| |
BGonline.org Forums
Opening 32Z: reply errors
Posted By: Nack Ballard In Response To: Opening 32Z: reply errors (David Rockwell)
Date: Saturday, 2 January 2010, at 2:05 a.m.
I feel that the best time to play 32Z is at DMP. I have [S D4 Z6] - 20k which is probably the rollout on Stick's site. Since DMP includes many scores which are not strictly DMP, there are quite a few chances for this.
Yes, [S D4.4 Z6.3] 20k. You'll see my reason for adding a decimal point in a moment.
Errors at DMP mean more than they do for Money. I'm not referring to MWC; I'm going purely by an empirical testing of EMG error sizes / margins of the closest few plays (as reported by rollouts) for DMP, GS and GG and Money. The ratio is roughly 9:14:14:20. The relationship of the last three make sense to me because GS and GG are cubeless scores while money is cubeful; why the 9 for DMP is lower than 14 I'm not sure but the sample size was large enough that it's at least in the ball park. Just compare a smattering of your own rollouts at the various scores and you'll see what I mean.
Using the 9:20 ratio, an error of .063 at DMP occurs as often as an error of .063 x 20/9 = .014 for money, which happens to be the margin I reported in my last post by which 32Z is weaker than 32S. (Actually, it would have been .0143 to another decimal place.)
My point is that (as far as I can tell) the chance of 32Z being a better move than 32S due to Opp's greater errors is just as great for money as it is at DMP.
Btw, 32Z at GS is [S Z11.8] (averaging your rollout and another, though their margins closely agreed). Multiplying the S/Z margin by 20/14 yields [S Z16.9]. That's higher than 14, so according to my 9:14:14:20 ratio, 32Z is a bit more out of reach at GS than it is for DMP and Money.
I think that 32S gets the fewest mistaken replies. The only one which is really tough at all is 32S-62 [H X21] - 15k. And, X isn't much of a mistake. I've never seen anyone play H here.
Which rollout is that? That's close to Gnu's GG result of [H X20] 5k. For 32S-62 money, I have Snowie [X S8 H14] 31k, Gnu [H X2 S22] 46k, and XG [X H1] 15k, averaging [X H4 S14].
However, that narrower X/H margin only reinforces your main point: that 32S gets the fewest mistaken replies. I opined the same, and I believe the reason 32S is easy for everyone to play against is that they've seen the right replies made repeatedly by bots or stronger players. Otherwise they might not play 41X because it looks ugly, or they might play 63R, 54R, 43S and 52S to duplicate 3s, or they might play 32K, 43K and 44A (like they used to).
32Z might or might not be harder to play against than 32S objectively -- let's say hypothetically by someone who retains all his backgammon talent except that he suddenly forgets all the rote replies he has learned -- from that perspective it's probably about a toss-up. The benefit of 32Z lies in replacing an entirely familiar position with an unfamiliar one and thereby inducing immediate errors that wouldn't otherwise be made. I think you agree with that; I'm just clarifying for others.
There aren't a lot of mistaken replies at 32D either. But, almost everyone plays 33B which is a sizable error. I also see things like 21S and 52D that add to a little more. Overall, I expect to see around 25% of the errors that 32Z generates.
I estimate 15-20%. We're on the same page.
Count me as making a "mistake" on 32Z-21 based on your Nacbrac. I'm not going to let my opponent make points with that builder on the 10 point. I'm going to play H rather than S. I'll play 21S after 52S, but not here.
Oops. I wrote [S H9 V40 U54 D79] from the Snowie data, but I just noticed Gnu data hiding on the page after, which is [H S7]. The average is [S H1], which is about as close as it gets. To anyone for whom it is relevant: please correct that in the table of my previous post: 32Z-21 [S H1 V40 U54 D79]>
I don't think you have seen the rollout below yet Nack. It is Supremo 46k. The gap between S and the others is a little wider than the MWC 46k posted on this site.
[S D8 Z14.7 $46 U47] 46k -- this is new, thanks! Z remains about the same (from -.0143 to -.0145 in the average), your D worsens Gnu's average so bot average goes from -.0103 to -.0107 or so. The average of U goes from -.042 to -.045, and $ from -.064 to -.053.
Stick might bray if he sees you rolled out $ and U to 46k, but we don't have to tell him.
(And, I don't use 1 for the seed any more either! This was started long ago.)
Glad for the assurance. This time you kept me from braying.
Nack
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.