Strategy Subclasses

RandomStrategy

This strategy randomly picks a Choice, compares it with the list of available choices, and returns if they match. Pretty simple, but also pretty stupid.

DealerStrategy / AltDealerStrategy

These are straightforward, implemented as conventionally found in any rules of blackjack. If the dealer has sixteen or less, she must draw additional cards until she reaches seventeen or more. The dealer must stand on hands totaling seventeen or higher, even if this value is lower than that of the player's hand. In AltDealerStrategy, the dealer hits on a soft 17.

SmartStrategy

This object implements the basic strategy as delineated in Lawrence Revere's Playing Blackjack as a Business. It does not count the cards, and thus does not vary its proportional bets or its plays depending on previous events. The strategy represents an optimal manner of playing any particular hand given the dealer's up card, knowing nothing about the present state of the shoe itself. When learning the game of blackjack, most players begin with the basic strategy. With standard rules, the player should just about break even utilizing this strategy. Note that dozens of runs of the game are necessary before any pattern of winning can be seen, since the standard deviation of winnings for any one shoe of blackjack is enormous, regardless of the strategy implemented.
We will now discuss some of the particular plays used by the basic strategy (the same principles apply below to the count strategy) so that people who are not closely familiar with blackjack can understand some of the intuition behind the statistical results.

Some players are surprised that you should stand on a 13. You have to realize that, to you, a final hand of 13 and 16 are exactly equivalent. The dealer has to hit until he gets a 17 or greater. So either he busts, in which case you win either way, or he doesn't bust, in which case you lose either way. Generally, you stand on most stiffs unless the dealer has a 7 or greater showing (a 7 or greater showing makes it likely that the dealer has a hand of 17 or greater, so he will likely beat you if you just stand).

You want to double down when the odds are very good that you will win the present hand, since it doubles your bet. The price for this is that you can only receive one more card. As such, the best time to double down is when you have an eleven, since there is a 4/13 chance that your one next card will be a 10 and give you 21 (an 8 or 9 would be pretty good here too). You would never double down on a hard hand greater than 11, since you may well bust. And you would never (unless counting cards) double on a hand of 8 or less, since your chances of getting a good hand are very slim. With soft hands it gets a little more complicated, since you may be able to get a decent hand with either a small or large card, but not with one in the middle. Looking at the plays and thinking about them should give some of the intuition as to why they are what they are.

Splits are a somewhat complicated concept, but the two hands that you always split give a pretty good idea of what they're used for in a basic sense. If you have 2 aces, you always split them because this gives you a very good chance of ending up with two excellent hands, each of which carry the full amount of the original bet. You split here to take advantage of a good situation. If you have 2 eights, you always split them, but for the opposite reason. Sixteen is an atrocious hand, the worst there is. You will always lose to a dealer that does not bust if you stand, but if you hit you have a very good chance of busting yourself and losing immediately. Splitting the two eights is a form of damage control. Yes, it does double the total amount of money on the table, but hopefully at least one of the two hands you get from splitting will be decent and you can eliminate your losses. It turns out that no matter what the dealer's up card, it is statistically advantageous to split the eights and hope that at least one of your new hands wins than to play the hard 16 and overwhelmingly likely lose.

CountStrategy

This object implements the Revere Point Count system as delineated in Lawrence Revere's Playing Blackjack as a Business. The strategy counts cards using a four-layered system. Every card has value -2, 0, 1, or 2 depending on how much better off it makes the player that the particular card has been removed from the deck. Cards of value 10/ACE are counted -2 (since the player would ideally like a shoe heavy in aces/tens, he reduces his count when such cards are removed), 2's and 7's are counted +1, 3-6 are counted +2, and 8's, 9's are counted 0. The player then determines the true count from this raw count, since the count system is built around a shoe with 26 cards remaining and a given raw count. Obviously, a raw count of +5 means far less when there are 8 decks remaining than when there is 1/2 deck (26 cards) remaining, so the player makes his decisions based on the corrected-for true count. Bets range from 1-12 units basedupon the present true count. Plays are generally very similar to those in the basic (smart) strategy, but here they are based on the present true count in addition to the player's hand and the dealer's up card. For some plays, if the true count is different enough from zero in a certain direction, this strategy will choose a different play than the basic strategy would in the same situation. This reflects the fact that the different play gives the player a better overall expected winnings for that hand than the basic play, given the present state of the shoe.

As previously mentioned, the mechanics of the count strategy are very similar to those of the smart strategy, so you can refer to the above description for some more detailed explanation of the intuition behind certain plays. We will go into one specific count play here. It involves playing an 8 against a dealer's 4, and the mechanics detailed below are simulated in the testcount program. On anything but an exceptionally high true count, it makes no sense for the player to double down in this situation. He does not have a great chance of getting a good hand, he has no chance of getting a great hand. Only with a 9, 10 or ACE will he even be able to beat a single dealer hand. If the true count is exceptionally high (+6 in this case), the player will want to double, since he can be pretty sure that his hit card will land him at 18 or 19. Also, it's very possible that the dealer has a 10 under the 4 that is showing, so he will have to hit, likely giving him another 10 and causing him to bust. In this case, therefore, the player should double.

With standard rules, the player should beat the house slightly utilizing this strategy. Note again that several dozen games are necessary before this pattern will emerge, and the standard deviation of winnings with this strategy, as with any other in the game of blackjack, is so great as to discourage all but the most disciplined and patient players. Strategies such as the Revere Point Count, which can be proven to beat the house in very large numbers of trials, are the genesis for the popular thought that blackjack is a beatable game. This is, without question, a true statement. Looking at the subclasses of play that are implemented here, as well as the addCard function in the countstrategy class, will give you an idea of just how difficult it is to keep an accurate true count and play the hands correctly. Even a computer will win only a small amount of money (compared to the total money put into play), and it never forgets the correct play, nor does it get distracted by cocktail waitresses and lose track of the count. There are, however, a dedicated few players who have come through the casinos over the years with correctly computed strategies (much of the computational work was done by IBM Corporation's Julian Braun) that have been honed to perfection by thousands of hours of practice. The expected winnings are positive, and playing enough shoes perfectly will eventually lead to a realization of that expectation. For those who do it right, an enormous profit awaits.

Back to BlackjackDesign