I have just run 3 separate 9,999,999 hand autoplay simulations of All
Am. full pay 100.7221%. all 3 of them did win, but they each
underperformed EV by about 1/2, specifically 100.37, 100.37, and
100.39%. I would think that 10 million hands would come closer to the
expected EV than that ?? Another oddity, the total RFs for each was
201, 197, and 198, but the expected RFs are 230. Seems too far off ??
Maybe I just have a unlucky computer??
Thoughts Please
Jim
My computer is lazier than yours, but luckier. It ran only one
9,999,999 hand autohold simulation (I think it had other things it
needed to do), but it got 100.81% return, and 242 royals.
I believe your computer may need a stern talking to.
--- In vpFREE@yahoogroups.com, "jimnkelli" <jbecker11@c...> wrote:
I have just run 3 separate 9,999,999 hand autoplay simulations of All
Am. full pay 100.7221%. all 3 of them did win, but they each
underperformed EV by about 1/2, specifically 100.37, 100.37, and
100.39%. I would think that 10 million hands would come closer to the
expected EV than that ?? Another oddity, the total RFs for each was
201, 197, and 198, but the expected RFs are 230. Seems too far off ??
Maybe I just have a unlucky computer??Thoughts Please
Jim
standard deviation per hand is sqrt(variance/hands)
for AA: sqrt(26.8/9,999,999)= .0016
1 standard deviation: .0072 +/-.0016 is 68% likely
2 standard deviations: .0072 +/-.0032 is 95% likely
3 standard deviations: .0072 +/-.0048 is 99.7% likely
3.75 standard deviations: .0072 +/-.006 is 99.98% likely
100.37% (.0037) is a deviation of .0035 (2.2 sd = unlucky)
100.81% (.0081) is a deviation of .0009 (0.6 sd = lucky)
>
> I have just run 3 separate 9,999,999 hand autoplay simulations of
All
> Am. full pay 100.7221%. all 3 of them did win, but they each
> underperformed EV by about 1/2, specifically 100.37, 100.37, and
> 100.39%. I would think that 10 million hands would come closer to
the
> expected EV than that ?? Another oddity, the total RFs for each
was
> 201, 197, and 198, but the expected RFs are 230. Seems too far
off ??
--- In vpFREE@yahoogroups.com, "clcjim2000" <clcjim@a...> wrote:
--- In vpFREE@yahoogroups.com, "jimnkelli" <jbecker11@c...> wrote:
> Maybe I just have a unlucky computer??
>
> Thoughts Please
> JimMy computer is lazier than yours, but luckier. It ran only one
9,999,999 hand autohold simulation (I think it had other things it
needed to do), but it got 100.81% return, and 242 royals.I believe your computer may need a stern talking to.
What program did you use to do this simulation?
I've found that the basic random number generator on a PC has a few
flaws. While the number is essentailly 'random' (or psuedo-random) by
most accounts, certain patterns will develop when you run a lot of
hands. Strangely enough these patterns will repeat with repeated
simulations. Without bothering to test the specifics, I've assumed
that with large runs, you basically spin thru the random number
generator so that eventually you hit the same potholes.
For example, in simulating some poker hands (I think it was while
analyzing Four Card Poker), I found that I was consistently getting
too many Four of a Kinds vs. the theoretical number. When I dug
deeper, I found it was consistently the same ranks that kept coming up
more often.
In the end, I essentially wound up having to modify how I drew Random
Cards for games (especially when I was looking to run large
simulations) as I frequently run 100 million hands or more. Generally
speaking my process seems to have worked pretty well.
I would imagine that the VP machines have customized random number
generators and do not rely on the simple RND command available on a
PC. One company that I spoke to did not reveal the specifics of their
RNG, but did let me know that they have their own and that it has gone
through rigorous testing to make sure it was truly random.
Elliot
--- In vpFREE@yahoogroups.com, "jimnkelli" <jbecker11@c...> wrote:
I have just run 3 separate 9,999,999 hand autoplay simulations of All
Am. full pay 100.7221%. all 3 of them did win, but they each
underperformed EV by about 1/2, specifically 100.37, 100.37, and
100.39%. I would think that 10 million hands would come closer to the
expected EV than that ?? Another oddity, the total RFs for each was
201, 197, and 198, but the expected RFs are 230. Seems too far off ??
Maybe I just have a unlucky computer??Thoughts Please
Jim
--- In vpFREE@yahoogroups.com, "Elliot Frome" <compuflyers@p...>
wrote:
I've found that the basic random number generator on a PC has a few
flaws.
While the number is essentailly 'random' (or psuedo-random) by
most accounts, certain patterns will develop when you run a lot of
hands. Strangely enough these patterns will repeat with repeated
simulations.
I would imagine that the VP machines have customized random number
generators and do not rely on the simple RND command available on a
PC. One company that I spoke to did not reveal the specifics of
their
RNG, but did let me know that they have their own and that it has
gone
through rigorous testing to make sure it was truly random.
Elliot
>
I would imagine that the casino pseudo rng regularly re-feed their
rng's with new seed numbers to prevent this "pothole"
problem...perhaps this is what is happening when machines seem to
pause every now and then during extended play.
In one of the Travel channel programs the producers asked two slot
manufacturere about the randomness of their rng's, one slot
manufacturer claimed their machines produced completely random
results, not pseudo-random, but random...clearly this spokesman didn't
understand the topic completely...interestingly the other company, I
believe it was Ballys gaming, gave a reasonable reply in which they
spoke about pseud-random number generators and how patterns could
develop, but that they work hard to prevent this problem.
I used WinPoker on a PC. I think I will run one more simulation today
to see if it also performs in the same 'zone' as the other 3, then I
will reboot, maybe that will help create a new seed or something for
some different 'random' numbers, interesting.
Jim
--- In vpFREE@yahoogroups.com, "Elliot Frome" <compuflyers@p...> wrote:
What program did you use to do this simulation?
I've found that the basic random number generator on a PC has a few
flaws. While the number is essentailly 'random' (or psuedo-random) by
most accounts, certain patterns will develop when you run a lot of
hands. Strangely enough these patterns will repeat with repeated
simulations. Without bothering to test the specifics, I've assumed
that with large runs, you basically spin thru the random number
generator so that eventually you hit the same potholes.
jimnkelli wrote:
I used WinPoker on a PC. I think I will run one more simulation today
to see if it also performs in the same 'zone' as the other 3, then I
will reboot, maybe that will help create a new seed or something for
some different 'random' numbers, interesting.
The developer (Dean Zamzow) has noted that winpoker isn't really a
simulation tool ... the RNG algorithm is too simple. It's intended as
a practice tool.
However, I still wouldn't expect to see the deviations from ER that
you're seeing over 10 million hands -- provided you're talking about a
single-play run (not multihand play).
I can tell you that my PC, which is about 4 years old (866 MHz
processessor), would likely require more than 200 hours to complete
your simulation in single play.
I suspect that you may be running your simulation in 10-play, so that
you're only looking at 1 million base deals. This would involve a
much greater expected absolute deviation from ER and may make sense of
your results to date.
(Not sure if I recall the math correctly ... and I'm really writing on
the fly ... but I think it takes sqrt(n)*x hands of n-play to converge
on the ER with the same confidence interval that x hands of single
play does.)
Confirm the winpoker play mode used and I, or someone else, will
calculate the probability of hitting the number of royals (or fewer)
in one of the sessions you reported.
- Harry
"Elliot Frome" <compuflyers@p...> wrote:
> > What program did you use to do this simulation?
I used WinPoker on a PC.
I ran one 9,999,999 sim overnight on FVP (single line play)
and got 100.60% with 227 royals.
On 12 Apr 2005 at 6:49, jr decker wrote:
On 12 Apr 2005 at 13:23, jimnkelli wrote:
LOL. I saw that show. The letter from the first casino should have
said, "..and we KNOW our machines are completely random because the guy
who made our perpetual motion machine said so. So there."
Seriously, I also remember the show didn't really question the letter
at all. I enjoy the Travel Channel shows about Las Vegas, but they
need to be taken with a grain of salt.
carlos
--- In vpFREE@yahoogroups.com, "jaydavidson118" <jaydavidson118@y...>
wrote:
[etc]...In one of the Travel channel programs the producers asked two
slot manufacturere about the randomness of their rng's, one slot
manufacturer claimed their machines produced completely random
results, not pseudo-random, but random...clearly this spokesman didn't
understand the topic completely...interestingly the other company, I
believe it was Ballys gaming, gave a reasonable reply in which they
spoke about pseud-random number generators and how patterns could
develop, but that they work hard to prevent this problem.
--- In vpFREE@yahoogroups.com, "jaydavidson118" <jaydavidson118@y...>
wrote:
--- In vpFREE@yahoogroups.com, "Elliot Frome" <compuflyers@p...>
wrote:
>
>
> I've found that the basic random number generator on a PC has a
few
> flaws.
>While the number is essentailly 'random' (or psuedo-random) by
> most accounts, certain patterns will develop when you run a lot
of
> hands. Strangely enough these patterns will repeat with repeated
> simulations.
> I would imagine that the VP machines have customized random
number
> generators and do not rely on the simple RND command available on
a
> PC. One company that I spoke to did not reveal the specifics of
their
> RNG, but did let me know that they have their own and that it has
gone
> through rigorous testing to make sure it was truly random.
>
> Elliot
>
> >I would imagine that the casino pseudo rng regularly re-feed their
rng's with new seed numbers to prevent this "pothole"
problem...perhaps this is what is happening when machines seem to
pause every now and then during extended play.In one of the Travel channel programs the producers asked two slot
manufacturere about the randomness of their rng's, one slot
manufacturer claimed their machines produced completely random
results, not pseudo-random, but random...clearly this spokesman
didn't
understand the topic completely...interestingly the other company,
I
believe it was Ballys gaming, gave a reasonable reply in which they
spoke about pseud-random number generators and how patterns could
develop, but that they work hard to prevent this problem.
The way random numbers are used in VP machines a pattern is very
unlikely if not impossible. When a simulation draws a single random
number to determine the outcome of a single event, a pattern may
develop. I have seen this happen on wargame simulations. But when
the number is drawn from a stream of continualy produced random
numbers a pattern is almost impossible. The question becomes does
Winpoker or FVP use a continuous stream of random numbers or a single
draw for each hand. I suspect it is the latter; requires less
programming and computer resources. Does anyone know?
Rick
--- In vpFREE@yahoogroups.com, "Rick & Lucky Lucy" <luckylucy39@e...>
wrote:
The way random numbers are used in VP machines a pattern is very
unlikely if not impossible. When a simulation draws a single random
number to determine the outcome of a single event, a pattern may
develop. I have seen this happen on wargame simulations. But when
the number is drawn from a stream of continualy produced random
numbers a pattern is almost impossible. The question becomes does
Winpoker or FVP use a continuous stream of random numbers or a single
draw for each hand. I suspect it is the latter; requires less
programming and computer resources. Does anyone know?Rick
Based on comments from Dean a few years back ... In winpoker I believe
that the seed is determined at initialization using mouse position
among other factors. It's all determinate after that. By checking
computer resources while running winpoker it's also clear the RNG does
not run between hands.
Dick