Timing issues for rating (and re-rating) events

The new ratings system will have something the current system lacks, the ability to schedule an event for re-rating if there are changes to it. The most common error is an incorrect ID, though we do get some reports of mis-reported results of individual games. (Based on the e-mail to the feedback addresses, at times I suspect that 5% of events have at least one ID error in them.)

This raises an interesting technical issue. If a tournament needs to be re-rated but some of the players who competed in it have already competed in other events, what should we do about any change to their rating in the re-rated event?

The logical answer is also the most complicated one: If an event has to be re-rated and there are followup events for those players, re-rate those events too!

This would allow us to do several things:

  1. We could rate events as soon as they pass all the validation steps. (See my lengthy post in another topic on this issue.) Such ratings would be PRELIMINARY.

  2. Right now the goal is to rate all events that have passed validation in a batch process once a week. That cuts down somewhat on the number of events that are rated out of chronological order, since within each batch we rate in ending-date (and section number) order.

  3. If we’re re-rating for other reasons, why not re-rate the entire file (back to some reasonable starting point) so that events are rated in chronological order?

Bill Goichberg and I have discussed this notion quite a bit, we feel it would be practical to re-rate events just ahead of a ratings supplement. (Bill actually wants me to go back and rate all of the events going back to 1992, but I see a lot of data quality problems with that.)