vpFREE2 Forums

$1 FPDW Question

#1

No, I do not know where there is a $1 FPDW machine ---
just quarter machines! However a friend asked me a question
and I do not know if I answered it correctly.
  Maybe Bob Dancer can contribute to this problem. A friend
gave me a copy of an article Bob wrote early this year for
"Casino Gaming". It was about $1 FPDW and why those machines
are never around for too long. My friend asked if it was
possible to make $40 an hour on a $1 FPDW machine and how
much money (bankroll) did it take for those who dominated
these machines.
  I told him that we do know that the long term expected return
for FPDW is 100.762%. I am guessing Bob was estimating that
a good player, can play 1,000 hands per hour (1050 to be exact)
and that will result in $5,250 coin in per hour. The return
expected would be $5,290 and the difference is the $40 per hour.
Am I correct so far?
   I explained this was a long term expectation. I assumed
1,050 hands of play per hour using perfect playing strategy.
Two individuals playing 24 hours a day could play 25,200 hands
each day. They could play 756,000 hands if the machine were
to last 30 days and of they stayed on the machine all 24 hours
for those 30 days I ran 10 computer trial runs of 756,000 hand
each using Winpoker. I received an average return of about
$26,590 ($37.93 per hour) which is close to the to the expected
average return of $28,804 ($40 per hour). The range was from
a gain of $51,930 to a loss of $43,110. The biggest negative
drawdown was in this losing session and it was $44,305.
   I believed I could show the my friend how $40 per hour could
be made playing $1 FPDW. However, he was not impressed by the
bankroll requirement. The total hands played were 7,560,000.
Would this drawdown be typical? It seems as though only those
bankrolled heavily could justify $1 FPDW if and when they did
exist. On the plus side only one of the ten 756,000 hand
sessions was a loss. However, it could have been the first
session instead of the fourth session! The largest gain was
$51,930 and the smallest gain was $5,435. I do not think that
I convinced him to search the world for a $1 FPDW machine.
I know 10 trials are not many, but how could he ignore the
drawdown in the fourth trial run? Maybe I do not do this
little demonstration correctly!

Bob

#2

Your expectation will be just shy of $40 per hour at a 1000 hand clip. BR
requirements depend on what ROR you are willing to tolerate. Why don't you
play with Cindy Liu's VP analyzer to get an appreciation of bankroll.

Chandler

···

-----Original Message-----
From: vpFREE@yahoogroups.com [mailto:vpF…@…com]On Behalf Of
futrend
Sent: Wednesday, August 09, 2006 1:41 AM
To: vpFREE@yahoogroups.com
Subject: [vpFREE] $1 FPDW Question

  No, I do not know where there is a $1 FPDW machine ---
just quarter machines! However a friend asked me a question
and I do not know if I answered it correctly.
  Maybe Bob Dancer can contribute to this problem. A friend
gave me a copy of an article Bob wrote early this year for
"Casino Gaming". It was about $1 FPDW and why those machines
are never around for too long. My friend asked if it was
possible to make $40 an hour on a $1 FPDW machine and how
much money (bankroll) did it take for those who dominated
these machines.

Bob

vpFREE Links: http://members.cox.net/vpfree/Links.htm

Yahoo! Groups Links

#3

I know 10 trials are not many, but how could he ignore the
drawdown in the fourth trial run? Maybe I do not do this
little demonstration correctly!

Bob

Or your interpretation of it is inaccurate. Losing that much 1 out of
10 times isn't representative of its real chance. Do 1000 runs and
you might not have another loss that big again. You can't tell much
from 10 trials. What's reasonably possible is more relevant than
what's theoretically possible.

#4

Thanks Chandler,
  I used 1,050 hands in my demonstration to reach $40 and hour.
That part was fine. It was only when I tried to demonstrate
such returns with software trials did the extreme loss appear
in one of the ten trail runs. My friend did not think making
$40 an hour was so great when he saw the 4th trial drawdown at
a negative $44,305.
  However, this loss does relate to a $11,076.25 drawdown to a
$.25 FPDW game which is available to play. This was only an
exercise and I shared it with the group since the computer loss
was was so extreme. I think Tom Robertson is saying that drawdown
would not bother him as it might not relate to real life playing
on $.25 FPDW machine.
  I also know that I cannot play 1050 hands an hour perfectly.
I generally understand bankroll and ROR from all the great post
on this site. I took your suggestion and plugged $11,076.25
bankroll into Cindy Liu's VP Analyser as 8861 units on a 5 coin
$.25 game of FPDW. The ROR is 0% at 8,861 units and only 1% with
3/4 of the loss or 6646 units. So, would you say, just as I
think Tom is saying, forget it since the size of drawdown the
computer trial showed probably would not happen in real life.

Bob

Your expectation will be just shy of $40 per hour at a 1000 hand

clip. BR requirements depend on what ROR you are willing to
tolerate. Why don't you play with Cindy Liu's VP analyzer to get an
appreciation of bankroll.

Chandler

From: vpFREE@yahoogroups.com [mailto:vpF…@…com]On

Behalf Of

···

--- In vpFREE@yahoogroups.com, "Chandler" <omnibibulous1@...> wrote:
-----Original Message-----
futrend
Sent: Wednesday, August 09, 2006 1:41 AM
To: vpFREE@yahoogroups.com
Subject: [vpFREE] $1 FPDW Question

  No, I do not know where there is a $1 FPDW machine ---
just quarter machines! However a friend asked me a question
and I do not know if I answered it correctly.
  Maybe Bob Dancer can contribute to this problem. A friend
gave me a copy of an article Bob wrote early this year for
"Casino Gaming". It was about $1 FPDW and why those machines
are never around for too long. My friend asked if it was
possible to make $40 an hour on a $1 FPDW machine and how
much money (bankroll) did it take for those who dominated
these machines.

Bob

vpFREE Links: http://members.cox.net/vpfree/Links.htm

Yahoo! Groups Links

#5

Tom,
I am not quite sure what you are saying. It was just an
exercise so I am not interested in doing more trials on $1 FPDW.
  
I think you are saying the probability of this ever happening
again, or in real life, is extremely low. I assume you would not
be concerned if a one time extreme computer trial result appeared
in 10 trials of a game you might play such as $.25 FPDW.

Thanks,
Bob

--- In vpFREE@yahoogroups.com, Tom Robertson <thomasrrobertson@...>
wrote:

I know 10 trials are not many, but how could he ignore the

? drawdown in the fourth trial run? Maybe I did not do this

···

little demonstration correctly!

>Bob

Or your interpretation of it is inaccurate. Losing that much
1 out of 10 times isn't representative of its real chance. Do
1000 runs and you might not have another loss that big again.
You > can't tell much from 10 trials. What's reasonably possible
is > more relevant than what's theoretically possible.

#6

Yes, Bob, I'd say that losing $11,000 at $.25 FPDW is beyond the worst
reasonable loss possible. I'd feel safe playing the game even with a
somewhat smaller bankroll than that.

···

Tom,
I am not quite sure what you are saying. It was just an
exercise so I am not interested in doing more trials on $1 FPDW.

I think you are saying the probability of this ever happening
again, or in real life, is extremely low. I assume you would not
be concerned if a one time extreme computer trial result appeared
in 10 trials of a game you might play such as $.25 FPDW.

Thanks,
Bob

--- In vpFREE@yahoogroups.com, Tom Robertson <thomasrrobertson@...>
wrote:

I know 10 trials are not many, but how could he ignore the

? drawdown in the fourth trial run? Maybe I did not do this

little demonstration correctly!

>Bob

Or your interpretation of it is inaccurate. Losing that much
1 out of 10 times isn't representative of its real chance. Do
1000 runs and you might not have another loss that big again.
You > can't tell much from 10 trials. What's reasonably possible
is > more relevant than what's theoretically possible.

vpFREE Links: http://members.cox.net/vpfree/Links.htm

Yahoo! Groups Links

#7

Maybe an OS / systems expert can answer this.

I've noticed when I've run large simulations -- from several hundred
thousand to 1,000,000+ the results do not straddle the expected value
but are biased to the negative.

I seem to remember reading that there was a known bias in the native
Windows rng algorithm, at least in earlier versions. Is my memory
correct and is this still a problem in XP?

One caveat -- my total sample size of simulations is small, perhaps
50 - 100. So the appearance of bias might be simply that.

I'd appreciate an answer from anyone who knows for sure.

Thanks,

Bill

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

#8

0.999347^($11,000/$1.25)= 0.32%

···

--- In vpFREE@yahoogroups.com, Tom Robertson <thomasrrobertson@...> wrote:

Yes, Bob, I'd say that losing $11,000 at $.25 FPDW is beyond the worst
reasonable loss possible. I'd feel safe playing the game even with a
somewhat smaller bankroll than that.

>Tom,
>I am not quite sure what you are saying. It was just an
>exercise so I am not interested in doing more trials on $1 FPDW.
>
>I think you are saying the probability of this ever happening
>again, or in real life, is extremely low. I assume you would not
>be concerned if a one time extreme computer trial result appeared
>in 10 trials of a game you might play such as $.25 FPDW.
>
>Thanks,
>Bob
>
>--- In vpFREE@yahoogroups.com, Tom Robertson <thomasrrobertson@>
>wrote:
>>
>> I know 10 trials are not many, but how could he ignore the
>? drawdown in the fourth trial run? Maybe I did not do this
>> little demonstration correctly!
>
>> >Bob
>>
>> Or your interpretation of it is inaccurate. Losing that much
>> 1 out of 10 times isn't representative of its real chance. Do
>> 1000 runs and you might not have another loss that big again.
>> You > can't tell much from 10 trials. What's reasonably possible
>> is > more relevant than what's theoretically possible.
>>
>
>
>
>
>
>
>
>
>vpFREE Links: http://members.cox.net/vpfree/Links.htm
>
>
>Yahoo! Groups Links
>
>
>
>

#9

I'm not an OS expert, but do know that the RNG that is part of Excel
has a relatively short cycle. After 16,700,000 or so numbers, the
random numbers loop back through the same cycle.

Because of this, I hat to use an outside RNG for the bigger sims in
Dunbar's Risk Analyzer. (the RNG I use is based on the Mersenne
Twister.)

--Dunbar

Maybe an OS / systems expert can answer this.

I've noticed when I've run large simulations -- from several

hundred

thousand to 1,000,000+ the results do not straddle the expected

value

but are biased to the negative.

I seem to remember reading that there was a known bias in the

native

Windows rng algorithm, at least in earlier versions. Is my memory
correct and is this still a problem in XP?

One caveat -- my total sample size of simulations is small,

perhaps

···

--- In vpFREE@yahoogroups.com, Bill Coleman <vphobby2@...> wrote:

50 - 100. So the appearance of bias might be simply that.

I'd appreciate an answer from anyone who knows for sure.

Thanks,

Bill

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

#10

Maybe an OS / systems expert can answer this

....

I'd appreciate an answer from anyone who knows for sure.

Thanks,

Bill

There are so many places that a simulation can go wrong. Even those with decades of
experience running them often make errors. As a rule of thumb, whenever you make a
new simulation, it is a good idea to collect all the randomn number you uses in the
simulation and analyze them. You need to check thier basic statistics (such as mean and
variance) and their distribution (PDF, CDF) before you worry about an OS problem. You
ought to check their autocorrelation function also. Once you check the inputs (the random
numbers) you then do the same checks on the outputs (the distributions of generated
hands). If all that checks out, you can start to think about OS or environment issues. That
said, there are real OS problems in windows, I would recommend not using their built-in
math code for any serious simulation work. Same goes for microsoft office products.
Excell should not be used blindly be used for rigorous mathematical work.

So as for your request: don't rush off to get the the opinion of an OS expert yet, since you
don't yet know what the problem isn't.

[One very common issue to consider: Most RNG's produce numbers with an approximately
even or gaussian distribution with zero mean. You want evenly distributed numbers. In
this case, the user usually sets the range as between -num to + num (and the random
seed). The next step is map these numbers to set of outcomes relevent for the simulation
at hand. This is usually done using a CDF. But often, errors creep in exactly at this point,
regardless of how it is done. The issue is that the original random numbers span a
symmetric inclusive interval about 0 but the output space does not. When you map these
numbers to a set of say integers (representing particular hands) it is very easy to make a
mistake and add a bias if you aren't extremely careful in your work). But it easy to check
for this error by making sure that both the original numbers and the mapped hands have
the correct distributions.]

···

--- In vpFREE@yahoogroups.com, Bill Coleman <vphobby2@...> wrote:

#11

If you make mistakes that cost you 1 bet ($1.25) an hour, then the
11,000 bankroll RoR rises from 0.32% to 1.6%. (that's assuming 500
hands/hour; if you make mistakes totally $1.25 every 400 hands, the
RoR "jumps" to 2.4%)

I realize that perfect play was implicit in this thread, but I want
to demonstrate what I consider an important point. Players tend to
underestimate the impact of errors on RoR.

--Dunbar

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

0.999347^($11,000/$1.25)= 0.32%

--- In vpFREE@yahoogroups.com, Tom Robertson <thomasrrobertson@>

wrote:

>
> Yes, Bob, I'd say that losing $11,000 at $.25 FPDW is beyond the

worst

> reasonable loss possible. I'd feel safe playing the game even

with a

···

> somewhat smaller bankroll than that.
>

#12

Several years ago, I ran well over 100 simulations of 1.5M hands (a year of
fulltime play) of a certain game. Because they were taking a while, I was
doing them on two computers. At the end of each one, I recorded the result.

I noticed that the sessions run on computer 1 seemed random like you'd
expect. The sessions run on computer 2 seemd to skew way negative, not just
compared to the sessions from computer 1, but compared to the game's ER.

I wondered if this may be an effect of timing between the SW running the
simulator and the computer's pRNG.

I have no definite answer for you, but I've seen the same thing.

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

···

On 8/10/06, Bill Coleman <vphobby2@cox.net> wrote:

I've noticed when I've run large simulations -- from several hundred
thousand to 1,000,000+ the results do not straddle the expected value
but are biased to the negative.

#13

I think it is to be expected that simulations will be skewed, with more
results below the ER and fewer above. I believe this happens even for
simulations as large as 1-1.5 millions games. It's just a consequence
of most simulations having fewer than the expected RF (and other rare
hands) whereas a few have more than expected. I guess it depends how
skewed you mean. I don't know the usual ratio exactly, although I
believe a post long ago mentioned it, but lets say in 10 simulations
about 6 are negative and 4 positive, with the overall average very
close to the expected ER. I ran about 50 large simulations several
years ago and got a similar result.

···

--- In vpFREE@yahoogroups.com, "cdfsrule" <groups.yahoo@...> wrote:

There are so many places that a simulation can go wrong.

#14

Considering that the Mersenne Twister RNG algorithm is 4+ years old,
and the code is freely available, one would think that thisis the
algorithm all the Video Poker programs use; even the game makers.

On FVP, I remember seeing a disclaimer that the RNG used there may not
be the same as that used by the game makers. I took that to be a pro
forma disclaimer.

Why wouldn't everybody play to the same score?

I'm not an OS expert, but do know that the RNG that is part of Excel

has a relatively short cycle. After 16,700,000 or so numbers, the
random numbers loop back through the same cycle.

Because of this, I hat to use an outside RNG for the bigger sims in

Dunbar's Risk Analyzer. (the RNG I use is based on the Mersenne
Twister.)

···

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

#15

Bill Coleman wrote:

Maybe an OS / systems expert can answer this.

I've noticed when I've run large simulations -- from several hundred
thousand to 1,000,000+ the results do not straddle the expected value
but are biased to the negative.

I seem to remember reading that there was a known bias in the native
Windows rng algorithm, at least in earlier versions. Is my memory
correct and is this still a problem in XP?

One caveat -- my total sample size of simulations is small, perhaps
50 - 100. So the appearance of bias might be simply that.

I'd appreciate an answer from anyone who knows for sure.

I'm not writing from the expertise that you're looking for, but
perhaps this might shed some light.

Dean Zamzow made clear that winpoker wasn't written to be simulation
tool; just a tutor and practice program. It's my recollection that he
indicated the rng call was a rudimentary one. Furthermore the rng
isn't reseeded during autoplay and the resultant cycle is too short
for an appropriate simulation. I believe that he suggested that if
you were determined to use winpoker for analysis, aggregating multiple
shorter simulations would be a more practical method.

FWIW, while I can't comment from certain knowledge, I expect that some
bias would be evident in a low variance game such as JB well past 2mm
hands.

- Harry

#16

If I'm not mistaken someone posted a reference to some software that,
among other things, helps calculate pot odds. I can't find the thread.

Can someone point me in the right direction?

Thanks,

Bill

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

#17

Look for Poker Stove. Freeware.

···

On 8/13/06, Bill Coleman <vphobby2@cox.net> wrote:

  If I'm not mistaken someone posted a reference to some software that,
among other things, helps calculate pot odds. I can't find the thread.

Can someone point me in the right direction?

Thanks,

Bill

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