Playing up policy

This player played 4 games in his “first” (Rapid) event, 8 in his “second” (Blitz), and 6 in his “third” (Open).

His pre- and post-event ratings, however – both on his MSA tournament list and on each individual crosstable – appear as follows:

Rapid: unrated → 911 (P/4)
Blitz: unrated → 828 (P/12)
Open: unrated → 857 (P/16)

This makes it look (maybe) as though he played 4 games, 12 games, and 16 games, respectively, in these three events, and that his rating was computed from scratch each time.

But that’s not what happened. So apparently the (P/n) numbers are totals, not individual counts. But if that’s the case, why is the third one (P/16) rather than (P/18)? The player played 6 games in that event.

Mike Nolan sheds some possible light on this oddity when he suggests that, in cross-system initializations, the number of assumed pre-event games (used for calculation purposes) may be different from (probably less than) the actual number of pre-event games. That could explain why the third one is (P/16) rather than (P/18). We may never know exactly how the rating algorithm determines this downward adjustment.

So, Micah is apparently right when he says “This is incorrect Bill” when it comes to P/0, but apparently partial downward adjustments in (P/n) are occasionally made for various reasons.

Bill Smythe

Bill, if you look at the rating system document, glicko.net/ratings/rating.system.pdf, you will see that when a US Chess rating is initialized from other sources, the maximum number of games it can be based on is 10. This is why the regular rating is P/16 and and not P/18 after the event was rated.

One of the things we have had to explain to new players frequently is that the number after the slash in a provisional rating is not designed to be the number of games actually played in that ratings system. With the new blended initialization formula, this is especially true.

Aha! That explains it.

It is a bit weird to look at an MSA crosstable, or the ‘Tmnt. Hst’ tab on a player’s MSA page, and see the pre- and post-event ratings listed as

Unrated —> 1234 (P/16)

making it look as though he went from no rating at all to a rating based on 16 games in his first (6-round) tournament.

I wonder if it might be better to expand the pre-event rating the same way the post-event rating is expanded. For example, in this case the pre and post might be shown as something like

789 (P/10) —> 1234 (P/16)

Come to think of it, though, that might be even worse, as it would cause readers to ask “what 10-round regular-rated event did this player previously attend?”. Oh well, leave well enough alone, I guess.

Bill Smythe

Space is at a priority on the MSA crosstable, so it is challenging to show that a players is unrated and have space to show how the rating is initialized.

Moreover, for unrated players under 25 (ie, most new players), their first pre-event event is generally initialized using their age, which would make it possible to look at someone’s initialized rating and figure out their birthdate, at least to within a few days. That’s a data leak we would prefer not to have.

I wasn’t suggesting that age-based pre-event “ratings” be listed if a player is brand new. Those players could continue to be shown with a pre-event rating of simply Unrated. Only if the pre-event rating came from another rating system would you see something like 1234 (P/4).

But, as I said, it probably isn’t worth it. Also, I suppose even 1234 (P/4) might still, somehow, allow some reconstruction of age data.

Bill Smythe

Not really. You would only be able to back out the age from the first tournament (not from subsequent ones) and only if the results were such that there was a flat spot in the possible ratings that covered the age-based initialization (50 x age). (For instance, 100W, 1200L is flat between 500 and 800 so an 11 year old would get a 550/2, while a 12 year old would be 600/2). If all you saw was a rating of 550/2, it wouldn’t tell you much of anything about the age.

MSA doesn’t know how a rating was initialized, we’d have to add that field to the MSA tournament database, which may not be possible for older events. Even if it is, that could require reloading a lot of data, possibly back to 2005 when the current ratings programming and database took over.

And the question arises: Is this worth doing? Does it provide any really useful information for TDs or a player’s opponents?

This topic was automatically closed 730 days after the last reply. New replies are no longer allowed.