There has been much discussion on r.g.b. about the possibility of the FIBS
dice being skewed, so I thought I should put in my two cents worth.
I have been playing on FIBS for about three years, and have logged plenty
of experience points. It is my personal judgment from my playing
experience that the dice are not biased in any way.
How qualified am I to make such a judgment? I have been playing
backgammon for over 30 years, both money play and tournaments. In
addition I have been active in many other gaming activities which involve
chance all my life.
When one is involved in gambling, there is always the danger of
cheating. It is important to be on the lookout for irregular
occurrences, and to distinguish them from the normal lucky and unlucky
sequences which will always happen. On more than one occasion I have
sensed that things weren't right, and walked away from a situation which
could have cost me a lot of money had I stuck around. Thus, from my
experiences in the gaming world, I believe my judgment in this area is
better than most people's judgment.
Strange things do happen by chance. When I was playing blackjack for a
living, I saw several "unbelievable" occurrences. Once I lost 19
consecutive hands! (that really was an outprice -- I've never seen
anything close to that before or since). Another time, I went through an
entire 4-deck shoe without losing a hand. If you play enough, these
types of things will happen.
It is very easy to make the mistake of having a selective memory when it
comes to unusual occurrences. This is because we tend to remember the
abnormal, while forgetting the normal. We remember the unusual times our
opponent won an "unwinnable" race, while forgetting the many times
nothing special happened and we won the race we were supposed to win.
This is just human nature, and often accounts for suspicions that the
dice are rigged.
A good example of this occurred in duplicate bridge, when computer-dealt
hands were first put into use. There were several complaints that the
hands were more distributional than they should be. Extensive testing
demonstrated conclusively that the hands were not more distributional
than one would expect. What happened was that players tended to remember
the freak hands, while dismissing the more normal hands, and they came to
the subjective conclusion that there was something wrong with the
hand-generating program. It takes quite a bit of experience to be able
to put things into perspective and not be fooled by selective memory.
Now to the FIBS dice question. The claim that the dice are intentionally
programmed to be rigged in some way is patently absurd. What would
marvin's motivation be? He has no profit from FIBS. He created the
program because he enjoys backgammon and wanted a free backgammon server
to be available on the net. In addition, programming the dice to be
rigged is much more difficult than programming random dice. Any such
assumption about the dice being intentionally programmed to be rigged is
completely contrary to logic.
The question of the random number generator being faulty in some way is
another matter. This is a possible danger, of course. Usually when a
random number generator (or a routine which takes the output from a
random number generator) is flawed the error is either so small that one
would not notice it with the naked eye, or so large that it jumps out and
hits you in the face.
One can, of course, test various hypotheses about the dice. When the talk
about rigged dice started (there has been such talk for some time -- this
isn't the first thread along these lines), I attempted to test the dice as
best as I could. I had over 150 matches played on FIBS in my data base.
I ran through the dice rolls from these matches. All the numbers came out
within normal tolerance limits. The hypothesis that there were more
doubles than usual was not supported -- in fact, slightly less than 1/6 of
the rolls were doubles. Similarly, there was no indication that
consecutive doubles occurred more often than normal.
For those of you who want to do your own testing, you don't have to
wait around until you have played thousands of games. Last year, Mark
Damish had a program called Big_Brother, which was a match logging
program. It logged onto FIBS, and automatically selected, watched, and
recorded matches. I believe these matches should be available at
www.gammon.com and there are quite a lot of them, so anybody with the
inclination and good programming skills should have no trouble taking
these matches and performing appropriate tests. I'm sure everybody would
be interested in the results.
It should be noted that statistical tests can't prove anything one way or
the other; they can only give us indications. For example, suppose I
flip a coin 5 times in a row and it comes up heads every time. My
reaction would be: so what. If it came up heads 10 times in a row, I
would start to raise a suspicious eyebrow. If it came up heads 20 times
in a row, I would decide that the coin was biased and would have to be
convinced otherwise to change my view. Not that it is impossible for a
coin to come up heads 20 times in a row, but it is very unlikely if the
coin is unbiased. These numbers represent my personal judgments --
another person may have different thresholds.
It should also be pointed out how important it is to formulate one's
hypothesis before testing. Otherwise, it is easy to be misled by after
the fact reasoning. If someone announced to me that a coin was biased
and then the coin came up heads 10 times in a row, I would be quite
convinced. On the other hand, if the coin was first flipped 10 times and
came up heads every time, and then the person announced to me the coin
was biased, I would be far less convinced. The difference is that in the
second case his claim was an after the fact claim. The same is true of
dice testing. For example, suppose you looked at a million dice rolls,
found that 5-4 came up more often than any other combination, and
claimed that the dice were biased in favor of 5-4. This would not be
very convincing -- after all, some combination will come up most often
over the million rolls. If you made the claim before looking at the
rolls, it would be far more convincing.
I have no objections to people stating that they think the dice on FIBS
are biased, even though I disagree with them. It's a free world on the
internet, and such points of view do spark interest and perhaps will show
that there is some bias. I do hope, however, that these people will make
more of an effort to substantiate their case with proper testing, rather
than mentioning a few isolated runs of good or bad luck or expressing
their feelings that the dice are biased.
My main reason for writing this is to persuade FIBS players not to let
their game be affected by thinking that the dice are unfair. Once you
start to believe there is something wrong with the dice it is very easy
to think you should distort your play in order to compensate for the
dice, and the results will not be good.