| |
BGonline.org Forums
The FIBS rating formula : why sqrt(n) ?
Posted By: Jake Jacobs In Response To: The FIBS rating formula : why sqrt(n) ? (Fabrice Liardet)
Date: Thursday, 15 September 2011, at 3:06 a.m.
Not to be picky, but Walter's contributions to Fish (which were many and significant) did not include writing any of it, not even his bio.
At the time Fish was written Kaufman's adaptation of the chess formula was what KG used for his ratings, and what FIBS used for its. It was clear that Kaufman's K factor was a hack, but a reasonable one. That it is flawed is obvious. The most clearcut example occurs with 2-point matches, a flaw that I gather players often exploited to boost their ratings.
The K factor for a 1-point match is 1, and for a 2-point match it is 1.414. If you double immediately in a 2-point match you are playing a 1-point match, but are credited as though still playing a 2-point match.
(That may have been fixed years ago. I don't play online, so don't pay attention to what the various servers are up to.)
Most of the work on the book was done in 1995, and this was all new. My gut feeling then, looking at the online ratings of players I knew, and could judge, was that the rating system was far from perfect. Walter's feeling at the time was that the math was the math, but conceded that the ratings seemed inflated. At least, he had no problem with my including comments to that effect. Of course Walter was a chess player, and perhaps brought confidence in the chess ratings to the backgammon table.
I used to take online rating difference between unknown players, and cut them in half when I wanted to handicap them. That is also a hack, but Tom Meyer made some money on auctions using my input.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.