| |
BGonline.org Forums
OTB and short rollout
Posted By: Nack Ballard In Response To: OTB and short rollout (Nack Ballard)
Date: Wednesday, 16 December 2009, at 7:28 p.m.
Correction of typo: Under (1) in my previous post, I accidentally left out a "not." I meant to write, "I mentioned there were a couple of exceptions to the useful go-with-distribution guideline... 41S-65R-43 may be one of those exceptions (and it would not be surprising, as Opp rolled her biggest number that makes no offensive point)."
It turns out that Paul Weaver had sent Gregg Cattanach copy of his excellent article Which Five Point is Better? long ago, and Gregg was kind enough to forward it to me this morning. (I've forwarded it back to Paul, in case he has misplaced his own copy.)
As it pertains to this thread, for 41S-xxx-43 (for money), Blue should avoid the offensive two-five distribution and make his 20pt, unless in the obvious case that White has split and is now directly threatening Blue's 9pt builder. Beyond that, 51S is the only clear reply exception (where Blue should make his 5pt by .026).
According to Paul's truncated rollouts, half-exceptions for the "xxx" referred to above are 21S, 65R and 64R (make the 5pt by .005, .004, and .001, respectively). I call them "half-exceptions" because I deem a margin within .01 to be a tie.
For early game positions, there is a lot more variance in a full 1k (1296 trials) than people realize. I've seen independent rollouts of full 1k's of the same position with margins .05 apart! My perspective will no doubt surprise many, but I've collected tons of data and based on my observations, I put as much faith in a 1k trunc as I do in a 5k full.
Without further data (which of course is always good to have), I average Tim's 41S-65R-11 [P @11] 1k full, with Paul's [P @4] 1k trunc, weighting towards the latter, and call value [P @6]. In other words, based on the data currently collected, the choice of making the 5pt and anchoring on the 20pt are within .01 and essentially tied for this position.
Paul also shared data for DMP, GS and GG. His only tool at the time was Snowie, so the checkers aren't played according to score, but most of the margins are large enough to trust the conclusions. As you might expect, for both 41S-xxx-43 and 43Z-xxx-41, one tends to make the 20pt at GS and the 5pt at GG, but not always.
There are many more details revealed in Paul's fine article, and although he will probably want to update the rollouts, I hope some day he gets around to publishing it.
Nack
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.