Backgammon

[Klaymen]

I was teaching my wife backgammon, and she was getting used to the game. She's not the competitive type, and sending someone's tile back to the bar seemed mean but she was getting used to the idea.

I rolled a 5&5 and took a pair of my 6's and knocked off her single tile deep in my home territory. Up on the bar it went.

She rolled 6&6 and was ready to make good use of them until I reminded her it was wasted because she couldn't get her tile off the bar.

She proceeded to roll:
6&6
6&6
6&1
:cry:

It was time to just stop the game and hope she'll play it again sometime.

[BC]

You know, I've found people always tell you they don't want you to "let" them win until you really, really beat them badly...

That said, I have been burned a few times by trying for an "artistic" win when a simple bash would have sufficed.

[Expunge]

I havn't studied backgammon all that much, but with it being a full information game isn't optimal strategy solvable, and therefore the game results in just luck of the dice?

[BC]

I havn't studied backgammon all that much, but with it being a full information game isn't optimal strategy solvable, and therefore the game results in just luck of the dice?

Yes. Just like chess. If you want a real challenge, you should take up bridge.

[3rookie]

Expunge, would you make that same statement about chess?
(Obviously w/o the dice)

The correct move is not to play

[Gandalf]

I havn't studied backgammon all that much, but with it being a full information game isn't optimal strategy solvable, and therefore the game results in just luck of the dice?
Theoretically solvable, yes. Like chess.

Practically solvable, no. Like chess.

What about compared to poker? In theory, shouldn't an optimal strategy be solvable? If you answer is of the form "my optimal poker strategy depends on what my opponent would do in certain situations", then the same is true of backgammon. My optimal backgammon strategy depends on my opponent's ability.

[Expunge]

Whoa, sorry I didn't mean to ask an offensive question, or imply any knock on any game. I'm really interested in figuring just how complex the game is. An individual game of backgammon has an element of luck just because there is a roll of the dice. Obviously luck balances out over the long run. Chess on the other hand does not have this randomness. There is no luck involved in deciding if you can move a piece or not. You either can or you can’t. Backgammon has what 24 spaces and 15 pieces per player. Chess has 64 spaces and 16 pieces per player, not to mention that there are 6 different types of pieces each of which can do different things. But then again backgammon adds in the complexity of probabilities of what someone will roll. As for Bridge now we get back to a partial information game. There is still luck involved since cards are randomly dealt to you.

Anyways I think I’m on the right track towards getting the answer I was looking for thanks to Gandalf. In magnitude of solvability complexity would you put it above or below that of chess, and to what degree?

[BC]

I havn't studied backgammon all that much, but with it being a full information game isn't optimal strategy solvable, and therefore the game results in just luck of the dice?

Yes. Just like chess. If you want a real challenge, you should take up bridge.

Edited to add proper punctuation: :D

I won't speak for anyone else, but I wasn't offended, I just meant my reply as dry sarcasm.

[Expunge]

What about compared to poker? In theory, shouldn't an optimal strategy be solvable? If you answer is of the form "my optimal poker strategy depends on what my opponent would do in certain situations", then the same is true of backgammon. My optimal backgammon strategy depends on my opponent's ability.

In backgammon i would think there is an optimal strategy that would win in the long run (tie vs. the same strategy) without knowing what your opponent would do in certain situations. Based on just what is infront of you at the time. As if your opponent does not follow this optimal strategy at certain times he is giving more away. Now while this optimal strategy can't be beat in the long run does not mean that it wins at the maximum amount.

In poker since you don't know what your opponent holds you must make some judgements. If you always assumed you were up against a completely random hand then i agree you are up against computable odds. I suppose though that you could do some heavy probability calculations and calculate vs random playable hands if they were optimal. The chances of someone acting behind you could also be calculated. Hmm... It might be possible. Lots of moving pieces lots of equations and lots of unknowns lots of assumptions. More unknowns than equations i'm not sure. I'll have to think more about it. Very interesting.

Full table poker obviously has more players to take into account, and well more than 21 possible combinations on any one turn. But i guess if you take multiple turn into consideration with backgammon there could be a significant number of different possibilities.

[Smooth Verona]

The luck of Backgammon.

It is interesting that I consider myself a much better player than my wife, but she is up in points over the last 10 years or so. She is an unbelievable roller (I know that is stupid to say, but seems to be true). Recently, I've been making her play with the doubling cube, and she has been making some bad decisions. Also, whenever we let our daughters roll for us, I consistently beat her. I really don't get it.

My father-in-law, who is well schooled in backgammon, says my wife is the smartest roller he's ever seen. :)

[BC]

The luck of Backgammon.

It is interesting that I consider myself a much better player than my wife, but she is up in points over the last 10 years or so. She is an unbelievable roller

Yes, that's what I'd believe in your position. :wink:

[Expunge]

Id be interested in a backgammon league / tournament here. I am by no mean volunteering myself. I've got enough to do with the poker stuff and that never ending chess league.

[RedSoxFan]

Backgammon does have dice so there is some luck component. But there is still a lot of skill involved. Not as much as chess, of course, but more than most people think at first or second glance.

I think backgammon particularly appeals to actuaries in that there a lot of probabilities involved, and it does come into play somewhat. For example, it usually is a good idea to leave a blot in your outer board, even if subjecting yourself to an indirect shot, since it increases the probability you can make an inner board point or your bar point next turn.

Today's backgammon programs are about as good as chess programs in terms of how close they are to playing the game perfectly, though they are quite different in their approach to the game. The best chess programs analyze as many positions as possible, seeing several moves or more in advance. This approach won't work in backgammon, since with the various die rolls and manners of playing those die rolls each move, it would take way too much calculation time to even look 3-4 moves in advance. This is the way backgammon computers were programmed up until several years ago, and of course they stunk. But a guy named Gary Tesauro started to apply neural net theory to the game, and computers have become very strong. Basically, the computer has 20-30 variables and assigns a value to each variable based on features of the position. Each variable has a different weight, and these are used to calculate its final judgment of the position. The weights are developed by having the computer play itself millions of times. The best programs today have three nets (or sets of weights), one for regular play, another for back-games, and I think the third is for prime vs. prime battles.

Anyway, backgammon is very challenging to master. It's like chess but with the dice involved every game is different, and even if you're playing a much better player, you still have a shot at winning. In particular, the doubling cube, which first started being used about 100 years ago, has been an excellent addition, adding a lot of strategy to it, and perfecting the game. Today's programs are very good at showing you your mistakes and rating your play (calculating the amount of "equity" you lose in the position as a result of each imperfect move, and calculating at average to show you how you're doing) - so it is a challenge to learn how to play the game better and see how your error rating improves. One of them happens to be free, a result of the Gnu open-source programming project. It is called Gnu Backgammon. It is just as strong as the #1 commercial program, has 95% of the features, and 90% of the user-friendliness, and like I said, its free (and the #1 commercial program is close to $400). The best place to play online that is free is FIBS (www.fibs.com).

[RedSoxFan]

In backgammon i would think there is an optimal strategy that would win in the long run (tie vs. the same strategy) without knowing what your opponent would do in certain situations. Based on just what is infront of you at the time. As if your opponent does not follow this optimal strategy at certain times he is giving more away. Now while this optimal strategy can't be beat in the long run does not mean that it wins at the maximum amount.


Yes, definitely. Though when playing a much weaker position you may e.g. make some sacrifices (i.e. leaving more and riskier blots) if you know your opponent is afraid of hitting them.

[RedSoxFan]

Id be interested in a backgammon league / tournament here. I am by no mean volunteering myself. I've got enough to do with the poker stuff and that never ending chess league.

www.dailygammon.com has a server much like our chess server.

[Expunge]

Thanks for the great info RSF.

[Expunge]

Id be interested in a backgammon league / tournament here. I am by no mean volunteering myself. I've got enough to do with the poker stuff and that never ending chess league.

www.dailygammon.com has a server much like our chess server.

Hmm Glenn?

[Gandalf]

Determining optimal strategy would be immensely complex for both games.

Human approximations of optimal strategy are more important in chess than in backgammon. You are never going to beat Kasparov in chess, even in a single game, because he understands the strategy so much better, and strategy wins chess.

I have some chance of beating the world's best backgammon player in a single game, because there is some element of luck. You can win a single hand of poker against anyone, because there is some element of luck.

In chess, looking ahead a single move, there are many possible moves your opponent can make, but a rather limited number that realistically he would make. As much as possible, you plan your game on the idea "suppose he makes his best move in response; what then?"

In backgammon, looking ahead a single move, I suspect there are more potential moves your opponent can make (potential, before he rolls the dice. Once he has rolled, the options he has available are fewer than the options a chess player has). With the probabilistic nature, you cannot predict as well what your opponent will do. You may choose to move in a position that leaves you very vulnerable to your opponent rolling double 2's, but strong against most other things. Alternatively, you can guard against the double 2's while being less ready to exploit his other rolls.

There is also a doubling cube, which adds complexity. With the cube, some games are worth more than others. As in poker, the decision to double is not purely a function of whether you expect to win. If you propose a double (like a bet in poker), your opponent may fold, call or "raise". The backgammon "raise" is a little different. Immediately, he folds or "calls". Thereafter, he retains the option to bet (to double again); while you have no option to bet. [Were he to double again and you to call, then you would again be in control of the cube, the only player eligible to propose another double.]

Yet another complexity to backgammon over chess is that, even in the absence of doubling, some games are worth more than others. If your opponent gets all 15 men off before you get any off, the game is worth 2 instead of 1. If he gets all 15 off while you have none off and at least one near your start, then the game is worth 3. In some positions, your chances of winning 1 are much better if you adopt a strategy that risks -2 or -3 than if you adopt a strategy that pretty much guarantees -1 with no chance of +1. If you're employing the strategy that risks -2 or -3, as the game develops you have to decide when to cut your losses and hope for the -1 result. In chess, every game counts the same.

[RedSoxFan]

Those are good comparisons Gandalf, thanks.

[E. Blackadder]

There is additional complexity in match play; not all points are equivalent.

[Expunge]

i think the gnu dice rolling program has a cheat in it.

[Gandalf]

We had a home electronic backgammon game. The dice rolling "random process" was that it started cycling through the numbers (not displaying them) and you stopped the process by pushing a button. The assumption was that the cycling was fast enough, and your pushing random enough, to produce random results. My wife figured out, even though the dice weren't shown, that certain flashing lights corresponded to the results, so she was able to control die1 (which cycled slower, changing only after die 2 had cycled through 1-6) extremely well, and was pretty good at die 2.

That's not a strategy available in playing chess against a machine.