Hello all,

None of the external dice dll's provided by Extreme Gammon (in the dice dll folder) work - the program complains they are not in the right format and insists upon switching back to the internal dice.

I've followed the instructions in the user guide to compile my own routine via Turbo Pascal but the same error is generated.

Has anyone managed to write their own dll or found one that the program is happy to use?

Many thanks.

Why do you need to use external dice?

Too many 2-in-36 shots being landed by the bot.

The number of times my first 3-2 roll using 24-21, 13-11 is met with a 6-4 response......you don't want to know. I just want to satisfy my paranoia

I know! I'll build the greatest engine of all time for a game and then have it cheat!

It's a fantastic bit of software and I never suggested it was cheating? Rather surprised you'd publish something like that on a public forum, that's probably actionable?

As I said, it's just something I'd like to do to satisfy my paranoia.

I have been playing this for a couple of years now and I never use the internal dice generator.

Of the three external dice generators only DiceLib.dll and MT19937.dll still work. RandomOrg.dll used to work but stopped some time ago. In my opinion even the external dice can be sketchy.

A good test it to predict to yourself your opponents dice rolls based on what you perceive you weaknesses to be. If you're calling the rolls correctly more often than not, particularly in the one game.... well.

Many times I will play entering manual dice rolls. It is time consuming but there is no fairer game and the wins-loses are more balanced. One game I played the other night, it was a fifteen round match and I was getting hammered 10-2. Change to manual dice and I ended up winning.

I also watch backgammon tournaments on YouTube and quite frankly whilst there are lucky rolls and all that I have never seen replicated in a live match what XG2 can pull off for itself.

Just my observations….

I've been using Extreme Gammon for 9 years now and see nothing unusual about the dice.

Just my observations ...

If you've never seen crazy rolls live, you haven't played/watched enough live backgammon. The craziest sequence I've ever seen was in a tournament I played a few years ago. I believe the sequence was 66, 66, 55, 66, 44 to close out a spectacular come-from-behind win. If that sequence had happened online, players would have sworn it was rigged...these things happen, though. (NOTE: I was not playing in the match. I just witnessed it.)

Then I might respectfully suggest that you've been playing with your eyes closed.

I read a lot about how "lucky" dice rolls can be explained by "the computer is playing at a high level and has positioned its pieces accordingly".

It doesn't wash with me. Sorry.

2-in-36 rolls anything less than twice in thirty six rolls over a prolonged period of time tells me there are rigged dice involved.

My stats vs Expert mode using internal dice: 25 wins, 70 losses.
My stats vs Expert mode using external dice: 53 wins, 22 losses.

Maybe I'm just cynical but the joy with facts are they are set in concrete.

I'd have to see proof before I believe they were rigged and your win/loss record is not proof. Just off the top of my head, I can think of several issues with that:

- Less than 200 games is nowhere near a suitable sample size to determine skill levels, and it's certainly not enough to prove dice are rigged. Let's see some statistics on the individual dice rolls.

- When did you play those games? It's very possible your skill level increased when you started using the external dice. If nothing else, given your pre-conceived notions about the dice, you certainly could have played with more confidence with external dice.

- You're assuming the internal dice are the dice at fault only because you lost with them. Perhaps it's the external dice that are biased, allowing you to win more than you should.

The best way I know of to truly test the dice is to analyze rolls. For each move, rank the potential rolls from best case to worst case scenario. Sometimes you roll your best potential roll; sometimes you roll your worst; usually, you roll somewhere in between; but over several thousand rolls, you should approach a ranking where your average roll is of average luck. Most other types of evidence are simply anecdotal.

I have seen crazy live rolls but not to the extent that XG2 internal dice or external dice can generate....

...but when you end your post with "Just my observation", you are already admitting you have absolutely no proof that the dice are not fair. It's just your perception, which, like all humans, is inaccurate. Without proof, I simply don't believe the dice are not fair and I haven't seen any proof whatsoever.

I am more than happy to play 200 games with internal dice rolls and then 200 games with external dice and report back here? Is that what you want?

As an aside, given that draughts has been solved as an AI experiment, surely it must be possible to program something to resolve every single permutation of backgammon positions so that every possible move is calculated optimally?

I'm more concerned about the number of rolls (for both the player and the opponent) than the number of games, but 200 games for both external and internal dice should produce enough rolls to show how often each roll is optimal (jokers), worst case scenario (anti-jokers), and average.

I'll take the forum view on this - over 200 games, how often would the bot be expected to repeatedly hit a 2-in-36 roll?

In my experience, anything involving me making a 13-11 first roll gets consistently screwed with a 6-4 roll.

I'm going to assume that's a rhetorical question.

When I play 13/11 on the first roll, XG hardly ever hits me.

Just my observation ----

I was joking with my other post, but I'll try to actually answer it here.

Asking "how often would the bot be expected to repeatedly hit a 2 in 36 roll?" is asking the wrong question. The real question is, how often does something rare happen more often than it should? The answer to that is, all the time.

For you, you find you get hit with an 18:1 shot too often. For me, in one site, it seems I often roll a 65 after my opponent opens with 13/7 8/7. On another site, it seems my opponent opens a disproportionate number of games. I've heard somebody claim they believe there are more doubles than there should be on a site. I've heard somebody state that their opponents seem to roll 66 or 55 in races too often. I've heard others say their opponent opens with 31 against them too often. There are literally millions of rarities that can and will happen.

So, to answer the question you should be asking, almost everybody notices (or perceives) that a rarity happens too often. The odds are, there will be a rarity that happens too often because there are so many rarities. One or two of them are bound to happen to you.

LMK when you're ready to play agaisnt XG for money.

Is there some algorithm masquerading as what we humans perceive to be "luck" that influences dice rolls? What is the purpose of the luck reference in the bottom left hand corner?

Ok, I've calmed down a bit now that I am on the point of creating my own DLL for use with ExtremeGammon. Took a while but I'm almost there.

I'll play 200 matches to 25 and see how that pans out.

As an aside, is it possible for someone to just be unlucky, even over a prolonged period?

I ask because whatever length of match I play ExtremeGammon, my luck rating is either quite unlucky, unlucky or very unlucky.

Ditto GnuGB, cpu is almost always "go to Vegas", I am "go to bed".

So you have 78 wins in 170 matches = ca. 46% wins against a bot that is probably better than you. That’s like....the expected outcome...

Reminds me of a player on Galaxy who complained about the ridiculous amounts of doubles he had in a match and it matched EXACTLY the long term expected distribution.

Complaining about rigged dice seems like some sort of masturbation for some people...

It is incredibly unlikely that a player will be lucky or unlucky for an extended period of time, but I don't put a whole lot of stock in the luck rating, anyway. It tends to be an assessment of a player's skill more often than an assessment of their luck. The better the player, the more "lucky" that player is rated, usually.

No one knows the answer to either question?

The 'Luck(Joker)' entry in the performance table gives a rough idea of who had the better luck in the match. As I understand it, for each roll it calculates your equity after each possible roll, assuming you made the best play with each roll. The difference between the equity after your actual roll and the average equity over all rolls is your 'luck' score for that roll. ('Joker' just means a roll that was an unusually large swing from the average roll.) Total up your luck for all the rolls in the match and you get a luck score.

It doesn't really tell you much that you wouldn't already know from looking at the match. If you win a 15-point match 15-0 against an equal player, your luck score should be extremely high, maybe in the 8-12 range. A close match against an equal player would have a very low luck score.

If you're spending a lot of time thinking about lucky and unlucky dice (as many of the people in this thread clearly are) then you're missing the point of backgammon. If you study the game and improve your play, you'll win more matches. If you spend your time instead fiddling with external dice generators and compiling elaborate statistics -- well, that's time wasted when you could have been studying something useful.