![]() | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
GameMasters Secret's Page |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Dealer's Ending Hand | % Probability |
| 17 | 14.6% |
| 18 | 13.8% |
| 19 | 13.5% |
| 20 | 17.6% |
| 21 | 7.4% |
| BJ | 4.8% |
| Bust | 28.4% |
Another important element in our analysis is the number of times each card in the deck appears. A very easy way to cheat is to remove from the deck(s) certain cards that benefit the player, mainly the 10s and Aces. We track each card that appears as the hands are played and, as you might expect, we verify that they are all showing in a proper proportion. Ignoring suits for the moment, there are 13 different cards in an honest deck and, within certain parameters, they should all show an equal number of times. Whether or not the casino suffles after every hand has no bearing in this, because the proportions should remain the same. Here is the raw data from a game we recently checked by playing about 7500 hands:
| Card Value | Number Seen |
| Ace | 3345 |
| 2 | 3222 |
| 3 | 3257 |
| 4 | 3215 |
| 5 | 3340 |
| 6 | 3243 |
| 7 | 3270 |
| 8 | 3312 |
| 9 | 3317 |
| 10 | 13330 |
First of all, you will notice that we track all 10s (10, Jack, Queen and King) as one card. It's only if we see an anomaly that we separate each one out in a follow-up trial. The total of all these cards is 42,851 and, since each card should show up one-thirteenth of the time, the "expectation" is 3296 (13,184 for the 10s). Right away, the Aces and the 5s pop out as being above what's expected, but is that a sign of cheating? Probably not. What we have here is normal variance that will likely diminish (in terms of percentage) as the size of the sample grows. Using these numbers, we can expect the appearance of each card to vary, plus or minus, by 57. The Aces are 49 (1.486%) over the expectation and that is within acceptable parameters. The number of 2s that were played is below expectation by 76 and the 4s are below by 81, which is troubling, but not a clear indication that something is wrong. Perhaps someone who is more skilled than me in statistical analysis can figure out if this casino is cheating, because I just don't see it.
One very easy hand to track is the number of "blackjacks" the player and dealer get. It seems reasonable to assume that a cheater could try to manipulate this, so it's worth recording. These are counted only if they occur on the first two cards and, since the probabilities are well known, it's a good test. Here are the probabilities of a 'natural' in various games:
| 1 Deck: | 4.8265% |
| 2 Decks: | 4.7797% |
| 4 Decks: | 4.7566% |
| 6 Decks: | 4.7489% |
| 8 Decks: | 4.7451% |
So, if you play, as an example, 1000 hands and the dealer received 55 naturals and you got only 45, is the casino cheating? Here's how to find out: First, figure the "expectation" (0.048 x 1000) and that's 48 for this example (remember, you can't get a fraction of a blackjack). Compute the square root of 48 and you'll get 6.92, but let's call it 7. This is the standard deviation. The dealer's 55 naturals is 7 over "expectation" and that's equal to one standard deviation which is quite normal. Your 45 is just half a standard deviation below expectation and that, too, is quite normal. What's not normal? About 3 SD is pushing things and 4 SD is a big warning flag. So, for our 1000-hand example, if the dealer had received 75 naturals, I would at least suspect cheating. If a similar number showed up in a follow-up sample, that would confirm my suspicions. Is it absolute proof of cheating? No, but it would tell me to stay away and, I promise, I would tell you.
This is probably the only means of measurement that most players use, but it's also the least reliable. While we can precisely calculate just how many hands we should win, lose or push, you need to remember that how well you play the game can effect these numbers, especially in relatively small samples. If someone is always hitting 17, then they are simply going to lose more hands than someone who always stands on 17. But, in a small sample, either of them could win more hands. See the problems this creates? That's why we don't put a lot of stock in it, yet ironically it's the whole reason why we do all this measuring: wins. losses and pushes are all that really matter in the game of Blackjack. It's just that they aren't good measuring devices. Nonetheless, we track the numbers and what we expect to see is 43.2% win, 47.9% lose and 8.9% push.
I think that about 95% of the people who play Blackjack at Internet casinos think they're being cheated when, most likely, only a small fraction are. Blackjack is a weird game and if you play it long enough, you'll see some amazing things. But don't operate off emotion. Record the data, such as the dealer's ending hands and the number of naturals you both get and you'll start to see that most games are honest. If the dealer "always" gets a BJ when you get one, start recording all the BJs in the game and you'll see that it isn't really "always". Taking the emotion out of playing will make the game more fun and,perhaps, more profitable.
Next time we'll cover Video Poker.
See you then.
| FRONT PAGE | | | STORY ARCHIVE | | | VIDEO BOOK REVIEWS | | | GM'S STORE | | | RESOURCES | | | ADVERTISING | | | CONTACT |