With respect, it's peace of mind.

In 1983 (this is the UK I'm talking about) I was playing Psion Backgammon on the ZX Spectrum. Back in those days, you could buy disassemblers that would reverse engineer the code behind the game; if you could understand assembly language (I could) then you could see what was going on in the code.

Guess what? I found exactly where it was cheating.

Fast forward 20 years and I found a very good Backgammon program for the pc which seemed to enjoy a lot of luck with the rolls. The guy behind the program was adamant that the random seed was dictated by the system date/time.

So it was child's play to write a batch file that reset the date/time to a certain value before calling the program and, voila, the rolls were the same for each game.......except, the computer's rolls varied depending on the board position. I could play a game in a certain fashion and note the dice rolls......shut the program down, fire it back up again via the batch file and replay it but make slightly different moves at certain times and laugh as the cheating kicked in.

The guy tried to wriggle by saying that if the cpu evaluated two rolls identically, it used the random number generator to decide which play to make. Except, that policy should have advanced the rolls by one, not rolled itself 6-6.

Long story short, if I use my own DLL to roll the dice, I know the dice are real.

You probably need to have been exposed to cheating programs; I can guarantee your outlook on backgammon programs would be very different.

Just because ExtremeGammon and GnuGB allegedly don't cheat doesn't mean that all backgammon programs are honest.

I'm very confused as to why ever care whether a bot cheats or not

when I play with a bot it's to improve my game, and not to ponder whether it deals itself a joker once in a while to boost its winning chances from 80% to 85% or whatever

I think that's an excellent point. I understand that some people have fragile egos and do not like it when they lose, even to a bot, but when I play XG, my goal is to make the best possible moves. I'd rather lose with a very low error rate than win with a high error rate, any day. I can't even tell you what my record is against XG, actually.

You're both missing the point

If you're playing against a bot that is cheating, or that you suspect is cheating, you *won't* make the best move or improve your game.

Assume your opening move is 5-2 and you think your best move is 13-8, 6-4.

If your bot is cheating, or that you suspect is cheating, you'll think "what's the point of playing 6-4, it's only going to roll itself a 3 and send me to the bar?" So you end up playing safe with 13-6....which is not the best move.

Maybe I've been mentally scarred by cheating bots

I think you are mentally scarred by cheating bots, but your logic is off, anyway.

If you think the bot is cheating, you should still make the best move, no matter what. If it actually is cheating, it's going to roll what it needs to win either way. Besides, beating the bot shouldn't be your goal, anyway. It should be to learn. So, making a sub-optimal move simply because you think it's cheating is counter productive. You make the best move, it will roll what it needs to beat you. You make a bad move, it will still beat you.

I'll just leave you with this. The best live players in the world have used XG to improve their games - some tremendously. When they play people over the board, I think it's fair to say the dice are fair and they win using what they learned from XG. So, it simply doesn't even matter if XG's rolls are fair (which I'm sure they are), it's still going to teach you the best moves in most situations.

Thank you, I'd never seen it like that. I will adjust my play in future and see how it all pans out.

And a further thank you for acknowledging the mental scarring aspect.....it's very real, actual bot cheating vs perceived bot cheating is exactly the same when you've had experience of it.

It's why I use my own DLL on ExtremeGammon (I programmed a beep before and a beep after each roll so I know that it really is using my numbers).

Don't get me started on GnuDung.........

both XG and GNU literally tell you the luck factor at the end of the match, if either were somehow rigged against you it would constantly show a bot as very lucky

you can stop fighting windmills it's not a very healthy activity

I know, I seem to end up after every match as "unlucky", "quite unlucky" or "very unlucky" (XG) and "go to bed" (GNU).

The thing you're missing is this: if you've been bitten by a dog when you were younger and then a few years later you're bitten by another dog, you *will* tend to mistrust dogs.

Human nature, can't escape that.

I had a game earlier today against GNU (intermediate level). He had three men on the bar, I had five points covered in his home bed......he had just the six point covered in my home bed.

He rolled double 2......all three men came off of the bar and he hit my one blot with the other 2. I then rolled double six twice in a roll and so danced.

I have no idea of the odds of that happening (and I'd be genuinely interested in anyone who can tell me) but I suspect they're into winning lottery territory?

46,655 to 1.

the thing is, what you're describing is a virtual dog bite - it's entirely in your head and it only hurts if you think it hurts

just focus on your own blunders which I'm sure quite a few. once you run out of blunders come back and we'll talk about those lucky bots

also, why are you playing gnu on intermediate?

it shouldn't be set to anything less than world class. do you even run analysis at the end of the game?

Thanks for that, I'd be interested in your workings because I think you're missing two or three zeroes off of the end.

Probably my fault, not explaining the situation properly.

The only roll of any value to him was 2-2.....he rolled it, got his three men off of the bar and managed to hit the only blot I had elsewhere on the board. So now I'm on the bar but he's only got his six point primed....and I rolled double six twice.

To be honest with you, I was trying to work my way up and also to gauge how good I actually am (I have been lulled into an inflated ego because I repeatedly smash AI Factory's backgammon at the strongest level.....I now realise the AI is woeful and so I'm not as good as I thought).

No, did you read my (quite excellent) post?

This is very real, two backgammon programs that I found out *were* cheating.

So it isn't virtual, it's real.

Do you know what? I have downloaded the GNUBG source code and it doesn't matter which compiler I use, I can't get back to the released version's hash?

I didn't miss anything

you're completely missing a point of playing with a bot which is to become a better player. whether bot cheats or not (it doesn't) is completely irrelevant

that's why I said the bite is entirely virtual - as soon as you stop caring about jokers and focus on your mistakes it would not be able to hurt you anymore

It may seem real, but it's totally misplaced and perceived. Your only goal when playing the bots should be to make the best moves possible and have the lowest error rate possible- and, of course, learn from the mistakes you make. Those are the only things you can control. You can't control the outcome of the match. You can't control the dice. The only thing you can control is your decisions and worrying about winning or losing when playing a bot is totally worthless.

So, what you are experiencing is totally of your own making. You're worrying about things that are completely irrelevant and because of it, missing out on the most important thing you can gain.

This game is programmed for the computer to bring the dice it requires, even the external dice which are slightly better are still heavily biased.

One can run various tests to observe that.

In addition, some of the errors that the computer registers as "errors" are objectively speaking not errors at all.

I am highly disappointed by this game as it does not really teach you much of anything, certainly not if you have been playing this game for years or since very early childhood.

If you are a newbie, sure go for it, the analysis is pretty good, an advanced player however has nothing to gain by playing this game other than destroying their game in a futile effort to escape the computer cheating.

Each roll you describe has a probability of 1/36. The conjunction of the three rolls therefore has probability 1/(36)^3 = 1/46656 or in odds terms as already posted 46655:1.

Yes, those particular three rolls happening as they did is improbable, but the sequence of ANY three rolls is actually pretty improbable. You only noticed because of what happened in the game. You would not be on here commenting about how lucky you were if you had the same situation except that you had your three point open rather than your two point and you went on to win a gammon because you rolled 66 a couple times in a row. Something similar has likely happened to you, but that probably was chalked up as being due to your superior skill level.

It’s called “selection bias”, and it seems you may be suffering from it. I have seen much worse than what you described when playing with live dice. It sometimes works against you, in which case “the dice are rigged”, and sometimes it works for you in which case “I am a great player.”