by Jake Jacobs
10 October 2012
Rick Janowski has been compiling error ratings for backgammon players from the pre-bot era. Digging through the archives, exploring dark chambers full of cobwebs and grinning skeletons on my hard drive, I unearthed three dozen of my own matches that range from ten to twenty years old. Much like old wine, if not sweet to begin with, there is a good chance that they will turn out to be vinegar now. The first one I put in XG gave me a thrill; during the two days it was being analyzed - I used rigorous settings, and it was a 23-point match - it showed that I had played 1.3, while my opponent was on the shady side of 25. That's when I learned that it is the summary of the first game that teases you during the rollout; I scored 6.0 that match, and my opponent came all the way back to 9.3.
|The rest of this article (10.05 K) is premium content. Please subscribe below.|
Article text Copyright © 1999-2013 Jake Jacobs and GammonVillage Inc.
Excellent article, Jake; one of your best!
I think that there are a few typos, though. Your 52, when trailing 3-5, can't legally be played 23/17. I would have played 24/22 23/18 here, but I'm guessing you're saying that 23/16 is best? Next, when trailing 4-6, the text description suggests that your diagram should either show a roll of 32, or show the checker configuration a turn later. Finally, in the last diagram, I think you mean to say that running cuts your gammon losses down to 70%.
By the way, how bad is it to play 24/23 24/22 in your "Split to the 21pt?" position?
Oh, and I'm very familiar with the Distracted rating, thank you very much! It's not all that hard to earn it with a cube error, given how XG calculates cube performance.
Studying the final position again, I now think that you meant to say, "My winning chances are tiny if I run."
No wait, it's still not making sense to me. Maybe it's, "My winning chances are lousy if I stay, but my chances of saving the gammon if I run are really lousy"?
Let's see ...
Yes, it should be 23/18 with the five, your play. Not really "once again ..."
The diagram with 64 originally had no dice roll to play, but the paragraph below had a typo, "13/6" in stead of "13/9," and while I was fixing that, I looked up at the diagram without rereading the entire paragraph, and ... you may guess the rest.
The split was .053 worse, but XG only rolled out its play versus mine, so a new rollout might narrow that gap.
Speaking of distracted cube errors, I just returned from Japan (expect an article next month). One of my opponents recorded our match, and sent the file overnight. I earned a respectable 4.66 despite one horrible game. Overall I made ten errors, two blunders with the checkers, and four blunders with the cube. Five of the blunders came in one game! I failed to split my back checkers in a last shot scenario, a position that I can and should calculate precisely. I was lucky enough to hit anyway, but then ignored the fact that he had twelve checkers on his ace, and when he entered immediately on my ace I missed four consecutive double outs.
Oops, one more typo. The winning chances are lousy if I stay, but my chances of saving the gammon are lousy if I run. The winning chances if I run are worse than "lousy," even worse than "distracted." :-) They are zero.
It's fun to hear about how the best players were playing years ago compared to today, and you and Frank were certainly two of the best back then. But what about the very best...people like, say Magriel and Malcolm, Kit and Robertie and Sly and Snelling--the better players in the world at that time. If we put their matches into XG how would they rate? Any idea?
Phil, Rick Janowski has been working on that question. Without looking it up, I think he found Senk playing around 5.5 pre-1990, making him the best of the best, and in the early 1990s Wil was in the high 4s, and Howard Ring was close. Of the ones you mention Kit and Sly had very good ratings for those eras. Magriel was probably the best pre- 1980, and was always good. Rick's findings were based on the number of decisions, with some players making more than 1000, some between 500 and 1000, and some fewer than 500 decisions. If you look at the number of matches analyzed I think fewer than 10, and maybe as few as five got you over 1000 decisions. Compilers of match stats, Iancho Hristov is currently the go to guy for this, use 50 recent matches for their findings, subject to availability, so a set of five matches played over a span of years may not accurately reflect a player's strength. What is clear is that when the bots arrived, ratings improved quickly.
You must be signed in to post comments.