I was trying to work out how ratings are determined for tournaments and found out that there's an element of uncertainty that isn't obvious at first.
The basic problem is this: prior to any major tournament, sections and pre-event ratings must be determined based on some rating snapshot in time. Ratings themselves are based on rated games which should be rated in event order, but the results for rated games must be processed or submitted by tournament directors which can take anywhere from a few hours to several days or weeks depending on how they process and submit (electronically, or by mail).
The USCF approaches this problem by publishing monthly "rating supplements" which are official snapshots of every player's "concurrent" rating for events scheduled for that month. For example, the "December 2010" rating supplement is intended to be use for any events held in December. (Tournament directors have the ability, however, to choose to use earlier supplements if the desire).
For the December supplement to be ready for December, it obviously needs to be finalized before the month of December begins, so it is generally published shortly after the first Friday of November, aiming to capture all events completed in October.
The wrinkle is that some events played late in October may not actually be submitted in time to be reflected in the supplement. Players really can't be certain what goes into the supplement until it's actually published.
One point I am still confused about is how re-ratings are considered. It's my current understanding that each week, the USCF actually re-rates recent tournaments to properly take into account the chronological order of events and correct for the problem of receiving results out of chronological order. (Incidentally, the rating algorithm itself uses a two-pass system, meaning that a first pass is made to estimate each player's post event rating, and then a second pass is made to actually rate every player for that event). This process is impacted by any new incoming data for any player so I believe the re-ratings are done in batch and I assume they go back over some reasonable window and do a roll-forward re-rating of every player and every event.
In other words, eventually, your rating is always computed the "right" way, using the latest rated results for you and your opponent at the time the event was played, regardless of when those recent events results were submitted to the USCF.
But I believe the ratings supplements cannot be altered once they are published. I could be wrong about this, but historically official supplements were published in Chess Life Magazine so once they published that was it and I doubt that would have changed. So whatever is in the supplement is going to be used for that month's events, regardless of any more up to game data, even if those games were held before the date-cutoff for that supplement.
Anyway this probably doesn't have any major impact on the vast majority of players, but in some situations you may find that your rating used for an event isn't based on what you thought it would be.
Monday, November 8, 2010
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment