Is my blackjack game working?

Is my blackjack game correct? Perhaps I don't understand the math, or I just missed something in my game.

I created a blackjack app, using these rules:

*8 decks
*Dealer stands on soft 17
*Double after split allowed
*Double on any two cards
*Split to 2 hands max
*Player cannot resplit aces (2 hands max)
*Player can hit split aces
*Player loses only original bet against dealer BJ
*No surrender
*BJ pays 3 to 2

My app uses a new shoe after every hand and so from the Blackjack House Edge Calculator I have an optimal house edge of 0.31183%. I used the Blackjack Basic Strategy Calculator with the following settings:

*Decks: 4 or more
*Soft 17: Dealer stands
*Double after split: Allowed
*Surrender: Not Allowed
*Dealer Peek: Dealer peeks for blackjack

Now I ran some simulations using perfect basic strategy and with a random bet of any of the following: 1, 2, 5, 10, 20, 50, 100

My assumption is after running x amount of simulations, the player will be down 0.31183% of the total amount bet, and I can safely say, my game is OK.

I ran about 85 million simulations and the player is UP 0.131049%, a difference of 0.442879%.

So I figure either:

a)my app has a bug and is paying out somewhere that it shouldn't
b)the amount won is statistically close enough to still assume my app is OK.
c)I didn't run enough simulations

I don't really understand standard deviation stuff unfortunately. Are results like mine likely in a fair game?

No need to vary the bet amount if you are testing the pay back. That just adds confusion and doesn't help you find out what you are looking for.

It adds more than just confusion; it adds variance, which means that you need to run your sim for longer before your results are meaningful. It's completely counter-productive.

I think randomizing the bets messes it up the ability to stress test actual advantage.

Stick with the same dollar value and run your simulation. You should be able to confirm a number near the "posted / ideal" statistics.

Changing bets is one of the strategy to beat the game--lose small bet, win big bets.

If you are shuffling every hand, I do not know why you need 8 decks. I suspect 3-4 would have very similar effects--again, run your simulation and stress test it.

-Kit

OK, thanks for the advice all. I will run the simulations again with a single bet and see if that helps.

Should the amount of simulations (85 million) be enough to get quite close (0.01% or so), or could I expect it to vary a bit still at this number?

Quote: ausername

OK, thanks for the advice all. I will run the simulations again with a single bet and see if that helps.

Should the amount of simulations (85 million) be enough to get quite close (0.01% or so), or could I expect it to vary a bit still at this number?

85 million is a little low for that. Each standard deviation should put your edge off by 0.01% or so (a little more). So there is only a 2/3 chance that you will be within 0.01% of the actual edge.

It should definitely show a gain (for the house) and not a loss, though. If you are off by more than 0.05% you have an error somewhere.