Game Plan Selection 1
by Mary Hickey - July 2006
Mary's articles are about "game plan selection".  Problems where you have truly reached a fork in the road and must make a strategic, rather than tactical, decision which will probably determine the course of the rest of the game.

Introduction to the new series

Welcome GammonLife!  I hope my column, and also those of the other columnists here, will help you reach your backgammon goals.  What, you don’t have any?  Well, some players like to win money and tournaments in exotic places like Paris, Monte Carlo, Costa Rica, and Columbus, Ohio.  (I had to throw that last one in because that’s where I play.)  And oh yeah, London, which is where GammonLife is based. 

Others would be content to become the unquestioned “owner” of the chouette at the coffee shop down at the corner.  Still others, probably Type Bs, play just for fun and the intellectual challenge of the game, but enjoy it more as they learn a few of its secrets here and there.  Backgammon has something for everyone, so set ‘em up, and let’s roll those dice!

Experienced players know it’s better to have a game plan than to fly blind from each roll and play to the next.  This is true even though the dice may force us to change it later on. Even a second-choice game plan well-executed is preferable to a hesitant, stumbling sort of plan, or none at all.  Ideal, of course, is to choose the best game plan in the first place, which is what we will try to do in this series of problems and their solutions.

Our bot-friends will help us, of course - for now I will usually roll out the positions at 2-ply with Snowie 3. GNU or JellyFish will often be consulted for a second opinion, and may occasionally be used for the rollouts instead of Snowie if it seems appropriate.  The minimum number of trials for Snowie rollouts will be 324, and in most cases there will be 648 or more.  If we haven't reached a clear decision at 972 or 1296, then the decision is too close for our purposes, and the problem will be sent to my "Cutting Room Floor" folder.  (I might find a use later for these close-call problems, but more likely I'll just let them gather cyber-cobwebs.) 

All rollouts will be full except for truncation at the database.  Truncated “mini-rollouts” will only be used for initial screening purposes, or for comparison problems where there’s a different roll to play, or a checker has been moved, etc.  These mini-rollouts will be 1-ply, 3888 trials, with truncation at depth 10.

Even if statistical significance is reached, no problems will be included in this series where logic plus rollouts don’t lead to a clear answer, meaning at least a .05 difference in standard equity terms.  If the technical decision is close, other considerations can easily supersede it.  These could include general game psychology, opponent-specific factors, or even your own comfort with the follow-up of one play versus the other.  In such cases it is also possible that the bots are making enough systemic errors after one play but not the other to cause a reversal when the next generation of bots appears.

All the problems for this series come from real games, either online or real life.  For the first columns, we will assume it is either money play or early in a long match.  Later, we may get into particular match scores, especially double match point (“DMP”) which is almost a game within a game.

Many of the problems will be examples of familiar, but perplexing, decisions such as run versus stay back, make a good point versus attack, or hold onto a backgame versus try to win going forward.

We will examine well-known principles, showing common applications and also situations where they may appear to apply, but in fact the lead concept is something else entirely.  I don’t think of such positions as “exceptions”, even though I slip and use that term sometimes, because what we call exceptions are actually situations we don’t fully understand.   It’s not that a good principle has been suspended, but rather that another one predominates.  Some of these principles may be as yet unknown to anyone - perhaps they are waiting for you to discover them! 

Problem 1:

Our first problem involves both running versus staying, and whether to try for another checker or commit to containing one already hit:

The decision to try for a second checker or commit to containing the one you already have is often difficult.  We will see more of these decisions in future columns.  This position permits you to commit to containment by running, 23/14, or stay back while making the useful 3 point.  If you skip ahead to the rollout at the end, you will see that making the 3 point is correct by a comfortable margin.  You will also see that the bot doesn’t offer an explanation of what is going on.  It’s up to us to ask the right “20 questions” to get Snowie to tell us not only what, but why.

The first piece of information we can discover is that this decision is not about racing per se.  If it were, then if the roll were 5-4, with the same number of pips, the decision would be close to the same, but it’s not.  A 5-4 allows you to make the better 4 point, but at the expense of an indirect shot and a weaker resulting builder distribution.  This comes in slightly behind running in a 3-ply and mini-rollout, but both plays are edged out by the game plan of radically committing to containment with 6/1*, 9/5.

Similarly, if the reason for making the point had only to do with the race, then if the blot were on the 21 point instead and the roll were 5-3, you’d probably make the 3 point despite the indirect shots, or at least it would be close.  However, running to the midpoint is clear (by .086 3-ply, .062 mini), even though several other options are available that allow you to try again for a second checker.  With 5-2, though, which can’t run cleanly but can make the 4 point very nicely, running becomes a triple-whoppin’ .321 blunder! 

So this is the second piece of information we can find here, and it’s a principle which arises often: When you have a play that runs your last man cleanly, it’s often time to say “Pips, schmips” and commit to containment of a single checker, even if you’re down in the race.  When “you can run, but you can’t hide”, that is, the run isn’t clean till your blot is missed in the outfield, it’s good to look for something better, particularly if racing isn’t necessarily a winning game plan even if you succeed in breaking contact.

There is a principle that Kit Woolsey has made famous, which advises: “When behind in the race, don’t race.”  Jake Jacobs tempers this with, “Don’t make a big you-know of yourself when you’re only a little behind!”  In this case, both ideas lead to the same play. 

The alternative to running makes a good point for the board or prime you will need to contain either one or two checkers, without taking excessive risk of being contained yourself.  These considerations make the point-making 9/3, 6/3 a clear winner.

Problem 1 with rollout:



Click here to return to Mary Hickey's section


Copyright © 2006-2011 by GammonLife.com. All rights reserved. This material may not be broadcast, published, re-written, or re-distributed in any form.