vpFREE2 Forums

For Chief/ Machine's life, etc

#1

In a message dated 10/26/2002 3:25:35 PM Eastern Daylight Time,
chiefvrwc@aol.com writes:

I love "the machine's life" concept that my buddy Cardfather (and others)
keep referencing. Is that the life of the RNG chip? The CPU? The power
pack?
What happens if a machine dies prematurely in an accident? Does it's unused

royal flushes get left to another machine? Can I have them?

I don't think it's that difficult a concept. The paytable determines the %
payback on a VP machine. Each individual machine will eventually payback
THAT %, as long as it is in play. The 'lifetime' of the machine is determined
by how long it stays in play.
Check with Ernie Moody, the developer of Triple Play & other such games.
He explained this to me 4 years ago at the Gaming conference in LV.
CF

[Non-text portions of this message have been removed]

#2

I don't think it's that difficult a concept. The paytable determines the %
payback on a VP machine. Each individual machine will eventually payback
THAT %, as long as it is in play.

  Absolutely, positively FALSE. The payback by the machine over its
entire lifetime is governed by the same things that govern what each of
us gets over our lifetime. After, let's say 10,000,000 hands, the machine
will almost certainly, but not certainly, be within a tenth of a percent or
so of the expected return. Note that the expected return is NOT the return
we see as the perfect strategy return, since most casino players make
a large number of errors. Instead, it is the average of the expected return
for the hands as played (usually misplayed) by the patrons. This makes it
more or less impossible to predict what a machine will return while on the
casino floor. Casinos know the theoretical maximum return, but they just
take sample data to arrive at the numbers that reflect the actual return.

The 'lifetime' of the machine is determined by how long it stays in play.
Check with Ernie Moody, the developer of Triple Play & other such games.
He explained this to me 4 years ago at the Gaming conference in LV.

  You and he had a failure to communicate. He probably omitted, for
brevity, the idea that the return will, as the number of hands increases,
move closer (as a percentage) to the expected return. However, when viewed
as an absolute value, the machine return is actually expected to diverge
from the theoretical value.

  QZ

CF

#3

QZ's points are correct in a qualitative sense, but for VP 10 million hands
isn't very many in a statistical sense.

VP has very high variance, and this implies that it takes a truly huge
number of hands for the results to "tighten up" to this level of precision.
Assuming a game with a variance of 30, the standard deviation after
10 million hands is sqrt(30/10,000,000) = 0.001732. So, the standard
deviation is 0.17%, and 0.1% represents about 0.6 standard deviations.
This implies that the probability of being within 0.1% after 10 million
hands is only about 48%, which isn't even a concensus.

If we take "almost certainly" to mean a 95% probability that the results
will be within a tenth of a percent of expected return, then we need
two sigma to equal 0.1%, or:

N = 30/(0.0005)^2 = 120,000,000 hands.

If we use the "Ivory Standard" of (99 + 44/100)% sure, then we need
2.8 standard deviations, or:

N = 30*(2.8/0.001)^2 = 235,200,000 hands. At a rate of 1200 hands
per hour, this represents over 22 years of continuous play, without
a potty break. Youch!

So, following a single machine for its entire lifetime and measuring its
payback will not give results that are statistically significant beyond
"roughly speaking."

···

On Saturday 26 October 2002 11:53 pm, Quad Zilla wrote:

I don't think it's that difficult a concept. The paytable determines the %
payback on a VP machine. Each individual machine will eventually payback
THAT %, as long as it is in play.

  Absolutely, positively FALSE. The payback by the machine over its
entire lifetime is governed by the same things that govern what each of
us gets over our lifetime. After, let's say 10,000,000 hands, the machine
will almost certainly, but not certainly, be within a tenth of a percent or
so of the expected return.

#4

I don't think it's that difficult a concept. The paytable determines the

%

payback on a VP machine. Each individual machine will eventually payback
THAT %, as long as it is in play.

      Absolutely, positively FALSE. The payback by the machine over its
entire lifetime is governed by the same things that govern what each of
us gets over our lifetime. After, let's say 10,000,000 hands, the machine
will almost certainly, but not certainly, be within a tenth of a percent

or

so of the expected return.

QZ's points are correct in a qualitative sense, but for VP 10 million hands
isn't very many in a statistical sense.

VP has very high variance, and this implies that it takes a truly huge
number of hands for the results to "tighten up" to this level of precision.
Assuming a game with a variance of 30, the standard deviation after
10 million hands is sqrt(30/10,000,000) = 0.001732. So, the standard
deviation is 0.17%, and 0.1% represents about 0.6 standard deviations.
This implies that the probability of being within 0.1% after 10 million
hands is only about 48%, which isn't even a concensus.

  OK, I admit it, I pulled those numbers out thin air. I did not
realize it took SO MANY hands to get that close. Nevertheless, play
enough hands, and the return will get very close to the expected
return.

  QZ

···

-----Original Message-----
From: Steve Jacobs [mailto:jac…@…com]
Sent: Sunday, October 27, 2002 5:02 AM
To: vpFREE@yahoogroups.com; Quad Zilla
Subject: Re: [vpFREE] For Chief/ Machine's life, etc

On Saturday 26 October 2002 11:53 pm, Quad Zilla wrote:

#5

Yes, in a percentage sense. As you correctly pointed out in your
post, the plot of results in terms of absolute dollars grows broader
as more hands are played, and thus the uncertainty in the outcome
grows when expressed in absolute terms.

···

On Sunday 27 October 2002 01:47 pm, Quad Zilla wrote:

-----Original Message-----
From: Steve Jacobs [mailto:jac…@…com]

This implies that the probability of being within 0.1% after 10 million
hands is only about 48%, which isn't even a concensus.

  OK, I admit it, I pulled those numbers out thin air. I did not
realize it took SO MANY hands to get that close. Nevertheless, play
enough hands, and the return will get very close to the expected
return.

#6

--- In vpFREE@y..., Steve Jacobs <jacobs@x> wrote:

QZ's points are correct in a qualitative sense, but for VP 10

million hands isn't very many in a statistical sense.
(snip)>

So, following a single machine for its entire lifetime and

measuring its payback will not give results that are statistically
significant beyond "roughly speaking."

Time Out. What you, Brian and QuadZilla have said may all be accurate
but misses the point. The point of confusion is that what is relevent
is the size of the sample and has NOTHING to do with the physical
machine that Cardfather speaks of. If you are talking about a sample
of ten million, (or ten billion if you prefer) there is no
predictable difference whether you played one hand on each of ten
million machines or all on one machine. There IS no machine "life."
What would it be if it existed? The major componants of similar
machines are interchangable. If the RNG or CPU gets fried, a new one
is put in. Power unit, mechanicals, box are exchanged frequently.
Does a new "life" start when a new CPU goes into an old box? IF a
machine (however you want to define a machine) could be expected to
produce a fixed percentage over it's "life", there would be a lot of
very rich technicians. All they would have to do is review the
payouts of all their older machines and see which ones have paid back
less than the precicted percentage. Those machine would have a lot of
catching up to do in their payouts. Right?

Reality is that each individual hand stands on its own. That's the
building block. If I tell a statistician that ten million hands have
been played perfectly on a 10/7 DB game, he can tell me (within
whatever percentage accuracy I ask for) what the range of expected
results will be. He doesn't ask me which, or how many, machines were
used for the sample. Anyone disagree (except Cardfather<G>)?

#7

If you are talking about a sample
of ten million, (or ten billion if you prefer) there is no
predictable difference whether you played one hand on each of ten
million machines or all on one machine.

Agreed. My point was that it take a lot of hands to predict the
outcome within 0.1%, which is only a precision of one part in 1000.
After playing the requisite 120 million hands, the 0.1% represents
an uncertainty of plus or minus $150,000 on a quarter machine.

Reality is that each individual hand stands on its own. That's the
building block.

I completely agree with that statement. Machines don't
know what their expected result is, and when they stray from
the average they do not "gravitate" back to the mean. Their
outcome is from a random drift that isn't supposed to be under
any control, and Nevada law would frown on any machine that
did exhibit

If I tell a statistician that ten million hands have
been played perfectly on a 10/7 DB game, he can tell me (within
whatever percentage accuracy I ask for) what the range of expected
results will be.

I disagree. You can ask for 0.00000001% accuracy, but the
statistician can't tell you anything _meaningful_ to greater accuracy
than about 0.17% for a ten million hand sample. Sure, the statistician
can calculate some 14 digit number, but only the first three digits
will have any real meaning. Anyone who claims to predict the
outcome more accurately is either cheating or

He doesn't ask me which, or how many, machines were
used for the sample. Anyone disagree (except Cardfather<G>)?

Agreed, as long as we're not talking about machines that use
a crummy RNG which allows them to become predictable. A few
years ago, some guy won a bunch of money from a Keno machine
that had an RNG that was predictable under certain conditions. I
heard reports of his arrest, but never heard the outcome of the trial.

···

On Sunday 27 October 2002 01:56 pm, chiefvrwcx wrote:

#8

--- In vpFREE@y..., Steve Jacobs <jacobs@x> wrote:

QZ's points are correct in a qualitative sense, but for VP 10

million hands isn't very many in a statistical sense.
(snip)>

So, following a single machine for its entire lifetime and

measuring its payback will not give results that are statistically
significant beyond "roughly speaking."

Time Out. What you, Brian and QuadZilla have said may all be accurate
but misses the point. The point of confusion is that what is relevent
is the size of the sample and has NOTHING to do with the physical
machine that Cardfather speaks of. If you are talking about a sample
of ten million, (or ten billion if you prefer) there is no
predictable difference whether you played one hand on each of ten
million machines or all on one machine.

  Maybe there is some confusion over who wrote what here.
I never supported the idea of a physical machine timeline or any
similar theory. I agree with and was trying to say what you just
said.

  QZ

···

-----Original Message-----
From: chiefvrwcx [mailto:chiefvr…@…com]
Sent: Sunday, October 27, 2002 10:56 AM
To: vpFREE@yahoogroups.com
Subject: [vpFREE] Re: For Chief/ Machine's life, etc

#9

Steve Jacobs wrote:

Agreed, as long as we're not talking about machines that use
a crummy RNG which allows them to become predictable. A few
years ago, some guy won a bunch of money from a Keno machine
that had an RNG that was predictable under certain conditions. I
heard reports of his arrest, but never heard the outcome of the trial.

I can't remember where I read this but here's the gist of what I remember as
the explanation of what happened:
- it was in Canada
- the man was not charged with any crime
- he hit more than one very long shot keno jackpot
- he did not use any sophisticated mathematical analysis to predict the PRNG
- he did notice repeating patterns each day
- the machines were powered off each night and restarted every morning
- the machines were designed to run continuosly

The likely explanation is that the PRNG restarted at the same point after a
power on restart. Apparently, the fellow in question detected and exploited
the pattern.

#10

--- In vpFREE@y..., Steve Jacobs <jacobs@x> wrote:

If we use the "Ivory Standard" of (99 + 44/100)% sure, then we need
2.8 standard deviations, or:

N = 30*(2.8/0.001)^2 = 235,200,000 hands. At a rate of 1200 hands
per hour, this represents over 22 years of continuous play, without
a potty break. Youch!

Great googlymoogly! Maybe I do need to get lucky!! LOL Amazing....