Chapter 9: Shuffle Machines

Blackjack Therapy by Clarke Cant

Chapter 9, how to beat those feared shuffle machines (and why they are probably illegal too).

Perhaps the best information on the internet so far about shufflers in the “Grimore” of Green Baize Vampire’s (that is his pen-name and he is a hell of a nice guy too) website (perhaps easiest to reach through the links page of He gives a detailed reading on the first generation of continuous shuffle machines that is simply great. Most of the details concern how to determine the latency, or the delay time, in rounds, until cards reappear. These early CS machines would place a round’s discards to the rear of the pack and shuffle them in, and quickly be ready to play another round. The clocking of the latency is important in that, even though the early machines would place the discards “to the rear,” the number of cards shuffled in would vary. For example: if an early CS used 6 decks altogether, this had to be tested too by clocking (it might actually be 5 or 7 decks), mainly by keying cards in a way but in the reverse of keying cards to spot aces. Instead 3 cards are memorized that are in the discard pile next to each other and the number of rounds for the same cards to appear together would be tested. With enough samples to have statistical significance, that would be used as the average delay. Once the average delay was established, the playing strategy was to count that number of rounds and play, using as your true count divisor the remaining cards left. The running count you played by is ONLY the number of rounds that corresponds to that delay. If you had a delay of 8 rounds, corresponding to just under 4 full decks at a full table (amazing that people did play them and play newer versions now), in a 6 deck game, you would ALWAYS divide the running count for ONLY the last 8 rounds, by 2.

The tedium here was the need for the clocking etc. People, GBV reports one team was lead by Stanford Wong, beat these machines in spectacular fashion. The detection teams at the casino would have their hands full having to keep an expiring count and would and will today having to reprogram tracking surveillance software, for such plays.

The newer King (shuffle master gaming ) and the Quick 5 ” (licensed to 3 makers I believe) machines are supposed to end the possibilities of such counting by having the possibility of the hands just discarded appearing on the same round. Well whoever sold the casinos on this idea didn’t do their math too well, because this machine is beatable still, and doesn’t require the clocking to determine “average latency,” that the other previous CS machines did.

Both are sold with literature that states they use a random number generator that moves the discards quickly (I am paraphrased from both so that the similar language doesn’t have to be duplicated) in a normal distribution into the cards being readied for play, but ends the dangers of card counting in that cards may reappear in 100 rounds or on the same round. The King uses 4 decks, while the Quick 5 uses 5. The King is taller and moves the stack up of cards up and down to a fixed shuffle mechanism. The Quick 5 uses a moving shuffle mechanism. For both there is one quick riffle of the discards into the main pack, so that your players will not have to wait or be led to walkaway during even the quickest delays while other machines have to be loaded and unloaded. That insertion is similar and has a normal distribution. Both sellers talk about improvements to the random number generators used being improved, to end the danger they can be shuffle tracked.

That normal distribution for insertions still causes a delay in the reappearance of the cards. The fact is, that you can select the latency you wish to play to. If you select rounds corresponding to 1 deck, you will adjust your true count always by counting the last one deck’s worth of cards, dividing the running count by 3 and multiplying the count again by ¾, or simply dividing by 4. If you select 2 decks worth or rounds, you would divide by 2, and then multiply by ½ or again divide by 4. Strange you say…damn right! But just because the equivalent true count divisor turns out to be invariant does NOT in any way mean that there is not any exploitable latency. Chapter 10 will take a whimsical trip through physics and philosophy to show how in-variance is nothing to be feared and not reason to jump to conclusions—except for bad mathematicians.

In this instance the shallower penetration is going to be better to play. The reason that you multiply by ¾ or by ½ is that the probability of having early reinsertion lower your count and profits goes up as you count more rounds.

Suppose you chose to count 3 rounds at a full table with 7 spots. That is close to .3 or 30% of the pack. You have a 70% chance that those cards will take at least 30% of the pack to reappear. You are facing 70% of the count; with 30% chance of the cards reappearing.

It is all reciprocal; as you choose to count more cards the significance of your count goes down.
Approximately the best profitability would be in counting 2 and ½ decks (about 2/3rds of the pack is the actual maximum) before you let the count “expire,” but for me the approximate 30% solution is best.

The BJF is: EC=effective depth or C; BJF=SA+3*(BA+PA+RA), using EC instead of C.

In both examples it would seem to be very shallow penetration, but from a careful reading of Theory of Blackjack (all editions would have this information) by Peter Griffin, all the hands played are at that depth. It is if you were allowed to play thousands of shoes by first seeing 75 cards + burned, played the round (and included the players cards now on the layout, which helps too –so never let the top counted round “expire,” in your running count until the end of each round). The edge is not spectacular but it is still an edge and few casinos will believe it is there; just look at the odd fact the true count divisor is always the total number of decks in the pack again! It is too strange, but the income player, with a black chip and higher bankroll, would do well to play these games, raise their bets ferociously (at least for now) and milk that casino for every possible comp. Even after it is proven this edge exists the surveillance staff is going to find it hard to prove that the player is counting (thankfully there have been enough outside — and even a few inside quiet stockholder — lawsuits where some degree of proof is needed to throw out possible counters –the Coast Casinos group excluded).

There are still some of the original non-continuous shuffle machines problems that need to be discussed. The biggest problem relates to bangers and the more tradition problems of warps, where normal peeking under tens and aces would bend cards (perhaps reread chapter 6). I had 2 opportunities to gain inside access to information on this. One was the casino manager at the Verdi Nevada Boomtown casino telling me about the original trials of the Shuffle Master machines. The other was a surveillance worker from the video room at the Atlantis, in Reno, telling me about early trials of tracking software there (management is now much more tolerant of counters and has come to realize that allowing some skilled players a narrow edge is necessary to their own profits as well, to keep others playing –we are supershills at best, to the casinos).

First I was told that it was at Boomtown that the first regulations were written up by the Nevada authorities (gaming is such a misleading term I won’t use it even though it is part of their proper title) banning having normal peeking under tens and aces, when shuffle machines were used. Even during the first tests, the uniform shuffle action, on cards with even the slightest bends, was enough to jam up the shufflers and cause massive clumping of cards. The most recent public comment on these closed regulations was reported in the Reno- Gazette Journal, when the Alamo casino, in Sparks Nevada, was approved to first add table games, on the condition that shuffle machines and peek readers be used on all tables to avoid (para.): the losses due to hole card play and malfunctions of shufflers that happened when regulations were violated at other properties….owned by the owners of the Alamo etc. That there is such a regulation is hard to verify openly in Nevada, but it is specific that procedures not be allowed that put bent damaged cards through any shuffle machines.

Yet from the chapter on bangers it is obvious that cards are still bent by the necessary guides used in the various practical peek reading systems in use. A careful reading of the open regulations on changing cards however admits that peek readers damage cards too. Just point your browsers to Gannet news service links or the state of Nevada sites.

This don’t-do-it-but-we-admit-that-this-happens-in-normal-use admission has been used by myself and others to argue that Nevada, and other gambling regulators, have not fully upheld their own regulations, including in recent class action lawsuits alleging deceptive video slot machine displays. Normally gambling (until the Federal Government comes in to, no doubt, eventually make things worse) is a state issue but this contradiction was used to reopen discovery in those cases. There is massive overlap between the peeking and shuffle device makers and the video slot and video poker machine makers. Normally there would be no case in that these “makers had been approved by competent state authorities: thus the US Federal Courts have no jurisdiction…”..That didn’t happen. As long as these dern things are used with peek readers they are probably illegal.

And even if the legal mess is untangled (read quietly settled with lots of “hush money”), there is no real danger that any shuffle machine not having at least normal distribution will ever be licensed. Already these makers have created a device that actually makes it likely for deck compositions to possibly remain in play longer. By not always taking discards deep (to the end of the pack) there is a delay in the time that the pack is “turned over,” and cards at the end are on average available for play. Not every discard packet is sent deep; when a riffle is done deep the cards at the back still have to wait to come forward, even though the wait becomes less and less. For the average player, in rare but noticeable circumstances, streaks would have some validity, due to this changing deck compositions, for the subset in play. What would take 8 rounds of full table play, to push forward rear cards into play would take, under the same circumstances (I am assuming a 4 deck shoe and half deck size inserts; these parameters vary in the new CS machines) would take 64 rounds, with the Quick 5 or King machines. Taking out the normal distribution of where the riffles are inserted. “To finally beat those damn card counters,” would be suicidal. Players playing all manner of otherwise, in normal games, bogus progressions and parley systems, would have, indeed already have, some real possible edges.

The clumping effects are present when banger bends are combined with shuffle machines. My Atlantis source quit when he was told to ignore an older couple he was checking out for possible card switching (a cheating move you can read more about in the books by Dustin D. Marks I already mentioned), and count down, both mentally and with the tracking software they were reportedly testing, a suspect counter. When I told him about what I was told at Boomtown, he begged his job back and gained access to files having records of about 10 million hands of checked play. He setup a script to filter out and examine basic strategy players and their results. I didn’t take proper notes but I would estimate that to be cut to about 1.5 million hands (I missed a golden opportunity). He was able to combine with files from several other beta test casinos, to get this number. He reported that basic strategy players did .2% better with the shufflers used with peek readers etc. This is comparable with the no-shuffle simulations Wong included in the latest edition of his, Profesional Blackjack, but still within normal ranges for fluctuations.

Anecdotally the boss at Boomtown stated that in trials they tended to beat counters (those trying to count, but who are losers in the long run) about as normally, while those who stuck to basic and flat bet slightly won. It is quite possible that both are just fluctuations and selective observation too, but it would make sense for the clinging of like bends to like bends to be amplified by this combination (shufflers and peek guides).

Im a sure however that the CS machines have more of a problem with this: that they damage cards more and have more clumping. Whenever the riffles are preformed deep, there is more pressure that can keep cards together. If you go back over the banger diagrams you can see where the typical bends are more unique too: aces have unique bends, tens have unique bends—more unique than the regular warps that were found to cause problems in the initial Shuffle Master trials. There is also more likelyhood that low cards are going to be placed deep than high cards. Even with the small sizes used before shuffle of the discards, a hand with more cards before busting is more likely to be madeup of small cards than large cards. The bust card is placed lowest on the discard pile, but the rest are more likely to be little cards. When sent deep the discard pack with lower cards is going to have a slight tendency to push the rest forward. These machines tend to do over time what “cut control lite” does.

All in all the casinos themselves have made a terrible mistake in listening to the idiotic consultants who helped develop these devices.

Chapter 10, Why I believe in Anarchy, Chaos and human progress.

The World Blackjack Portal!