vpFREE2 Forums

Some VERY Disturbing Statistics

vp_player wrote:

This is why the question was asked about given just a single data
point, how many standard deviations away from the mean would it
need to be before one would wonder about a machine being "gaffed".
My orignal response was 6 SD, but now, after having actually done
the math and graphed the curve, I wouldn't argue with a case being
made for 4 SD (only 1 or 2 RFs in 20.7 cycles) to make one wonder
about the fairness.

The "single data point" is extraordinarily unusual.

I see an alternate question being that given observation over the last
year, what is the probability of some even of similar rarity and
negative consequence being observed (and, as a result, give rise to a
post of similar nature to the original). Obviously that probability
is much higher.

In that context, does this event still raise concern of a "gaffe".
For that matter, might it not be reasonable to suggest that the
appropriate measurement period would be over the player's lifetime
play, for certainly had it occurred at any time during that period it
might give rise to similar question with equal concern.

I won't argue that this experience doesn't give cause to ask the
question about fairness. After all, it's a certainty that the
postulated gaffe (double royal cycle extension) would give rise to
such an event with a decent probability (other potential gaffes that
would also be to the advantage of the casino, such as restricting the
probability of a quad, wouldn't).

As such, it gives space to question existance of a RF gaffe. However,
if you closely evaluate any short-term run of play it's likely that
you'll find some aberration in play results (flushes, quads, dealt 3
card SF's, whatever) that could prompt you to speculate on a specific
possible machine gaffe.

Given this big picture, I find it difficult to class this particular
observation as "very disturbing" (aside from the fact that the related
consequence of the hand shortfall in $ terms might be larger than
other possible scenarios not involving a RF ... though not necessarily
so, given higher win frequencies involved in other cases that amplify
their consequence).

I see only one significance: This frames a specific possible gaffe
that can not newly be tested in a fixed trial that will give rise to a
valid probability that this gaffe "hypothesis" can't be rejected on
the basis of the statistics -- hard evidence to submit to the casinos
gaming authorities. (And this is as good as it gets with any
statistical test.)

- H.

So, in other words, it would be extremely difficult for gaming control
to catch this type of gaffe, even if some players were complaining
about not getting their share of royals. Does that give you more
confidence that it isn't done?

It is my contention that this gaffe can be at least detected with some
probability, but not sufficiently for a court of law or gaming
control. I would say that once the Bayes' probability of gaffe exceeds
50%, it's time to start scouting for other plays. Complain to
authorities if you wish, but don't expect to convince any of the
validity of Bayes' theorem. The FBI uses it, but probably not lesser
enforcement. Maybe you can convince the Wizard of Odds (Michael
Shackleford) to take up your case.

···

--- In vpFREE@yahoogroups.com, "Harry Porter" <harry.porter@...> wrote:

I see only one significance: This frames a specific possible gaffe
that can not newly be tested in a fixed trial that will give rise to a
valid probability that this gaffe "hypothesis" can't be rejected on
the basis of the statistics -- hard evidence to submit to the casinos
gaming authorities. (And this is as good as it gets with any
statistical test.)

Harry Porter" <harry.porter@> wrote:

> I see only one significance: This frames a specific possible
> gaffe that can not newly be tested in a fixed trial that will
> give rise to a valid probability that this gaffe "hypothesis"
> can't be rejected on the basis of the statistics -- hard evidence
> to submit to the casinos gaming authorities. (And this is as
> good as it gets with any statistical test.)

nightoftheiguana2000 wrote:

So, in other words, it would be extremely difficult for gaming
control to catch this type of gaffe, even if some players were
complaining about not getting their share of royals. Does that give
you more confidence that it isn't done?

It is my contention that this gaffe can be at least detected with
some probability, but not sufficiently for a court of law or gaming
control. I would say that once the Bayes' probability of gaffe
exceeds 50%, it's time to start scouting for other plays. Complain
to authorities if you wish, but don't expect to convince any of the
validity of Bayes' theorem. The FBI uses it, but probably not lesser
enforcement. Maybe you can convince the Wizard of Odds (Michael
Shackleford) to take up your case.

Well, you caught be in my own "gaffe" (drawing on the alternate
meaning of the word). I need to edit my fast-fingered flurried posts
a little more carefully. The brain spoke "that can be now ...", the
fingers substituted "not".

So, no, I don't think it's extremely difficult for gaming officials to
detect such a gaffe (meaning, in this case, an intended skewing of the
machine results) in most real-life cases. If they're competent and do
a decent investigation into a complaint that is appropriately backed
up by statistics, it shouldn't be too much problem at all.

Both those criteria are sticklers, however.

I've no confidence that gaming personal are inclined to carefully
examine a complaint to assess whether there's some meat to it or if
it's just coming anecdotally from someone who happens to be doing
poorly in their gaming. Like you, I see room to question the
statistical competence of gaming officials (at least the "front line"
to whose attention your complaint might first rise).

As far as the complaint itself, I'll again suggest that someone who
says "I've had no royals in x cycles of play" doesn't offer up
anything firm to warrant investigation. However, if on the basis of
that observed experience they form a hypothesis of a possible gaffe
that would give rise to those results with a decently strong
probability, and then proceeds to test that specific hypothesis
against a defined trial going forward, then the outcrop of that test
does present something that can be handed over to gaming with some
credibility as cause for further investigation (assuming that those
statistics satisfy probable cause for continued suspicion).

As you suggest, it may be necessary to place those results into the
hands of someone with sufficient credentials to add weight to them.

- Harry

--- In vpFREE@yahoogroups.com, "nightoftheiguana2000"
<nightoftheiguana2000@...> wrote:

> Bottom line:
> I agree with you when you said, "The important thing is to

understand

> what the tests that were done mean. (how to interpret them)."

So,

> how would you interpret the Red Rock data? IMO, there is nothing

to

> interpret, in the same way that there is nothing to interpret if

I

> just keep playing until I am 3 standard deviations behind.

At what point would you begin to suspect an altered machine?

It's hard to answer such a subjective question. I can say this,
though. If I did start to worry about whether my results
were "legit", I would do one of two things. If the game was not
important to me--for example, if I could walk across the street and
play a game just as good--I would stop playing it. What's the
downside of that?

But if an equally good game was NOT available elsewhere, then I'd set
up some kind of test. Because I'm already somewhat suspicious of
this game, I'd shoot for a test where a threshold bad result has
something like a 10% chance of happening by luck. For example, for
25c FPDW I could say I'll play another 37 hours, but if I ever fall
$1000 behind, I'll give up on this game. Falling $1000 behind at
some point during 37 hrs of play has about a 10% chance of happening
by chance alone. If I use 10% events as my criteria, I will be
walking away from perfectly good games 10% of the time. That seems
like a reasonable approach for a serious player.

Note that the proposed test has a specified starting point, a
specified end point (loss of $1000 or 18,500 hands, whichever comes
first) and a specified criterion for deciding the game is not
playable. The probability of satisfying the criterion is understood
ahead of time.

One could also use a test that relied only on the end result of a
specified number of hands, but that would be difficult to follow
through on if (1) you are already worried about the game, and (2) you
start off losing quickly.

--Dunbar

···

--- In vpFREE@yahoogroups.com, "dunbar_dra" <h_dunbar@> wrote:

Agreed.

···

--- In vpFREE@yahoogroups.com, "dunbar_dra" <h_dunbar@...> wrote:

If I use 10% events as my criteria, I will be
walking away from perfectly good games 10% of the time. That seems
like a reasonable approach for a serious player.

But if an equally good game was NOT available elsewhere, then I'd set
up some kind of test. Because I'm already somewhat suspicious of
this game, I'd shoot for a test where a threshold bad result has
something like a 10% chance of happening by luck. For example, for
25c FPDW I could say I'll play another 37 hours, but if I ever fall
$1000 behind, I'll give up on this game. Falling $1000 behind at
some point during 37 hrs of play has about a 10% chance of happening
by chance alone.

It's interesting that, on my own, I've done exactly what you
suggested. We played an additional 18,390 hands in the week since I
posted the orginal message and lost $1,250. Consequently, we're done
playing at Red Rock.

Thanks for all the comments.