control are used to initialize a player's first regular rating, if the player gets rated in that order.

Limiting the number of games of regular rating used to initialize the other ratings, contributes to the often large gap between a player's ratings on the separate scales.

The rating-algorithm document recognizes that the ratings vary in quality, and says that the over-the-board regular rating is to be given the most weight.

Surely this should be done.

It would be ideal if USCF were to change the implementation of the rating system, to match what the rating-algorithm document says is done.

Failing that, at the very least the document should reflect what is actually being done: use only up to 5 games of a player's regular rating to initialize any other new rating, but use up to 10 games of any of these faster and/or online ratings to initialize the over-the-board regular rating of a player who hadn't got one yet.

Tom Worley

For example:


Player 2 in section 5 here had a fresh over-the-board-regular rating based on 9 games, only 5 games of which were used as his initial over-the-board-quick rating.


Conversely, all seven games of this player's online-blitz rating were used as his initial over-the-board-regular rating. 


Is this right?


Here's a player who was quick-unrated, but he had a fresh regular rating of 1523 based on 9 games, 5 games of which were used to initialize his quick rating. He drew one game at quick against a 1388. So his new quick rating, based on 6 games, should be something near


(5*1523 + 1*1388)/6 = 1500.5


Instead, his new quick rating is 1466 based on 6 games. 




Tom Worley


It appears that fewer games worth of the player's existing rating are used, even than the "based-on ____ games" numbers indicate:*1523+%2B+1*1388%29%2F%281.+%2B+x%29+%3D%3D+1466. 

That is, 1.37 games at 1523, averaged with 1 game at 1388, gives 1466.


(x*1523 + 1*1388)/(1. + x) == 1466.


is true for x equals approximately 1.37 games.


Not the 5 games that the new rating "based on 6 games" would have us expect.

Now a full 10 games of a player's existing over-the-board regular rating are used to initialize any other rating, as advertised. 

Thank you!!

FIDE recently adjusted its rapid and blitz ratings so as not to be lower than 200 points below the standard rating.


It would be good if USCF would do something similar. Often, quick ratings are many hundreds of points below the players' regular ratings.

Where all events are either regular-only or dual-rated, quick ratings never recover, once they lag behind.

Add new comment

Restricted HTML

  • Allowed HTML tags: <a href hreflang> <em> <strong> <cite> <blockquote cite> <code> <ul type> <ol start type> <li> <dl> <dt> <dd> <h2 id> <h3 id> <h4 id> <h5 id> <h6 id>
  • Lines and paragraphs break automatically.
  • Web page addresses and email addresses turn into links automatically.

Plain Text Comments