the dorbel daily

Monday, 13 February 2012

Backwoods v. zyxtcba part 4

Two questions left over from the last post and here's the first, with Black leading 6-4 to 13.

Black (zyxtcba) went for 20/10, encouraged by those two blots that White has in his homeboard no doubt. Some knowledgeable commenters also liked this, but it's a blunder. At present, this is a positional struggle, with Black holding a clear advantage. He has much the better anchor and all his checkers are in play. White on the other hand has four checkers already behind Black's 20pt anchor. These are useless for priming but can suddenly become useful if this turns into a hitting contest. Abandoning the anchor initiates exactly that sort of game. My own personal rule is, if it isn't obviously right to run, choose something else! So, what else? Some commenters and also XG++ like 13/7, 13/9, an ambitious attempt to build a winning prime at the cost of two blots. Old Skool masters liked this sort of play a lot, reasoning that if you own the 20pt, you can put your checkers where you want them and damn the torpedoes. It's actually still a very sound guideline and I can imagine a world class player making this play. The other contender is 13/3, less ambitious but it retains the midpoint and the anchor and forces White to quit his anchor if he wants to hit. Subtly it also activates the two spares on Black's 6pt to do something useful. I like this a lot, moving steadily towards a winning prime with little risk. A 2592 game 3-ply rollout also makes this a small but not definitive favourite. It's hard to spot like so many quiet strong plays. Humans are hard-wired to take positive action to deal with a problem now, but in backgammon, keeping all your options open and adopting a quieter approach is often as good or better. Erroll Flynn had the big style, much preferring to swing in to the banqueting hall on the chandelier, but one wonders whether walking quietly down the stairs wouldn't have served as well. An interesting and instructive position.

White (Backwoods) also had a roll that is commonly mishandled by beginners. Here it is.

Sadly Backwoods chose the beginner play, 23/22, 13/7 but hitting on the ace is absolutely clear. White must take advantage of his stronger board and go for the gammonish closeout. This is not an isolated case. It's worth taking a bigger risk than this to try for a closeout and here it is so important to stop Black making an anchor on the ace that White should probably hit there even if he knows that Black is about to roll a one! This also comes up a lot, well worth remembering.

Other news. We're into the last two weeks of Fibs League session 55 and both Master A and Master B are coming nicely to the boil. In Master A your correspondent has 8-3 with one match to play, but everybody else has played fewer matches and nine other players can still put a run together and equal or surpass this score. In Master B Germany's runnerup is 8-4, also with one match to play (Master B has 14 players in this session) but again, most of the division will still be hoping for a burst of form to catch or pass that.
The 2011 Master Playoffs kicked off, with Schigolch (Germany) winning his first match with Mano (France). All these ties are best of three 9 pointers.

A word about rollouts. This blog attracts some mild criticism because I don't post the results of rollouts. It may be useful for you to know that I use XG for this, with 3-ply checker play. It's possible, likely even, that a higher level of play can produce a different result, but you do have to remember that players who can play as well as XG 3-ply are few and far between, so by using more advanced settings, you move into areas where no human can play well enough to achieve the theoretical equity. As the differences usually amount to a few hundreths of a point of equity, the value of a more accurate answer is limited anyway. I've also found that posting rollout results is boring and counter-productive, as they stifle debate. I would rather listen to our correspondents discussing why they would make a play any day, even if they are wrong!! Backgammon isn't about knowing what the right answer is, it's about knowing why the right answer is right, so that when in future we meet something like it, we have something to base a play on.
Anybody can do rollouts and I am always pleased to hear from somebody who has rolled out a position from here and got a different answer, but I won't be posting mine.

I'll have some more action from Backwoods v. zyxtcba tomorrow, perhaps even later today, so stay tuned. Until then, enjoy the game!


Timothy Chow said...

Paul, I understand your point about how rollout results stifle debates. However, it strikes me that you're not consistently following your own philosophy. You do, in fact, post rollout results, because you say, for example, that "a 2592 game 3-ply rollout also makes this a small but not definitive favourite." What I notice is that you just post verbal summaries like that, without posting the full information provided by the bot. I find that, if anything, posting a summary like that stifles debate more than posting the full rollout information. Often, when someone posts the full output, I'll notice certain things about it, such as the standard deviations, or the pseudocubeless win and gammon numbers, that provoke further questions and ideas for further analysis of the position. If you just say that 13/3 is a "small but not definitive favorite," there's nothing to stimulate further thought (how small? how definitive? because it wins more games or because it loses fewer gammons? etc.)

I'd recommend that if you are serious about not posting rollout results, then you also refrain from mentioning the results of any rollouts as well. That is the philosophically consistent thing to do.

dorbel said...

Thanks Timothy. I appreciate your comments and suggestions always and I quite understand the points that you are trying to make. My approach to rollout results may not be consistent, but it is very important for me to write in my own way, a way that reflects my particular approach to the game. The minutiae of rollouts is not interesting to me and those who do find it helpful can do their own rollouts with their own settings.
"Not definitive" in this particular position means that 13/3 is ahead by 1/100th of a point, but with confidence intervals that do not make this a definite result. To all intents and purposes, the two plays are equal. The really interesting point is that it might be right to select 13/3 if playing a master (in order to keep the game as simple as possible) but go for 13/7, 13/10 if playing a weaker player (in order to seek complication).
I have thought about this a lot. I can post full rollout results, I can post IDs for Gnu and XG, I can post the pipcount, but the more work I do for the reader, the less he or she will engage with the position. Things thatyou find out for yourself are much more likely to stay in your head than things that one is told.