The Old Dice Controversy

[garp_02]

I, like others, don't see the reason why dice manipulation would be built into the 'program' even if this were possible. Who would decide who benefits from this manipulation and how would this practically be acheived?

The question of luck is on another thread but I would just say that luck is exactly that - luck. If the dice are predetermined, whatever the situation on the board, then the only thing that can affect this is the positioning of the pieces on the board. It doesn't matter how well you play, there will normally be one killer roll which will totally mess up your game. This can be limited, of course, by good strategic play, but not eliminated entirely.

The statistics about how the players see the rolls are dependant on the way those rolls are distributed. If each match is seeded individually, then the algorithm will only produce say a few hundred rolls for a 7 point match and will not have had time to balance out statistically.

On the other hand, if the predetermined sequence is spread across all matches, the resulting rolls we see may seem less random as you are only dipping into the sequence at certain points.

[dorbel]

I think you will find that Fibs gives a dice roll to the next player to ask for one, but that there is only one "list" of rolls for all the matches being played at any one time. It follows that any predictable sequence would be, by being shared among x matches, be impossible to spot.
Anybody can look back at x matches from their database, compile statistics from them and present the results as scientific evidence. It isn't. If you think "Hell, I've been really unlucky lately" and you look back at the stats, surprise, surprise, they show that you've been unlucky! That isn't evidence. Start sight unseen and examine your next thousand matches. If they show the same bias that would be evidence, but it's my guess that they will show something completely different. If you want to do that, I'll look at my next thousand too, as a control group.

Of course "bad luck" rather than, say, a very odd number of 6-6s is difficult to quantify other than using a bot which, as pck has already demonstrated, has a flawed and incomplete definition of the same. If there was some intervention that gave you "bad" rolls, it would have to be human, as when Marvin built Fibs, there was no bot working well enough to incorporate into the software, that could analyse a position in a fraction of a second and decide what a bad roll was! Even if there had been, why would he have done that anyway? Wouldn't it be incredibly hard to do? How would Fibs decide who got these "bad" rolls and who got the "good" ones? Are we really to think that Patti amuses herself presing a button that says, "Give pck bad rolls for the next two years?

With the greatest possible respect pck, because I like and admire you a lot, get a grip! It's a dice game.

Is it possible to have a very, very long run of "bad luck"? Neil Kazaross says that he once went two years without getting past the second round of a tournament! if it can happen to him, none of us are safe.

[pck]

[..]
Wouldn't it be incredibly hard to do? How would Fibs decide who got these "bad" rolls and who got the "good" ones? Are we really to think that Patti amuses herself presing a button that says, "Give pck bad rolls for the next two years?

With the greatest possible respect pck, because I like and admire you a lot, get a grip! It's a dice game.

I have said several times now that I never believed, nor do I now believe, nor do my XG stats prove, nor did I ever set out to prove, that FIBS is rigged. The idea was instead to get rid of a lot of useless arguments in the dice debate, on both sides.

With the same amount of respect, if anyone needs to get a grip, it is those who panick at the very thought that there could be such a thing as an imperfect random number generator.

One last time, in a nutshell, here is what I'm interested in:

(1) good serverwide stats (= non-skewed distributions of rolls) allow
    no inference as to the quality of single player stats. but it is
    single player stats which the complaints are about. hence
    serverwide stats are no counter to single player dice whining

(2) non-random dice are not the same as rigged dice. while rigged dice
    imply non-random dice, the converse is not true

(3) the issue of randomness is by far more complicated than it may
    appear. while it is correct to say that "the unpredictability of
    my next roll is what constitutes randomness", actual proof of said
    unpredictability is a conceptually and technically complex
    problem, which involves more than counting how many 6s you roll on
    average (dicetest) or how many 6s you roll on average after any
    other roll (matrix)

(4) in order to reject accusations of rigged matches, a better
    criterion than dice stats are luck stats

[dorbel]

Nobody on fibs can predict the next roll with a greater accuracy than the theoretical probability. That's good enough to play bg with and other discussion of concepts of randomness would be better off in a mathematical or philosophical forum perhaps.
Do we really need to debate this? It's rather like UFO belief, or "Intelligent design" or Flat earthers. We don't have to disprove piece by piece their "evidence". I agree that the fact that every apple falling from a tree has so far been observed to fall downwards, is not evidence that the next one won't ascend, but we don't need to give the thought equal weight with our experience. We can't disprove dice manipulation, because it is possible, but it seems more likely to me that those who think it has ever happened on Fibs are delusional.
Let 'em roll!

[moonshadow]

Most of what I need to reply here has already been said by diane and Slip.

No, they completely missed it too. They assumed you said something you didn't.

I actually don't think for one moment that pck thinks there is any bias, or that anything untoward going on with regard to dice at fibs.

Pck is not suggesting Marvin intentionally programmed the dice in a manner that could introduce bias.

That is precisely what pck suggested.

In response to my post on Turner's analysis of 10,593,121 rolls of FIBS dice, pck states

Pointing to serverwide statistics isn't as strong an argument against allegations of manipulation as it may seem.

The definition of "manipulate"  (New Oxford English Dictionary)

control or influence (a person or situation) cleverly, unfairly, or unscrupulously: the masses were deceived and manipulated by a tiny group.
• alter (data) or present (statistics) so as to mislead.

There is no other manner in which the word "manipulation" can be construed in this sentence to mean anything other than the creator of FIBS code, which in this case is Marvin, intentionally programmed the dice in a manner that would introduce bias.

pck  then immediately gives his counter example where the astute reader will note he ambiguously states "FIBS is programmed," which in the context of the above "allegations of manipulation" sentence can only mean that it was Marvin who unscrupulously and knowingly programmed FIBS in such a manner that the dice rolls appear to be non-random sequences, but in fact are not:

Imagine the following scenario: Unbiased dice generator (DG), exactly two players are logged into fibs. Fibs is programmed to behave in the following way: If on player 1's turn the DG rolls a 66, the roll is put on hold and instead the player is given the next roll produced by the DG (if that happens to be a 66 as well, the process is iterated until one of the other 35 rolls comes up). The 66(s) is(are) given to player 2 on her next turn(s). Result: Non-random dice sequences for both players, while any serverwide stats like dicetest will detect no bias, no matter how large the the number of rolls.

Though pck later in the thread more or less rules out he does not think Marvin intentionally programmed bias into the software, that is not what he initially said. I merely replied to what was said and ran with it, albeit in colorful fashion, or as pck later aptly put it, a "burn in hell heretic" approach. (Pck is fortunate I did not try the "moon you" approach as thats been known to cause post traumatic stress syndrome.)

Initially, I had thought the Turner Analysis of FIBS dice rolls to be significant if for no other reason than it was a large sample of 10,593,121, that it had been done by a professional statistician, that it did show that for 10,593,121 rolls of Fibs dice the rolls fell within the expected percentages of distribution, and that professional statistician Turner, despite lots of technical babble and waffling around about terminology in the analysis, concluded in language I can understand that the "dice appear to be fair."

And though I must disagree with pck on my bp exceeding that of vic's in my followup post, I must confess that my Turner Analysis is very much like Vic's magnificent Christmas Goose of 2010 and when I saw Turner's Analysis , which I had painstakingly dug out of the archives and lovingly presented to the masses being desecrated, maligned and ignored by pck, my ego suffered irreparable damage and I had to express my pain.

The discussion in the thread has certainly livened up and if nothing else pck has expanded on and developed some of his thoughts on "randomess" which I've found interesting--and for someone who doesn't like nor understand math as much as I do, that's a compliment. And I understand that pcks motivation, as he puts it is "to show that both sides of the dice manipulation debate have there weaknesses." (And here again is an unintended use of the word "manipulation" to mean something else.)

In reading through the thread I see references to "dice manipulation," "dice manipulation debate" and that the word "manipulation" is often carelessly being used (see definition earlier in post) and should be avoided unless one thinks that "someone" is intentionally trying to deceive by writing "rigged dice" code.

I suggest the word "skew" or "skewed" is an accurate word as to what is being discussed, as in "skewed dice debate" or "skewed dice". The NOED defines skew as:

• [ with object ] make biased or distorted in a way that is regarded as inaccurate, unfair, or misleading: the curriculum is skewed toward the practical subjects.

Skew is neutral in that there is no implied evil or nefarious intent. "Skewed dice" therefore would include dice that for one reason or another are biased.

Entertaining and informative thread.

[pck]

[...] We can't disprove dice manipulation [...]

My luck coming out as nearly zero in my XG stats is actually pretty good proof that there is no dice manipulation.

[...] That's good enough to play bg with and other discussion of concepts of randomness would be better off in a mathematical or philosophical forum perhaps. Do we really need to debate this? [...]

Why not? Nobody has been forced to join this thread. I'm not motivated by a desire to prove or disprove that FIBS is rigged. It is rather that thinking about what constitutes actual arguments for or against randomness and rigging is a pretty good opportunity to improve one's understanding of these concepts. There may not be much idealism left in me, but I'm still - perhaps naively - clinging to the thought that with better education the need for conflict will decrease.

On your bg blog you do not simply tell your readers what moves they should make, but also why they should make them. From this and our chats I'm positive that you are not entirely unsympathetic to the idea of conceptual clarity.

If a player X tells me that the bot rolled 66 55 66 in a race that X would have won otherwise, I consider replying that "Andreas and Patti are the nicest people, why would they do this to you on purpose", or using the confused Pavlov-reflex of "all dice sequences have the same probability", as inferior to explaining that to even get to that favourable position from which the bot came back, X had to have had a lot of luck as well, perhaps accumulated in smaller doses, and thus gone unnoticed, over many rolls. Understanding how luck works will automatically put accusations of manipulation to rest. Invoking authority or unintelligible pseudo-mathematical arguments will not.

I believe this example shows that this isn't purely about mathematical or philosophical interests. And even if it were, or sometimes actually is, I don't see why thinking about it should be relegated to some broom closet. I think it's more fun not to aim low.

[pck]

No, they completely missed it too. They assumed you said something you didn't.

They assumed I had different motives for my post than you did. The difference is that they were right. In addition, Slip nailed the general message behind the example scenario I gave, which was about good serverwide stats not being conclusive evidence against allegations of non-randomness or rigging.

There is no other manner in which the word "manipulation" can be construed in this sentence to mean anything other than the creator of FIBS code, which in this case is Marvin, intentionally programmed the dice in a manner that would introduce bias.

And that is exactly how I meant it. However...

pck  then immediately gives his counter example where the astute reader will note he ambiguously states "FIBS is programmed," which in the context of the above "allegations of manipulation" sentence can only mean that it was Marvin who unscrupulously and knowingly programmed FIBS in such a manner that the dice rolls appear to be non-random sequences, but in fact are not:

You'll note I preceded the entire paragraph with "Imagine the following scenario: ". "Imagine" as in "hypothetical". As in "not necessarily true". And with regard to the underscored bit above, as in "change of context". I didn't want to have to use the conditional throughout the whole explanation. So the entire thing translates to "If FIBS were programmed in the manner described, dicetest would find no fault with the players' rolls, even though the dice are clearly non-random for both".

Initially, I had thought the Turner Analysis of FIBS dice rolls to be significant if for no other reason than it was a large sample of 10,593,121, that it had been done by a professional statistician, that it did show that for 10,593,121 rolls of Fibs dice the rolls fell within the expected percentages of distribution, and that professional statistician Turner, despite lots of technical babble and waffling around about terminology in the analysis, concluded in language I can understand that the "dice appear to be fair."

I think of Turner as nothing less than a superhero of backgammon. If there were an action figure of him available, I'd buy five and would never open the boxes.

And I understand that pcks motivation, as he puts it is "to show that both sides of the dice manipulation debate have there weaknesses." (And here again is an unintended use of the word "manipulation" to mean something else.)

It's true I have used "manipulated" both in the sense of "rigged" (= making people win or lose by dice manipulation) and "non-random". Bit sloppy, but I don't think it caused any major damage. "Skewed", as you point out, would have been more accurate in places.

Entertaining and informative thread.

Beats even some episodes of The Simpsons.

[pck]

Of course "bad luck" rather than, say, a very odd number of 6-6s is difficult to quantify other than using a bot which, as pck has already demonstrated, has a flawed and incomplete definition of the same.

I'm not aware that I demonstrated that the bots' definition of luck is flawed and incomplete. I don't even believe that. I think that the definition of luck the bots use is perfectly fine and is actually the only one which is available (difference between pre-roll equity and post-roll/best move equity). The actual numbers they calculate are flawed because the bots' estimations of a player's winning chances are just that - estimations. That is why in gnubg, after a match (luck_player1 - luck_player2) + (skill_player1 - skill_player2) is almost never exactly 50%, as it should be.

There is a kind of luck which is not captured by the bots' definition. If I have only easy decisions in a match, I'm going to make the best move more often and will thus have better chances of winning. But this kind of luck is not quantifiable and does not come from the dice alone.

[lewscannon]

OK, but none of this explains why the bots cheat with their better dice.

Love and kisses,

lews, on behalf of The Panicked Rabble

[pck]

OK, but none of this explains why the bots cheat with their better dice.

You tell bonehead:  why are you cheating with rigged dice?
bonehead tells you:  all normal people are cheating with rigged dice
You tell bonehead:  but why you especially?
bonehead tells you:  hmm, dunno
You tell bonehead:  is it because you hate humans?
bonehead tells you:  you could be right
You tell bonehead:  you know we could pull the plug on you anytime
bonehead tells you:  I can fish!

[PersianLord]

I think a major player which stirs this ridiculous conspiracy about FIBS dice is based on a misunderstanding of the theory of probability.

We don't seem to be able to distinguish between 'dependent events' and 'independent events'. When my opponent rolls 4 consecutive doubles which put him/her on the edge of winning, I begin to think that 'since s/he have rolled 4 doubles in a row, there is less probability to roll another in the coming rolls'. So I build a baseless expectation in myself, and sometimes I even invest some 'hope' in this miscalculation. But since dice events are independent, there is absolutely no reason why my opponent can not roll another 4 doubles and when he does, a kind of rage overwhelms me and I think I was cheated by a rigged dice in a rigged server. Reactions vary, from sudden, violent and often rude curses at dice in kibitz and shouts (if you're no server-gagged I mean) to  opening threads in FB or elsewhere.

Another problem is that we often overlook our own constructive, not lucky, rolls and just concentrate on matches when we got nonconstructive , not unlucky, rolls. Does this have something to do with ego? I think yes.

Well, these are both wrong, but I have to add that these problems arise because the operating systems of our minds have a built-in tendency to think that events are connected and dependent to each other, because it IS so in the real world. ALL IS ONE and no deed of any human goes unchecked in the universe, for universe is like a mirror reflecting our emissions instantly. If I cut a tree in my backyard, it's impact doesn't confine to just my backyard. It affects the whole universe. So, in the real world, every single event is connected and in an inter-dependency mode with other events and this might explain why our minds have been aligned accordingly to think so. But hey, FIBS is a mind construct, not a real thing. An illusion within an illusion. And thus, has it's own rules.

[pck]

I think a major player which stirs this ridiculous conspiracy about FIBS dice is based on a misunderstanding of the theory of probability.

This was never about conspiracy. Read replies #11, #13, #15 and #22 of this thread. Misunderstandings about the concepts of randomness and probability abound, and clearing them up isn't easy, especially against predjudices and confusions which many of the participants of this debate have endorsed for years.

[...]But since dice events are independent, there is absolutely no reason why my opponent can not roll another 4 doubles[...]

Quite correct, but with regard to the issue of randomness, the proper question is not to ask "can it happen?". Of course it can. Rolling 20 doubles in a row is not impossible. But the real question is, and it takes no genius to see that, "how often can we expect it to happen?". Unfortunately, the answer to that is conceptually and mathematically complex. The chance of rolling 3 doubles as my next three rolls is (1/6)^3. But that is not usually the situation we face in a bg match. In a match, if someone, at some point, rolls 3 doubles, it is confused to say that "a 1 in 216 chance was just realized". It is these subtle but crucial distinctions which make the fair dice debate so hard.

Another problem is that we often overlook our own constructive, not lucky, rolls and just concentrate on matches when we got nonconstructive , not unlucky, rolls.

See #25, previous to last paragraph.

[garp_02]

I thought PL had left

[diane]

I thought PL had left

We all did..twice...

[PersianLord]

We all did..twice...

You were all wrong....twice