“I’d rather be lucky than good.” — Lefty Gomez

I think of backgammon as similar to poker, in that it combines generous measures of luck and skill, and even a little bit of bluffing at times. That fact has some interesting ramifications that make the game both complex and unpredictable at times. For example, unlike in chess where strategy and skill rule the game, in backgammon on any given day, a complete beginner may beat a world-class expert. That is the nature of backgammon; with a few lucky rolls at a critical time, you can pin your opponent to the bar and practically coast to the finish line!  Yes, dumb luck can prevail over skill, at least in the short run.

That luck sword tends to cut both ways though, and on average, you are about as likely to have “good” luck as you are to have “bad” luck.  So, over the long haul, what makes some players better than others?  Do they find some magical way to “get luckier” than the rest?  Why is it that when we are winning, we perceive it as skill, but when the opponent is winning it is perceived as luck?

I think that game skill actually works to make you luckier in backgammon.  That’s right, as we become more skillful players, we also become luckier players.  A beginning player will tend to make very “safe” moves that avoid getting hit, but may result in checker positions that will be vulnerable on future rolls, or that are inflexible when trying to build new points.  A more experienced player will be better able to evaluate the probabilities of success with more risky and aggressive moves.  These moves often leave him more vulnerable to getting sent to the bar, but yield board positions that may offer much more versatility to take advantage of future rolls.  So there is a potential risk vs. potential reward that must be assessed and understood.

When analyzing games with a computer tool such as GNU, it will rank possible moves according to the game “equity” they produce.  Often times I find the move that I made was not the “best” move, at least according to GNU.  In fact at times, GNU finds moves that I didn’t even consider, leaving me wondering “Why would GNU make that move?”  Since the computer program has the ability to perform thousands (or even millions) of roll-out predictions, it tries to choose the move that has the highest probability of yielding the best “equity” going forward.  Often that means making somewhat risky moves in order to cover more points that allow you more opportunities to block or capture your opponent on future rolls.  There is always the chance that the risk won’t pay off, but the probability is that you will win more games.

Novice players tend to make more conservative moves that they view as “safe”, rather than risk getting sent to the bar, when possible.  The expert player may at times leave open checkers scattered all over the board, in order to improve the probability of making a prime or capturing the opponent’s checkers.  This tends to lead to more interesting games with a broader range of outcomes.  That is, the expert player tends to win more games than the novice, but may also get gammoned (or even backgammoned) more often than the novice player.

So as you play, or as you analyze your games after the fact, be observant of where you might be able to make moves that position your checkers most effectively to take advantage of future rolls.  One example might be where you get a high doubles (fives or sixes) roll.  Normally considered to be “good” rolls for moving your checkers around the board quickly, they can be wasted if your checkers are all piled up on “safe” points that happen to be blocked by your opponent five (or six) points away.  Some forethought might have prevented that inflexible checker positioning, giving you a better chance to take advantage of a fortuitous double-six roll.

Similarly, as you are approaching the point of moving your checkers into your home or “inner table” while you opponent still has anchor points there, think about how high doubles on your next roll would affect your positioning.  Would double sixes force you to move checkers in such a way that you end up leaving blots open and vulnerable?  This would be viewed as an “unlucky” roll, that could cost you the game if your opponent then hits your blots.  Positioned a bit differently, that same double-six roll could be a very “lucky” roll that propels you to to an easy gammon.   A little planning goes a long way to reducing that probability.

This is an introductory article to GNU Backgammon, a free, open-source software (FOSS) backgammon program that is a World-Class opponent, a live Tutor, and an analysis tool.  You can get the program by following the link in the side bar menu.  This is a very powerful program with many, many features.  In this introduction, I will show how you can use the analysis tool to review your games; both GNU games, and games that you import from other sites.

The entire subject of computer analysis of backgammon positions, moves, doubling cube decisions, games and matches, is far too complex to explain in detail here.  It is based on “neural networks” that simulate thousands (or even millions) of potential outcomes, through a “roll-out” process, where the computer actually evaluates all of the possibilities for the next move, or even the next several moves.  From all of those possibilities, it assigns an “equity” value to the various possible decisions that can be made for the current situation.  With that information, it can then determine not only what is the “best” move, but it can actually tell you how much better it is than other possible moves.  Using the “Hint” tool, you can even check your move against the computers equity table to see how good (or bad!) of a move you are about to make.

GNU has a Tutor Mode that automates this process somewhat.  With Tutor Mode active, GNU will actually give you pop-up warnings to let you know when you are about to make a sub-optimum move.  You can set the thresholds that  GNU uses to classify moves as “doubtful”, “bad” or “very bad”.  As a beginning player, you may want to set these thresholds loosely, and then tighten them up as you get better.  GNU has so many customizable settings, that you can fine-tune it to your liking.  The computer’s playing skill level can be adjusted down to match your skill level, giving you more of a chance of winning occasionally.  Other things like the basic appearance of the board colors can all be customized too.

GNU Backgammon Screen

After playing a game (or match) against GNU, you can run the analyzer.  This will show you the history of all the moves in the game(s) and will highlight moves where sub-optimal moves or doubling cube decisions were made.  Here is an example of what the analysis for one recent game looked like:

Game Analysis

That certainly is a lot of numbers!  Let’s decipher it a bit.  At the start of each row is the move number, starting with zero.  The next group of numbers is Player A’s move, followed by Player B’s move.  For each move, it shows the dice roll, and how it was played.  For example, on Move zero, Player A shows “64: 8/2 6/2″.  That means that Player A rolled a six and a four, and moved one checker from point number eight to point number two, and moved the other checker from point six to point two.  In the actual game analysis, when you click on a move to select it, you also see the board diagram next to the analysis, making it much easier to  visualize the moves.

Notice also that some of the moves are highlighted with either colored font, colored background, or bold font.  These are all highlighting special information about that move.  When you click on each move to select it, an additional information box shows up at the bottom, giving the details on that move.  Let’s look at a few examples. In the move list above, Player A’s move 1 is marked in green font.  Let’s click on that move to select it for further info.



Move 1: Player A makes a "Doubtful" move

In this diagram, Move 1 for Player A is selected in the move list, moving one checker from the 13 point to the 8 point.  Below the move list is the info box that shows the details.  Note that the analysis box has two tabs: one for Checker Play, and one for the Cube Decision.

On this play, the move is marked “doubtful” and shows up fifth on the list of possible moves.  From an equity perspective, this move was 0.22% worse than the number one move, which would have been to move one checker from the 13 point to the 9 point, and another checker from the 6 point to the 5 point.



Move 1: Player B makes a "Bad" move

In this diagram, Move 1 for Player B is highlighted.  As the info box at the bottom shows, this was marked as a “Bad” move, 0.42% worse than the preferred move.

Next, let’s look at Player B’s move 19, which is marked in red.



Move 19: Player B makes a "Very Bad" move

Player B’s move 19 is selected in this diagram.  The GNU analysis marked this move as “Very Bad” at 1.59% worse than the preferred move!

So the checker plays are highlighted with colored text where:

• Green = Doubtful
• Blue = Bad
• Red = Very Bad

In addition to the font colors, bold font is used to designate a particularly lucky roll.

In addition to checker play, the GNU analysis also evaluates doubling cube decisions, and highlights them with the colored backgrounds on each move.  For example, on move 6, Player A’s move is marked with a gray background, and Player B’s move is marked with a yellow background.  Let’s select Player A’s move and see what it says.



Player A Doubles: "Doubtful"

Now the analysis box shows the cube analysis, rather than the checker play.  It shows lots of probabilities of winning a single game, a gammon and a backgammon for each player, followed by the equity impact of various doubling decisions.  GNU’s conclusion is that Player A’s position is too good to double, because he has a 47.2% of winning a gammon.  So the preferred decision is, don’t double.

Now, you would think that if Player A was in that strong of a position, that Player B would surely refuse the cub and drop, right?  But that’s not what happened here.  Let’s look at Player B’s move.



Player B Accepts: Very Bad

Here we see that Player B’s acceptance of the cube is a whopping 4.715% worse than rejecting the cube and dropping the game!

Well, that is the brief overview of the GNU analysis tool.  In addition to the checker play and cube decisions, GNU also analyzes the luck factor for each move and even gives the players a rating comparable to that used by FIBS.  These are all summarized in the various charts and graphs in the GNU analysis tool.

If you want a very powerful tool to use to make your backgammon game stronger, consider trying out GNU backgammon.  It is free, and it is a world class player.  Because of all the options, there is a bit of a learning curve, so just take it one step at a time.  Soon you will be playing a better game!

Watch this space for backgammon problems and solutions, according to GNU.