Do we want to blitz, or build an outside prime? Neither decision feels all that great, since the blitz goes deeper than we prefer, but the prime we can make is an “outside prime”, notoriously difficult to work with especially when all its back points are stripped.
If we choose to blitz by playing 7/3*(2), the fourth 4 should be played 13/9. We need more wood in the Zone to complete an attack, and this brings another builder to bear on the points we want most to make. It also "won" the evaluation and the Mini (mini-rollout shown at the end of this article), by .034 over the blitz with 24/20 for the last one.
Similarly, for the priming approach we are dropping out 24/21, 13/9(3) because it decisively lost the Mini (-.124) and was also fourth in the Snowie 3-ply evaluation. Only reason we even considered it was that JellyFish gave it second place, though it still rated it a blunder. The priming game plan will be represented by bar/21, 24/20, 13/9(2).
The attacking play is unsatisfying, forcing you to lead off by making “the wrong point, with the wrong checkers, at the wrong time”, as you might express it in a debate with your team captain. It leaves you with a stacked 6 point, a stripped 8 point, no bar, and three blots lying about, making continuation difficult. Its main benefit is taking away half your opponent’s roll, which of course isn’t a bad thought.
The priming play also does nothing about that stack on the 6 point, but at least it brings two new checkers into the Zone, gives you four in a row, and doesn’t create any new blots. The last 4 played 24/20 is actually a strong move, since it reduces the value of that extra builder on White’s 3 point - it becomes what is sometimes called a "dilly builder", because it has nothing to attack and can only be used to make deep points at a time when that’s not his priority.
The rollout below shows that the priming play is clearly preferable here. In backgammon it's often right to be aggressive - but not when you can't back it up. It's psychologically harder to just sit there and let him do what he wants, when you have a chance to make the first strike. Here, though, it’s better to amass the material you need rather than attack too early. Playing passively is wrong when you had a legitimate opportunity to “bring the game to your opponent”, but sometimes the most prudent game plan is to prepare first, attack later.
As we noted last time, priming is usually more effective against two or more checkers, since it’s harder for them to make a fast getaway. In this case, even an outside prime works pretty well against his two back men.
In contrast to last time, the Snowie evaluations show an increasing difference ending at .076 on 3-ply. This concurs with JellyFish's L7 match score 0-2/5 evaluations, which showed an increasing difference of .040/.107/.1135 in favor of priming even when you "need a gammon". Last time we noted that a decreasing or unsteady difference means you need a rollout. Conversely, these increasing differences mean you probably don’t. A steadily increasing difference usually means the bot knows what it's talking about!
If you want to make optimal use of your rollout resources, in terms of both computer time and your available waking hours, you will seldom go wrong by just accepting the 3-ply answer in these cases. You can then devote the time you and your computer have for backgammon to rolling out something more questionable. The rollout below shows that you wouldn’t have gone wrong by just believing the evaluations here.
Problem 1 with rollout: