Nothingness: if you resign after bearing off at least one, it will be 3-0... otherwise it will be 5-0 (or 7-0 if you're pieces are still in opponents area).

So get one piece off first, then resign if you can't mathematically win.

if i resign during a game of probackgammom while the score is 1-0 in a 5pt mathc and the doubling cube is at 2. essentially gving me a 3-0 deficit? or does it give me the whole 5-0 loss?

I have another question regarding GnuBG.

I want to play Hypergammon with GnuBG - so I ran makehyper to generate the database files. But now I have no clue where to put them. I tried putting them in my home directory, into the build directory of GnuBG, and into /usr/share/gnubg and /var/lib/gnubg - I could never selebt Hypergammon as the game type, only Backgammon and Nackgammon. The GnuBG documentation only tells me to generate the files, not where to put them. So has anybody of you done this before and has an idea that might get me on the right track?

Thanks :)

toedder: I suspect that a few of the add-on programs that come with various distros are behind the cutting edge, especially with a product like GnuBg that is officially still unreleased yet with a fairly frequently updated development version available to the public. Bugs pop in and pop out again and a distro can end up with a buggy snapshot. I'm glad to hear that changing the version - in both directions, lol - worked for you.

When I said I downloaded and installed the latest version, I was wrong. I got a little confused by the naming conventions for their tarballs and apparently I downloaded an older one from 2008. I now installed the snapshot from today, and it still works as intended.

Position ID: AwAAEAEAAAAAAA
Match ID: MAGgAkAAKAAA

playBunny: I was using the latest version from the Ubuntu repositories, so until your post I just assumed it was fairly recent. I just downloaded the latest snapshot and installed from source, and voilà! it changed and tells me that "Double, take" is indeed the correct action.

So it seems like some glitch that was introduced and fixed between your versions, or maybe some obscure error resulting from the package building process. It could always be a bug in one of the libraries used, etc.

Anyway, thank you all for your assistance, and for restoring my faith in GnuBG, and my certainty that it is smarter than I am ;)

I don't have the mat file on this computer atm, and do not remember what match it was, but the position ID was AwAAEAEAAAAAAA

toedder: g11 is a GnuBg-generated MET that has values created using rollouts for scores up to 11-away, 11-away. g11 is the default MET for that version. I think the latest GnuBg uses, the Rockwell-Kazaross MET.

Yes, METs are only for match play.

It's very curious that you have this error and I don't have it in versions either side of yours. That's assuming that I'm putting in the corretc position. Perhaps you could upload the match somewhere?

Alternatively, the problem may simply go away if you download and install the latest version.

Ah, okay! Thank you

It was set to g11, whatever that means. Mec25 yielded the same result, as did the different Jacobs METs. And again, it even told me that spot was a "No double, take" in money play, and METs shouldn't be taken into account there, right?

(A lot of this is new to me, I haven't gone past the "looking for huge mistakes I made" and the "let's see whether I really was as unlucky as it felt" stages of analysis yet)

toedder:  there are various MET (Match Equity Tables).  gnubg--and any other capable playing software--should let you specify which MET you want to use for cube decisions.  The MET are tables showing the theoretically expected match winning percentages from various match scores...assuming equal-strength opponents.

The MET are less important at scores such as your 17 away / 16 away.  But I'm currently trying to memorize those values out to about 7 away / 7 away for real-life tournament play.

The version is Version GNU Backgammon 0.90.0 Dec 9 2010

What is the MET and how do I find out about it?

toedder: I can't reproduce the error on any of the versions that I have (Oct 2005, Nov 2009 and July 2011). Which version of GnuBg are you using and what MET is speciied?

wetware: I am on my phone atm, so I cannot easily look it up. But I am fairly certain that I was 17-away and my opponent 16-away. And as mentioned, GnuBG told me the same when I set this up as a money game. So to me it seems as if the expert level simply has some big flaws when it comes to end game situations.

grenv: Yes, it was removed by Mr. Beaver.

Pedro Martínez: Was the post removed? It was a proper link to a respectable slang dictionary (beaver is an animal, any other usage is slang). If that offended someone then they need to grow up and join the rest of us in real life.

grenv:

rod03801: Thank you rod. An interesting word for a betting action.

We used to use "beaver" to mean someone that watches you play pinball and gets so close to the action that they bother you while asking questions about the play.

toedder: i did do something stupid. Anyway, must be a bug in the program then, since the calculation involves probabilities 1 and 0 only.

Walter Montego: I found this :

A rule often used in money play (but never in match play) which says: A player who accepts a double may immediately redouble (beaver) without giving up possession of the cube. The opponent (the player who originally doubled) may refuse the beaver, in which case he resigns the game and loses the current (doubled) stakes. Otherwise, he must accept the beaver and continue the game at quadruple the stakes prior to the double

What does the term "beaver" mean in this?

toedder:  discussion of a similar position can be found on pp. 66-68 of Improve Your Backgammon by Lamford and Gasquoine.

However, their initial example is at a much different match score: with the cube already "yours" at 2, and with you leading in the match by a score of 3 away / 5 away.  Given that situation, the correct cube action is: No Double / Take.  (By their methods, you would need to be a 62.5% favorite to bear off in that situation for the redouble to be correct.)

They do also discuss a general approach to the cube decision, given any particular match score.  (So, I'd love to know the exact match score in your case--just how early in the match was it?)

Also mentioned in the book is the correct cube action if the position were played as a money game:  double/redouble/take

You counted 65 twice ;)

I don't think it's a user error. I don't know how I would make an error there. Also, the only thing I changed for it to come to the same conclusion as myself was setting the analysis level to grandmaster instead of expert.

52, 53, 54, 55, 56, 62, 63, 64, 65, 66, 44, 33, 22

isn't that 21 chances? did you forget 33 and 22? or did i do something stupid...

Either way double is correct, and simple since control of cube is not in the calculation. Must be either a bug in the program or user error. ;)

Is the expert level of GnuBG even useful? Or is just that it sucks at simple end game problems?

I and my opponent both had two checkers left - I had one on the 5 and one on the 2 point, my opponent's both were on the 1 point. It's my turn, center cube. So it's hit or miss. I calculate I am a 19:17 favorite and doubled. After the game, GnuBG told me that was a mistake, despite agreeing that I was a favorite in that spot. But it somehow came to the conclusion that after me doubling and my opponent taking it, my opponent would become the favorite. I was confused, as that didn't make any sense to me. It was early in a 21 points match, and the score was close, so I didn't think that would have any impact - but I let GnuBG analyze it as a money game regardless. It told me No double, beaver would be the right move. Didn't make sense to me, and I don't see how beavering would change that simple of a problem, where beavering HAD to be bad imo. Still, deactivated beavers, GnuBG tells me it's no double, take, despite me being a favorite. I recalculated, and messaged two people about it, because it didn't make any sense to me. Finally I had the expert idea to change GnuBGs analysis level from the default expert to grandmaster. And there it was, the Double, take, that I had envisioned.

So why would expert level fail to get such a super simple problem right? I mean I have no doubt GnuBG is worlds better than I am even at the expert setting. But at what point does it become unreliable? Does it have a general weakness regarding endgame patterns?

Anjil: Further to what grenv said, random is not the same as varied. In order for the dice to be random, these patterns have to occur every now and then.

You say that you noticed it before. I take that to imply that you've also stopped noticing it, presumably because that clustering went away .. until the next time you noticed. That's what random does. :-)

Anjil: 11 games, or 22 dice rolls, is not enough to know anything... this set of results is random enough. I guess the biggest anomaly is 8 fives and only 1 one .. but you need to collect about 100 starting rolls to even begin the discussion.

Of course, someone could query the database and post the stats... :)

I've just made opening move on 11 games,Why do I get the same opening dice rolls across multiple games?

I've noticed this before but now it's really bugging me

My opening dice for the 11 games were as follows:

3 games (5/4)
3 games (5/3)
2 games (5/2)
2 games (6/2)
1 game (6/1)

Lots of same opening dice numbers, this doesn't seem very random to me

Anyone else seeing similar or just my luck?

Games links below:

Backgammon (Rainbow Days vs. Anjil) (5/4)

Backgammon (Anjil vs. Rainbow Days) (5/4)

Backgammon (Slavek vs. Anjil) (5/4)

Backgammon (lenkacz vs. Anjil) (5/2)

Backgammon (Anjil vs. Slavek) (5/2)

Backgammon (Slavek vs. Anjil) (5/3)

Backgammon (Anjil vs. Thom27) (5/3)

Backgammon (Anjil vs. Thom27) (5/3)

Backgammon (Anjil vs. slon4e) (6/1)

Backgammon (Anjil vs. Magrata) (6/2)

Backgammon (Anjil vs. AlterMann) (6/2)

rod03801: Yes, the odds are considered in a way yet they don't get deliberately calculated for their own sake.

Lol. You can't talk about being slow when this is your introduction to a very challenging topic! I've wanted to explain something about neural network backgammon here for the last few years and have written a few lengthy posts on the topic but I've always discarded them before posting. Painting a reasonably accurate but understandable picture of how the bots work is difficult and this is the first time that I'm happy enough with what I've written to actually post it. Even so, this is more of a sketch than a picture so I'd be impressed if anyone new to neural networks gets what they're about just from this.

playBunny: Thank you for yours and other's thoughts over the last few days to my questions and comments. I'll return a few more times as I see how they fit in with what I have observed from playing.

I do have an opponent that does have the games in private. Not saying nothing about, but it seems like a strange thing to do for any game, but especially for a game like Backgammon. I can see an argument for it in Dark Chess, Screen Chess, and Battleboats, but Backgammon? I asked him about it and all I got was a smile for a reply. I do not like playing private games. It makes it extremely difficult to share a great or unusual game with anyone else that might be interested. And now I have a reason to think about it that has implications about it. This particular player makes double offers in places none of my other opponents do, and he is the only person I have seen that is luckier than I am, well one other player is this lucky, but he knows the dice guy and pays him a lot.

playBunny: I need to absorb that more. On first reading, it sounds like in a way, it actually IS considering odds? Maybe not in a straight forward way. I need to stew on that. Interesting though. I'm slow sometimes!

rod03801: But is there a specific reason to NOT consider odds?

The main reason is that doing so wouldn't add any new information. A position evaluation already includes the odds. In fact it includes the odds, and outcomes, of everything that could possibly happen from that position until each end of the game. All of that exploration into future possibilities is condensed into how much it'll win and how much it'll lose, on average, from that position.

How a bot learns is by playing a million or trillion or gazillion games through to the end and, for each position along the way, it records the outcome. If that position already has a value from one or more outcomes then the latest outcome is merged into the value. Many positions occur again and again and so the value for each of these positions becomes more and more accurate. Positions that occur more rarely will accumulate fewer outcomes and be less accurate. Also, the closer a position is to the start of the game the less accurate it'll be because a smaller percentage of the myriad possible paths will be travelled by the set of games that are explored.

However, that's an as if kind of explanation. No database can store each and every position; that would be impossible given the sheer number of possible positions. (Although it is possible with hypergammon because, with only 6 checkers rather than 30, there are many fewer positions). The neural network method that the bots use is very clever. During the explorations mentioned above, when adding in the outcome for a given position, what they do is record and merge the value for the position with that of positions that are like the given position. By recording "positions that look like <this>" instead of actual positions, the storage requirements are greatly reduced, although this is at the expense of some degree less accuracy for a given position.

In a well designed neural network those "positions that look like <this>" will be fine-grained enough to capture the subtleties of positions that are fairly similar visually but different backgammonly, such as one with a blot that can be hit directly using a 6 and a matching position where the blot is just one pip further away and needs both dice in order to hit it.

Apart from the ability to store all that information with a practically sized database, the other, and huge, advantage of the neural network method is the ability to generalise. Because it doesn't store actual positions you can give it a position that it's never seen before and it can always find a similar position whose values can be used. The mathematical "distance" from the position to the similar position will vary but often it's not so far that the accuracy suffers too much. Thus the computer can respond as if it knows the position, even having never seen it.

playBunny: Very VERY interesting.

I am surprised about a program not considering odds. Why would that be? Maybe I'm naive, but I would assume with all the power of computers, that it would be easy to incorporate that into it as well.

I mean, I understand that what it DOES consider, is obviously "good enough" for the most part. But is there a specific reason to NOT consider odds?

Walter Montego: Loadsaquestions!

Sometimes playing for a long shot is so much more worth it than playing safe. Which way will a computer decide?

You might be surprised to hear this but the computer doesn't care about the odds of a shot and never calculates them. The same goes for the odds of a leaving a blot and having it hit, closing a prime, etc. It doesn't a care. In fact it doesn't even know what odds are!

What it does is make each of the possible moves and then evaluate the resulting position. The one with the highest evaluation is the winner and the move that got to that position is the one that gets made. The computer's database is all about knowing the value of every possible position to a reasonable degree of accuracy. How it gets to any position (eg. through a long-shot hit or just by moving pieces) is irrelevant. A backgammon program is very much more a stats machine than a conceptual player.

I'm sure you've made a move that you know was not as good as a different move, but if your opponent doesn't get the roll to make you sorry for taking the chance you will win the game. So, do you go for it, or not?

Unlike the computer player, a human player will often consider the odds. That's because we don't have quite the same kind of calculating ability for comparing moves. Our intuition is the best match to the system that the computer uses but we have much, much less experience. We do know, however, that the position evaluation is related to the odds of various shots. (The computer doesn't need to know this as it can compare exact(ish) outcomes directly). The other thing that we do that the computer doesn't do (and really cannot do), is think conceptually.

I might know, for instance, that, being behind in the race, I must hit my opponent, despite the risk of being hit back. For example, hitting might leave a blot in my 5-point home table. In a case like this I'll know that the chance of being hit is 11/30 but if I don't hit then a 5 or 6, at odds of 20/36, would let my opponent escape with a most likely win resulting. I don't need to know the equities of the positions, I can compare the odds and they dictate the hit. The computer goes straight to the equities, compares those and comes to the same conclusion.

So, do you go for it, or not?

Having said the above. There are times when I don't play rationally but play "psychically", ie. I try and second-guess the dice. And hey, sometimes it works! And sometimes I might do the incorrect move just for the hell of it, because getting the luck would give me a buzz. (For some players that's their predominant style, especially cube gamblers). By and large, though, I try to play what the backgammon programs (mainly GnuBg) and my own experience have taught me over the years is the optimal move.

the rating points for Backgammon on this site. Single games seem disproportionally in favor of the player who is rated lower when his rating is more than 60 point

Absolutely. The chess rating formula used for backgammon makes a mockery of the rating system. The risk/reward ratio is totally wrong for high rated players. It's impossible to consistently win as many matches as needed to maintain your position unless you go for longer match lengths which reduce the luck element. A 60% winning ration for single pointers is very good but the chess rating formula means that you might need to win 75% in order not to be penalised.

Is a person that uses a computer to help with his moves on this site play noticeably different from someone that doesn't use it?

Anyone whose rating chart has anomalies is worthy of suspicion. The rating cheats who have opponents dropping like flies for them can be spotted by sudden streaks of wins. A person using a bot to determine their play is less noticeable but, unless they really are good, they'll progress too easily.

What would be an example of such play?

Usually you can't really tell but a good player can sometimes spot moves that are unusual or advanced. Moves that wouldn't occur to most players, such as hitting in their home table and leaving two blots. That's more risk than most players can accept but there are times when it's definitely the right thing to do.

Then there are other occasions when a bot-trained player will spot something that only a bot would do. In some circumstances, for example in a holding game or anchor game, when there will be a few moves before there's any chance of hitting action, there's time to build one's home table. Doing that efficiently might mean making a vulnerable-looking mess (that will suddenly firm up into a solid board). The bots have a particular way of doing this at times, maybe having three blots in their home table. Human players tend to be more cautious and/or neater in how they do it. Seeing a player do such a bot-style spread suggests that they're either good, and bot-trained, or using a bot.

Is it something noticeable at once, or is it much more subtle that it takes many games and certain move situations to notice or even have a suspicion that it might be going on?

Both nabla and I spotted the bot-player that I mentioned before. I seem to remember seeing some of the bot-style moves and being suspicious. I was also suspicious of the rather too perfect rating climb. nabla and I analysed some his matches using GnuBg. Initially it looked suspicious but not conclusive because there were still mistakes being made. However that analysis was done at 2-ply, which is a strong setting. nabla had the idea of redoing the analysis at 0-ply and then the mistakes vanished and the player was show to be playing exactly as the program dictated.

Since he was exposed, he plays all of his matches in private mode so that nobody can see and analyse his moves!

Does the computer ever conclude that that the winning player made the most mistakes or the worst of them?

Absolutely. Almost every time you beat a bot then you'll have played mistakes against an opponent that made none, yet you won. Similarly, good players who lose to lower rated players will usually have played better but still lost. That's how powerful a factor luck is in the game, and what keeps people coming back for more, even if they're rubbish players.

What's a mistake?

When a bot analyses a position and a roll it will create a list of the possible moves and an evaluation for the resulting position. The top move, the one with the best evaluation, is the best move and all others are mistakes. - according to that bot at that analysis level

But how do we know whether that list of moves, and in particular the top move, is correct? The only way to know is to analyse even more deeply.

Thus the 0-ply analysis, used by the bot cheat mentioned above, showed mistakes when the matches were analysed at 2-ply. As 2-ply is more accurate than 0-ply, we can assume that 0-ply's mistakes are mistakes. (or at least very probably)

However, it's important to note that the bots evaluate moves and positions as if they are playing themselves at that same analysis level. In real play, with a human opponent, you can be sure that your opponent has weaknesses. A bot will never play to exploit its opponents weakness. It can't, because it has no such concept as "opponent", let alone "weakness". The computer knows how *it* would play a given position and roll and that's all it needs to know. When I said "as if they are playing themselves" just above, what I mean is that their statistical database (ie. the neural network) has values that only reflect play against itself.

And what happens if it is the winning move despite being a mistake? A move is a mistake because on average it produces a suboptimal outcome (the lower evaluation mentioned before). But you can often make a mistake and get away with it because, in any given game, the dice are being themselves, not trying to be average.

I've many times made a move and clicked Submit just fractions of a second before realising that it was the wrong move, and then breathed a sigh of relief because my opponent didn't get the roll that could capitalise on my mistake. Similarly, when analysing a match, the computer will show me many mistakes that I didn't realise (or wondered about) at the time and the follow up shows that I got away with it.

grenv: Does the computer ever conclude that that the winning player made the most mistakes or the worst of them? What's a mistake? And what happens if it is the winning move despite being a mistake?

Walter Montego: Easiest way is to play the game through a computer yourself afterwards and see if the player made any mistakes. Most players will make a few mistakes.

in reference to your longshot vs safe... the computer would decide just as easily as any other move... with math.

For example if you had a choice.. keep the game 50/50 or make a move that is 30% likely to win and 70% likely to give the opponent a 70/30 advantage... what would the computer do?

move 1: 50%
move 2: 30% + 70x30% = 51%

Move 2 wins.

The human may think this is a difficult decision (because the %ages are hard to calculate so it feels like a gut decision).. but not the computer.

playBunny: Is a person that uses a computer to help with his moves on this site play noticeably different from someone that doesn't use it? What would be an example of such play? Is it something noticeable at once, or is it much more subtle that it takes many games and certain move situations to notice or even have a suspicion that it might be going on?

playBunny: It seems like related subject to this is how to reward the rating points for Backgammon on this site. Single games seem disproportionally in favor of the player who is rated lower when his rating is more than 60 point behind his opponent. I am thinking that as the match length is increased it becomes more fair from the higher rated player's perspective and this accounts for why the higher rated players will play me even though I am 200 or more points behind them in rating.

coan.net: Just knowing the common rolls and what to expect isn't enough. Sometimes playing for a long shot is so much more worth it than playing safe. Which way will a computer decide? I'm sure you've made a move that you know was not as good as a different move, but if your opponent doesn't get the roll to make you sorry for taking the chance you will win the game. So, do you go for it, or not? And this obviously has to take into account if it's a single, a betting game, or a match, and if a match the current score.

Walter Montego: And what about luck? I get real lucky against some of my opponents and other I just get stomped.

Luck will always play a large part of success in games, matches and sets of matches. It's only when you make those sets large that you can discount the luck, because it will tend to balance out.

In talking about the advantage that a computer's knowledge can give in a given game then it's really a question of how many errors the given player is likely to make. That's a function of their skill level, the familiarity of the game type and the complexity. As you noted, some positions are easy and the move is "standard" or "obvious". The set of standard moves becomes larger the more skill a player has and the value of the computer in those positions diminishes.

coan.net: For a computer - they can look ahead 20 moves - they can easily see what the more common rolls will result (More people will roll a total of 7 then roll a total of 3.)

Backgammon computers typically look only a few moves ahead, 2 or 3 for Snowie and GnuBg, 3 or 4 with eXtreme Gammon.

However, that's something of a misleading value because even looking no moves ahead, a backgammon program will still be extremely strong. That's because the static evaluation of a position already includes the statistics of looking ahead to the end of millions of games.

Walter Montego: There is no argument that chance is a part of backgammon - but you will get a wide variety of people who argue how much luck there really is involved.

For a computer - they can look ahead 20 moves - they can easily see what the more common rolls will result (More people will roll a total of 7 then roll a total of 3.)

Walter Montego: It depends on what you mean by "using a computer" and "advantage".

If you mean having the computer choose the moves then there's a perfect example in the top three of the backgammon rankings. That is, there's enough of an advantage that you'll get to the top of the rankings, beatable only by a very strong human player who plays the longest matches possible to take advantage of his cube skills relative to the much weaker playing pool.

Walter Montego:

Hmm, maybe part of my question might depend on the actual position as it seem that a computer might help in some positions a lot more than others.

(A post of mine from a game while having a discussion about computers)

Just how much of an improvement can using a computer make in Backgammon? In Chess there's no arguing about the improvement, but in something so dependent on the roll of the dice and the 11 different rolls and 36 combinations of those rolls and each turn multiplying it and the permutations from where to move one's pieces, it seems like it'd not be all that much of a help. All the same, I agree with you as it seems it might help in some way, but how much?. Even the world's champion would have trouble defeating either of us more than 60% of the time in individual games, but remember these guys aren't playing individual games, they're playing matches. Even 55% of the wins would make for a match win of over 80%. Count up your games won and lost in a couple of matches and see what it adds up to. And what about luck? I get real lucky against some of my opponents and other I just get stomped.

Single games? Short matches? Long matches? Individual games with money bet on them? Does any of this matter?

Walter Montego: Read the latest posts on the Feature Requests board.

playBunny: Then the problem should be fixed. Either start over with a new dice program, or find out where this link between the rolls is being made and sever it.

I haven't posted to this board in many years. It looks from the links you and pedestrian posted here that this has been discussed and reported about from last year. So where's Fencer at on this? I am sure he wants the games here to play correctly. And any type of random dice program must be a fun challenge to make it work randomly, which they aren't doing now.

Walter Montego: Can't we just have the dice to simulate a real pair of dice being toss

Actually I think that that could be exactly what's done and where the problem lies.
http://brainking.com/en/Board?bc=26&plla=1067194

pedestrian: Well then!
IF this is the case, what's being done about fixing the problem? Can't we just have the dice to simulate a real pair of dice being toss instead of some fluke that can't be right and might explain why I win more than I should, or less?

These links confirm that tony's and others' charting of the dice have shown a pattern of something being wrong with the dice.

It is a fairly simple thing to have a well working random dice simulator, so what's the deal? A day or two's work on a program should be enough. This is a major problem since it affects the integrity of the game even if the flaw is fair for both players, and I am not so sure that it is fair for both players since in an identical paired start one player or the other will have an advantage depending on the what rolls were made. Maybe not 6-1 or 3-1, but what about 6-3?

coan.net: There's an open bug report on the problem, plus a lot of previous discussion on this board: 

tonyh: Out of those 57 games that had the same - did you happen to pay attention who won the opening roll? (Player 1 or Player 2)?

And were they all no-cube matches or with the double cube?

I'm just wondering if maybe we can determine if we can see more of a pattern here... I myself have noticed it also, just never felt like sitting down and counting.

For some time now, I have felt that my opponent's and my opening moves were identical. I shared this view with Walter Montego and he agreed. We decided that I do a test of 150 games to see how often the opening moves were the same, expecting a result of about 10, at odds of 15/1.
Out of the last 153 of my games, 57 had identical opening moves. Surely, this cannot be right; it needs looking into.