Probability counts on strategy engine?

[vectro]

Hello, looking at the strategy engine for Atlantic City rules (http://www.blackjackinfo.com/bjbse.php?numdecks=8+decks&soft17=s17&dbl=all&das=yes&surr=ls&peek=yes), I see a "house edge" of 0.39%. I have a few questions about this number:

1. Where does this figure come from?
2. What is it a percentage of? Basic bet or overall average bet? (I calculate that following this table with double/split equal to original bet results in overall average wager of 1.076 times the basic bet)

The main problem is that I can't reproduce it; in my own simulation I'm seeing a house edge of about $0.178 on a $5 basic bet, or a 3.328% house edge. You can see my simulation code (in perl) here: (Dead link: http://vectro.org/blackjack.pl)

Thanks for your advice. The output of one such simulation follows:

Outcome statistics (200000 games, final outcome $-35677.50):
Mean: -0.17838750 (expected -0.0209 or -0.0185) Average Bet/Hand: 5.36
Standard Deviation: Sample: 5.89394 Population: 5.89395
Worst game: -30.00 Best game: 35.00
Expected mean and stddev for various number of trials:
10 trials: Mean $ -1.78 Stddev: 18.6383
99%-confidence: $ -49.79 to $ 46.22
100 trials: Mean $ -17.84 Stddev: 58.9395
99%-confidence: $ -169.66 to $ 133.98
500 trials: Mean $ -89.19 Stddev: 131.7928
99%-confidence: $ -428.67 to $ 250.28
1000 trials: Mean $ -178.39 Stddev: 186.3832
99%-confidence: $ -658.47 to $ 301.70
5000 trials: Mean $ -891.94 Stddev: 416.7654
99%-confidence: $ -1965.44 to $ 181.57

 

[Sonny]

1 Where does this figure come from?

Each rule will change the house edge by a specific amount. You can use those numbers to find the house edge for any games based on the rules. Check out these links:

http://wizardofodds.com/blackjack
(scroll down to the “House Edge” and “Rule Variations” sections)

http://www.blackjackforumonline.com...he_House_Edge_for_Any_Blackjack_Rules_Set.htm

2 What is it a percentage of? Basic bet or overall average bet?

It is a percentage of the initial bet (IBA). With a house edge of 0.39% you would expect to lose $0.39 for every $100 bet you make (or 100 $1 bets, 20 $5 bets, etc.). The total bet advantage (TBA) would be somewhat different.

The main problem is that I can't reproduce it; in my own simulation I'm seeing a house edge of about $0.178 on a $5 basic bet, or a 3.328% house edge.

Yeah that’s way off, possibly because you only ran 200,000 hands. You’ll want around 2 billion rounds for very accurate results. You might be able to get away with 200 million if you just need general answers. I would expect a somewhat bigger average bet also. Try running more hands and see if it approaches a more reasonable EV (your SD of 5.89 sounds about right though). If not then there's some debugging to do.

-Sonny-

 

[dacium]

I doupt anyone here knows object perl very well. Most of us a C programmers (at least as far as I see). Perl is extremely hard to follow.

I would start by disabling splits and see what your EV compares with the strat engine here with no splits. This is where most people get stuck. You definatly have a major problem somewhere to be out that far. 200,000 hands his a extremely high chance of getting with in -1% to +0.2% for a game on -0.4%. Thats surely not your problem.

Probably the best way is run your program hand by hand and check what is going on. It won't take long for you to see if the player or dealer is drawing a card wrong or the bet is being settled wrong. Just play long enough to get splits and double downs and blackjacks and make sure they are all working.

Also we would need to see the other utils used to check the shuffle etc. Or please post a distriubtion of each card dealt and the dealer up card so we can check for bias or mistakes in the shuffle etc.

edit: I think i see the problem. The results of the hand are not done for split hands.