Backgammon: Bad luck or bad move?

Here is a diagram of a position that occurred in a game I just lost. (I am white, playing vs an AI.)

Brief description: I have a pip count advantage of 50 vs 100 or so. I have two of his checkers still in my home board while I am beginning to bear off my own checkers. As well as the two checkers he has in my home board, he also has two additional checkers not yet in his own home board. This, I thought, was a sure win for me.

I rolled 5 and 3 (see diagram) and I bore off* one of my checkers on 20, and moved the other from 20 to 23.

He rolled 4 and 2, and moved 13/9 and 3/1

I now rolled 5 and 1. Only one move was possible: 19/20 and then bear off the 20.

This left a blot on 19, and of course he hits it on his next move. I didn’t think this was cause for panic, though–he now had three blots out on the board, and I still had a massive pip advantage.

But for the next I don’t know how many turns, I only rolled 1s, 2s, 3s and 4s, meaning I could never get my guy off the bar, and giving him all the time in the world to completely turn the game around. I lost.

My question is, did I make a mistake? WasI wrong in my evaluation of the diagrammed position as a nearly sure win for me? Or was I right about that, but did the wrong thing with the roll of 5 and 3? Any other move would have left a blot behind, so it seemed best to do what I did. But is this wrong?

Did I just get hit by some really bad luck, or did I make a bad move?

*Is that how you conjugate “bear off” in the past tense?

Yeah, you got boned. I’d call that position a possible gammon and a near sure win. I’d have offered a double as soon as I left the blot open though, to force him to concede. Even once you got hit however, you’d have expected to get in within a few rolls - there’s only a 3% chance of not getting in within 4 rolls.

ETA: There was nothing else you reasonably could have done with the 5/3 - you have to bear off a 5, so your only other moves are to move a 5 to a 2, bear off a 3 and leave a blot in 5, or move 6 to 3 and leave two blots. Leaving a blot you don’t have to is utterly insane at that point.

Good to know I wasn’t off-base. Thanks!

How do you know it wasn’t the AI cheating? AIs love to cheat.

Can’t tell how serious…

I am serious. I don’t know about backgammon software in particular, but lots of games with AI opponents cheat to make the AI better.

This is the AIFactory Backgammon app. There is a manual dice option they made available to prove the AI doesn’t cheat, though I think this should have been completely unnecessary. I would bet money that no commercial Backgammon program has ever been programmed to cheat. I can see absolutely zero reasons for that to ever have occurred.

Cool, a backgammon thread. I’m by no means an expert and I can’t access your link at work, but basically you can only chastise yourself for bad play if there was a move you could have made with the 53 roll that would have left you with no blots, and less chance of a blot next roll. In other words, if it was possible to play the 53 such that no possible roll left a blot on the next move, you probably erred. But I doubt that was the case, so you were just unlucky, most likely. In fact, having written this, I see muldoonthief has already confirmed that nothing else was reasonable.

As I’m sure you know, given the opponent hit the blot immediately, there was no opportunity for the OP to offer a double, until he had already been hit, when a double would be risky.

I agree it is extremely unlikely that the computer cheated. There is no reason to do so, it was just good luck for them. Any appearance of computer cheating in backgammon is usually down to confirmation bias, and the fact the software is good enough to set itself up to cope well with as many rolls as possible, so more “good rolls” occur. Doesn’t apply in this particular situation of course. But play that same position 100 times against the AI, and it would show that hitting is just random chance.

Yeah, I use a program called Backgammon NJ for my iPhone, and the developer had to release a very thorough explanation about how the program does not cheat, and how to test/prove it. If you read the App store ratings, the low-starred ratings claim rampant cheating, but this is most clearly absolutely not the case. It’s just a very good backgammon engine.

As to the OP, I’m pretty sure I remember GNU Backgammon (which I think Backgammon HJ is based on) having a way to analyze board positions and giving you a ranking of the best moves. It’s been awhile since I’ve used GNU Backgammon, though.

For the first move, I’m getting 5/2 5/off as the best play, followed very closely by 5/off 3/off, followed by 6/3 5/off. (I think you have the numbers reversed from your diagram.) According to the analysis, you should win 94% of the time from that position. I guess you could have offered a double, but the analyzer is also showing that it’s a position too good to double, at least at the initial position. It’s also showing me that you should win with a gammon 30% of the time, and a backgammon 1.3% of the time.

Oh, and I haven’t had a chance to analyze the subsequent moves, but Google GNU backgammon, and you can set up the positions and dice and see if there’s anywhere you played sub optimally.

I’m not a professional, but I think I would have offered the double as soon as I was hit. He had three blots open and you had a huge pip lead. I think he would concede (and you no longer have the gammon equity to play for).

I wonder GNU would say about doubling strategy directly after getting hit.

ETA: Although now I see that his home board was very strong, so maybe doubling is unplayable there and you just got hit by the 5% of the time you can lose from that position.

You are correct, my mistake. But I absolutely would have offered a double right after he hit my blot. The gammon is almost certainly ruined, and he’s still got a way to go to beat me - so I’ll see if he’s got the belly for it.

Fair enough - I haven’t seen the initial position, but if he was unlikely to make a fifth home board point in the next 2 rolls, I might well do the same.

NJ : It analyzes the statistical pattern and previews the dice roll outs. This is as fraudulent dices in a real game because good real dices dont have strange patterns, otherwise If you play the dices manually, it is a median player. Test yourself !

I’ve tested it manually, and it works as advertised. I have no idea what you’re trying to say here.