|
BGonline.org Forums
Gammon rate in $ Games?
Posted By: Jake Jacobs In Response To: Gammon rate in $ Games? (Albert Steg)
Date: Thursday, 9 July 2015, at 2:19 a.m.
It's an interesting question.
Danny Kleinman published the first widely used MET in 1980, and used 20% as the gammon rate. When Heinrich and Woolsey used Hal's database of matches a decade later they used 26%. Roy Friedman iconoclastically claimed the rate was 36%! When I wrote FISH I set the gammon rate for an even match at 24% based on not much more than what others had done and the seat of my own pants. (I was tempted to go a bit lower, having used Danny's table for so many years, but Walter leaned toward Hal's and Kit's 26%, so 24% was a compromise.) And then for each 50 Elo I gave the stronger player 1% more game wins, and 1% more gammon wins, and reduced the fish by the same. While there was no empirical data supporting any of this, those numbers, besides being rather neat, required a minimum of additional tweaking to produce tables that met the requirements across scores. In other words: they worked. Today XG seems to think the gammon rate is 28%, so Hal and Kit's historic match data was what humans were achieving back then; the bot being a better player does slightly better.
The rate is derived from games always played to conclusion (I've always assumed), and that shouldn't change too much from match to money, the gammon save and gammon go scores should tend to balance each other.
So your question comes down to this: what is the gammon rate for games that may or may not end with a cube turn, or perhaps we could say the actual results of human play? If 1/3 of games (per Stick below) end in cube turns and the gammon rate is 28% per XG, then 18.67% off all games should end in gammons, but does that figure differ for match and money? One would expect that players play on for gammons more often absent Jacoby. But at the same time they pass gammonish games more often because of the score.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.