[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Nacbracs

Posted By: Nack Ballard
Date: Thursday, 24 December 2009, at 11:01 p.m.

In Response To: How do you nactate..? (Matt Ryder)

Thanks for your kind and enthusiastic feedback, Matt. The more interest that is shown in Nactation, the more motivated I'll feel to complete the next update to the tutorial. (It is underway, but I tend to work in spurts.)

I am particularly curious about the "shorthand equity nactation" you mention. Have you developed this further?

By that, I was referring to Nactation brackets, or "nacbracs," a word you may have seen tossed around, so far by David Rockwell, Stick, myself and maybe MCG (apology to anyone I overlooked). Nacbracs are a condensed way to communicate rollout results, either from one rollout or a combination of rollouts averaged.

In nacbracs, the plays are listed in order of rollout strength, and after each letter/character the relative margin is shown in thousandths of a point. For example, Gnu's rollout (whether one trusts it or not) for the 21 reply to opening 51S is

51S-21 [\$ U3 D28 S34].

This means that \$ rolls out best, U has a margin (relative equity) of -.003, D of -.028, and S of -.034. (For this example, and in my others in this post, it is redundant/unnecessary to add "51S-21" before the brackets if it is stated close by in your text.)

You will often see a number after the bracket that refers to the number of trials in thousands rounded down. Most commonly, it is a multiple of 1296. For example, "5k" means some number of trials between 5000 and 5999 (the person in possession of the rollout knows exactly) but almost always 5184. Likewise, 10k means between 10000 and 10999, but usually 10368. And 15k usually means 15552, etc. Note that above 25k (25920), adding another 5184 becomes 31k (31104), which is why you rarely see the oddity "30k," and for the same reason you see 46k discussed instead of 45k, etc. (5k x 9 = 46k).

If not all plays are rolled to the same number of trials, the largest number of trials is listed first, followed by exceptions -- the smaller number of trials and its play(s). In the example rollout, \$ and S were each rolled out 46k, and D and S were each rolled out 5k. This information is condensed to "46k 5kDS." It is possible to omit the number of trials, though I usually don't unless all close plays listed have been rolled out at least to 15-20k for money or to 10k for cubeless scores (otherwise it can be misleading, for variance reasons).

Adding the number-of-trial information yields either of the two sample nacbracs shown below. The first line represents two plays, the second line four plays.

51S-21 [\$ S3] 46k
51S-21 [\$ U3 D28 S34] 46k 5kDS

[When I'm listing rollout data en masse (and not in bot columns), I add the symbol # before the number of trials in a Gnu rollout, and ^ before the number of trials in an XG rollout. When neither symbol is present, it's a Snowie rollout. For example, the nacbrac string "[U \$14 S20 D32] 20k 5kD, [\$ U3 D28 S34] #46k 5kDS, [U \$12 S23 D30] ^36k 9kDS" contains the 51S-21 rollout results of all three bots. (This can be condensed further to "[U \$14 S20 D32, #-3 34 28, ^12 23 30] 20**5, 46*5*, 36*9*," but it's better not to use up space explaining that here.) In this forum, I recommend stating the name of the bot and using a distinct bracket for each, but I thought you might be interested in some shortcuts for personal use.]

Another useful set of abbreviations is d, s and g, representing double match point, gammon save and gammon go (commonly referred to as DMP, GS and GG). The lower-case letter appears prior to the best play.

There are two methods of insertion. The simple way is to place the d, s or g before the brackets; for example,

51S-21 d[\$ U10 S22] 46k 5kDS

... is Gnu's rollout at DMP. (For added clarity, you can use "m" for money, but if d, s or g does not appear, money is assumed.)

The other way to insert d, s and/or g (usually when there is more than one) is inside the brackets. For example, to represent Gnu rollouts at all four scores for the reply roll 43 to opening 53, you would write

53P-43 [D Z5 U22 S22, dZ D1 S6 U6, sS D3 Z16 U42, gD Z20] 46k 5kUS, d10k 5kSU, s5k, g5k.

[This can be condensed further to "[D Z5 U22 S22, d-1 6 6, sS 3 16 42, g20] 46*5*, 10*5*, 5, 5," in case you want to glean the meaning.]

Finally, if you want to report accuracy to ten-thousandths instead of thousandths, insert a decimal point. Repeating Gnu's money rollout with the added accuracy yields

53P-43 [D Z4.7 U21.6 S22.3]

Accuracy of margins beyond some threshold (10, 15 or maybe 20) is less important. Therefore, even if you want to show a more accurate margin for that of Z, you probably won't want to do so for U or S. Thus,

53P-43 [D Z4.7 U22 S22]

If Z's margin happens instead to be 5.0, you can list it that way or simply as "5" as long as the recipient of your communication knows that you have a convention of decimalizing under a certain threshold.

Nack

Post Response

Subject:
Message: