| Back to Reports page | |
| Shuffle Tracking and the Game of Baccarat | |
The answer is for most people - NO. Linear counting systems for the game of blackjack are very good at approximating the actual effect of card removal and expressing this effect in terms of advantage. They do not fair so well for the game of baccarat however, particularly for the tie bet. Work by the late Peter Griffin and a paper by Michael Hall suggest that linear counting techniques, while technically able to show an advantage, are exploitable only by those with huge bankrolls and then only in a manner easily detectable (heavy end play). If an advantage is to be gained in baccarat by card counting then it has to be by perfect card counting, which involves the use of a computer. Perfect card counting does not actually involve counting cards. Removed (played) cards are entered into a computer. Sufficient cards are then selected at random from those remaining in the shoe to complete a coup. The result is logged, the process is repeated many millions of times and the results tallied. From these results the advantage (disadvantage) is calculated. Computerized simulation is the only truly accurate way to determine the effect of card removal in the game of baccarat. As Michael Hall says "even perfect card counting yields pathetic results" but this isn't the complete story. Griffin and Hall both indicate that the two parameters required to beat the game are small subsets and a very large bankroll. The latter goes without saying when discussing professional syndicates. You can see from their work that as the subsets get larger, even by just a few cards, the advantage drops off dramatically. Nevertheless if you use a computer to simulate card removal you will have a very slight edge. Even with a slight overall disadvantage a computer-equipped and well-financed team will make good money by exploiting the player incentive schemes offered. Assume the computer-equipped punter is playing a break even game. Whatever the casino returns to him by way of rebate on turnover is his profit. If you do your own numbers, even based on non-negotiable chips you will see that the game is worth several thousand dollars an hour to a big player. Bear in mind that obtaining a break-even game using a computer is relatively easy with any baccarat game dealt down to half a deck. Some casinos use two colour cards in baccarat to prevent seconds dealing. In effect this is the equivalent of playing with two four-deck stacks. Much smaller subsets will be simulated and this significantly improves the advantage. (The colour of the card back is entered into the computer along with the card value). It goes without saying that with a little inside help the punter would end up owning the casino in a week or two. Shuffle tracking Baccarat games (and their derivatives) that shuffle and re-use cards are attacked in a different way. Before I go into the details, a little background information will not go astray. The best concealable computers use a self-learning shuffle tracking and mapping algorithm. Discards are entered into the computer as the game unfolds. As the cards are about to be shuffled the operator keys in a dealer ID number. The computer has already learnt this dealers average grab size, unevenness (i.e. heavy left-light right), riffle roughness etc. by having several shuffles previously entered. The computer then shuffles the cards exactly the same as the dealer. The theory is that the computer knew the make up of the stack pre-shuffle so can reconstruct the make-up, post shuffle. It's not quite that simple - it works like this: - As the second shoe of the day is played, the discards are again entered. Let's say the last card dealt was a 6C. The computer runs through its list of cards and flags every 6C. The next card dealt is a 9D. The computer then looks at all the flagged 6Cs to see if a 9D follow any of them. Let's say that a 9D follows three of the 6Cs. All the other 6Cs are dropped. Next card dealt is a QH. If the computer finds that one of its 6C-9D pairs is followed by a QH then it knows exactly where it is in the array. We still have a long way to go. It's all well and good for the computer to know where it is in the array but it still needs to be sure that it can predict enough cards to complete a coup. This is where its learning ability comes into play. The algorithm knows that this sequence came from a left-hand grab of thirty-two cards and that this sequence is twenty cards into the grab. It therefore knows the value of the next twelve cards. Still not there yet. Any sequences of cards that have been shuffled are made up of cards from the left grab and the right grab. So far we only have information about the next twelve cards from one side. With a simple shuffle, one-sided information is sufficient to produce meaningful simulation results. The computer takes the next six cards from the known grab and six random values and performs millions of simulations, the outcome of which is expressed as a percentage advantage and conveyed to the operator. You may say, why simulate with two six-card subsets when you only need a maximum of six cards to complete a baccarat coup? The answer is that if the riffle ratio is rough the cards could be clumped and come from one side only. More complex shuffles require information from left and right grabs to produce meaningful simulation results. It should be borne in mind that these twin six-card subset simulations do not place the cards in random sequential order, a sequence could be separated by more cards than you expect but they will still follow in sequence. In effect every coup is the same as the last coup in a big baccarat game dealt down to twelve cards that are not in random order.There's more. If our dealer is like most dealers and picks heavy grabs on one side, situations occur where our computer locks into the end of one of these grabs and knows the exact value and order of sufficient hands to complete a coup. This is of course a 100% advantage. Friendly dealers A very bad and clumpy dealer is a godsend to a computer-equipped team but so is a very good one. If a dealer picks identical sized grabs and performs perfect riffles, in other words performs the shuffle exactly as laid down by the casino operator, the computer is able to track it much more accurately. What do you think the chances are of getting a collusion charge to stick against this dealer? Over friendly dealers The above methods are very effective and conducive to longevity but if our dealer is a little braver and our team a little greedier, the next step is to make the clumps a bit bigger, one way or another. Nothing new here, you've seen it before, but it's a little harder to detect if a computer is used to track and predict. Conclusion The casino industry is able to call on much technology and some very clever people to provide them with risk management tools and sophisticated player tracking software. They would do well to remember that some very talented people have chosen to work against them. |