Clarification on game counts for provisional ratings

There continue to be questions as to how the game counts for provisional ratings are determined when a rating is initialized using other ratings information. The Ratings Committee has approved the following explanation:

For US Chess players who have a provisional rating (normally those who have completed 25 or fewer games), the “based-on” value connected to their rating is not always the number of games they have played; it is the “adjusted” number of games which is usually larger than the actual number of games.

The only situation in which the adjusted number of games is the same as the actual number of games played is when the player’s initial rating was determined solely based on their age. In every other situation, i.e., when players’ initial ratings are derived from ratings converted from other systems (including other US Chess systems), the adjusted number of games is the sum of the actual number of games played and a value that reflects the information used to initialize a player’s rating. That value, which can never be larger than 10, depends upon the possibly multiple sources of converted ratings (the other US Chess online or over-the-board systems, FIDE, or CFC) to obtain an initialized rating, and the staleness of the converted rating.

However, after a first tournament, the adjusted number of games increases only by the actual number of games completed in subsequent events. The details of the calculation for the adjusted number of games can be found in section 2 of the US Chess rating system specifications:

1 Like

Some historical background may be useful.

Prior to the introduction of the blended ratings initialization process several years ago, most ratings initialized from other US Chess ratings information had an initial game count of 0 while ratings initialized from FIDE ratings had a game count of 5 or 10, based on the player’s FIDE rating.

Under blended ratings, any rating initialized from other ratings information (FIDE, CFC or US Chess) will have an adjusted game count between 1 and 10.

As a result, we are seeing more questions about why someone’s provisional rating game count is different from their number of competed games in that ratings system.

Whether this will reduce the number of questions raised about this issue is uncertain.

Links to this Forums topic have been added to several pages on MSA, including the MSA FAQ.

1 Like

Let’s say a player has their US Chess OTB regular rating initialized from the other ratings they have and the game count they get from that initialization is five. Usually a provisional rating based on 5 games is governed by the “special” rating formula but is it governed by the special rating formula in this case?

As long as the game count is 8 or lower, the special formula applies.

The Special Formula exists because the Standard Formula can get a bit funky when N’ (from whatever source) is small. The Standard Formula does one step in a Newton’s method procedure, where the first step can be a doozy when N’ is small.

Note that It’s possible to have N’<=8 when the player has more than 8 credited games if the rating is very low, but it can only be barely less than 8 even for R=100, and a possible overshoot isn’t a serious issue for someone with such a low rating.

I see the link to this forum topic in this version of the MSA FAQ US Chess MSA - Main Page but not this version Frequently Asked Questions: Member Services Area | US

In the second link, the question “I just played in my first Blitz, Quick or dual-rated chess event? Why doesn’t my tournament record indicate I was previously Unrated in quick chess?” Should be revised due to the new “blended” rating process.

Is it an intended outcome that a unrated player who plays no rated games in their first tournament can nonetheless have an initialized rating? Originally I cited this player but they anomalously appear twice in the cross-table for the only tournament in which that player has played so far.

My expectation would be that a player who plays no rated games in a tournament should not have any rating change (even initialization from other rating systems).

I’m not sure what’s going on here, it is possible the duplicated event is confusing things.

I don’t have access to the ratings server right now because my laptop died on Friday, so I can’t look at the trace logs to see what’s going on. I should be back online by Tuesday.

The second entry is a forfeit loss and no other games. There was either some error with R1 pairings or the player showed a bit late and they did a re-pair. At any rate, the second entry has no rateable games, so it has no effect on the rating.

It should have combined the games from both pairing numbers under the same ID, maybe the fact that there weren’t any ratable games in the second pairing number confused it (and me.)

This may be the first time we’ve seen this. I doubt we’ll try to correct the current programming, but we may add this to the test events for the new ratings system.

Didn’t it do exactly what it’s supposed to? Both pairing numbers are showing the xxx/5 from the combined entries.

Hmm, yeah I think you’re right.

Yes, this is what I discovered once I found the duplicate entry.

Would be good to have a test case to check this scenario:

  1. New player appears in the cross-table for a four-round quick tournament but plays no rated games.
  2. Same player plays 4 rated games at regular rating (only).
  3. Same player then plays 4 quick-rated games.

The player should now have a provisional rating based on 8 games (not 4).

Again, I can’t readily find any examples to determine if the current system matches the test scenario described above.

I believe Tom is correct that nothing is wrong at this time, this player had a second pairing number, for whatever reason, one of which had 5 ratable games and the other had none, so the same provisional rating based on 5 games is appropriate for both pairing numbers.

1 Like