This is a truncated version of a post I wrote a few days ago that
somehow got obliterated by the board's server.
The anomaly regarding multiline flush draws: This is the collected
data from six players over a course of approximately 14 months in
the years 2001-2002. Almost all play was .25 3/5 line 10/7 DB, or
Super Aces.
With a sample size that reached 1.4 million draws (at which point we
felt we had reached a decent level of confidence), we had the
following results: flushes were filled at a rate of 1 in 6.1. This
is considerably worse than the mathematical odds of 1 in 5.2. We
calculated that this alone cost the player something like 1.7% of
his EV.
The really interesting (if that's the word) aspect of this is how
CONSTANT the ratio remained as our sample size grew. The first
collation of data stopped at 200,000 hands, at which time the
completion ratio was...1 in 6.1. It pretty much stayed there, never
dropping below 6.0 or rising above 6.2. This bothers me a LOT more
than if the results had been all over the map. The constancy of the
results suggests intervention in mid-draw, for instance if the
processor was "told" to underweight cards that would complete the
flush. This is contrary to what we "know", that the machine deals
ten cards when the initial bet is made. I see no reason why the
program couldn't stop after five cards, examine the resulting hand,
and then assign different values to the remaining 47 undealt cards.
We also found that draws to three of a kind were anomalous, not so
much in completing 4-of-kinds (although these were somewhat
below "par"), but in the number of full houses drawn. In this case
we felt our sample size, 200,000 hands, was too small to draw a
conclusion. (The trouble was, it's 51 to 1 against getting dealt
trips, so we couldn't collect much data.) For what it's worth,
though, we completed quads at a rate of 27 to 1 (nominal: 22.5/1)
and hit the full houses a horrifying 1 in 18 times (instead of the
nominal 1 in 15.3).
Kevin and I and two other individuals have now drawn to four-to-a-
royal 435 times since we started keeping track. Kevin missed 127 in
a row, hit one, then missed his next 150 before hitting another one.
The rest of us have missed EVERY SINGLE ONE. A priori odds: 46 to 1.
We "should" have hit NINE royals....not TWO.
The collective results of fifteen people who all play at a
professional level (although differing amounts of play) for 2003: of
fifteen players polled, twelve report a losing year, eight of whom
report a loss of over 1% of their total action (disregarding noncash
comps). In the five previous years, no more than one of these
players in any given year, booked a loser for that year.
Over the course of four months, six expert local players lost
$23,000 playing fourplay Odyssey Lucky Draw fullpay Deuces Wild
(100.7%, ha).
These are just some of the "strange" goings-on that we have
observed. To answer a question many have asked, BTW, why would the
casinos gaff fullpay machines, why wouldn't they just reduce the
paytables? Well, you can have your cake and eat it too: you can
continue to shout from the rooftops that you have "FULL PAY VIDEO
POKER", while actually returning 98% or less. The downside of this
policy, having your casino shut down due to nonrandom machines, HAS
NEVER HAPPENED and WILL NEVER HAPPEN, DESPITE the fact that casinos
have been caught with nonrandom slots on their floor over thirty
times in Gaming Control Board Enforcement history. None of those
casinos lost their license, and were subjected to no discipline or
to a slap on the wrist at most.
What conclusion should we draw? It's hard to say.