Streakiness myth?

Apologies if this has been covered previously.

You often hear the claim (incorrectly, I believe) that online casino play is 'streakier' than regular casino play. Whilst I accept that online casino play with reshuffling every hand is different to regular casino play with penetration into a shoe-dealt game (see http://wizardofodds.com/blackjack/appendix10.html), it is not much different, if at all, to CSM play, as explained per the link above. So, I did two 1 billion hand simulations (6 decks, DA2, DAS, S17); one simulating online play of reshuffle after every hand, the other simulating penetration of 75% and compared the streak percentages and they are identical. Unless I either did something wrong or am reading the streak data incrrectly, then there is no evidence that online play is streakier than a shoe dealt 75%. The flow of the cards may be different (less/more 10's and less/more 2's-5's) but there is only the most marginal of impacts on the house edge (and in favour of the player) and certainly no impact on the length or frequency of winning/losing streaks. There is a marginal difference between the actual number of hands at each streak (as there would be because no two 1 billion hand samples need be identical), but only by a few here or a few there. When expressed as a percentage, the figure is identical in just about every case with a difference of 0.01% in a few cases. Thus, the myth of 'streakiness' is just that; a myth. Am I correct? Anyone else come up with a similar conclusion?

Your conclusion should be correct, assuming a fairly dealt game. At one time there was an abundance of playable video blackjack games, where a fast player could play about 700+ hands per hour. I used to regularly see some amazing streaks on these machines (both good and bad), but I'm confident this was just because the high number of hands per hour compressed the unusual events into a much smaller time frame. You would see something once or twice an hour that would occur only once a day at an 80 - 100 hand per hour live blackjack game.

EE

The original question was about the myth of unusual winning or losing streaks when playing on-line blackjack vs. live table games. My response was that I thought this observation probably was due to the higher hand per hour rate that one could play on-line (and thus a greater opportunity within a given time frame to observe infrequent / unusual events compared to the relatively slow play at a live table), and I gave an anecdotal example from video blackjack machine play.

EE

I've seen a few 20+ hand streaks online.

I totally agree, EE Counter. Selective memory is the real reason. And lack of sufficient play in a land-casino to be able to make fair comparisons. People like to find a reason why they're losing, or worse, why their progression system is losing so they blame something other than the facts.

I don't feel that online casinos are unfair, however, I think there is a possibility that the games are streakier than you'd find with a hand or CSM shuffle.

Anecdote: I used to work at a videogame developer. We had a wargame with a 'dud warhead' percentage that was near 50%. In practice, it was ridiculously streaky. Well, our head nerd dug down into more research into how the random number was being generated. No computer can TRULY generate a random number, it has to be seeded from somewhere. We were seeding it from something crappy and repetitive. He "upgraded" it to something with more innate variability (number of microsecond on the computer's system clock, cosmic radiation radio frequencies, etc.). Streakiness reduced.

My point? I don't really have one. But I have two comments about the simulations:

1) If you simulate a billion hands of anything, that's going to mask any streaks that may have developed. What could conceivably be interesting is to simulate 2 million batches of 500 hands, and see how the distributions of wins and losses hashes out in the comparison.

2) The simulations are done on a *computer*, and therefore would inherent any freaky tendencies of that software's RNG algorithm. The test would be to test an RNG algorithm used by a casino, versus a super-fancy-cool RNG simulation.

With all that being said, my bias against random number generators is really just personal hunch. It is, honest, more likely that they are equally streaky as real life, and we just report it because of the speed of play.