[zxcvbn4038]

If your AI is any good it will refuse to play because it’s realized the house always wins.

I spent a few hours watching card counting videos on YouTube to see if there was anything to it - didn’t matter who it was, they spent the entire time belly-aching because other players were getting “their” cards. Amazing how all these card counting schemes depend on everyone else playing perfect basic strategy.

[Zarath]

Card counting schemes have almost nothing to do with the cards of others. I have no idea who you were watching but they sound like frauds. It's trivially easy to verify that counting systems work in theory. If you have a sufficiently good counting system, you will know exactly the cards remaining in the deck (Suppose I knew only 8s were left in the deck. I just never hit and let the dealer bust continually). Even the best human-capable counting systems only have an edge of 1-2%, meaning that you still need to get lucky or have a ton of time and a near infinite bankroll (minimize risk of ruin).

Having more players at the table hurts the efficiency of the count because more cards come out between when you place your bet, and when you receive both of your cards. This means that the correlation between the count when you place your bet, and the hand you receive, decreases.

House edge at good casinos is normally below 1%, and you're telling me an AI with perfect information couldn't beat a dealer whose moves are incapable of changing given new information?

Source: I've card counted for 6-7 years now

[zxcvbn4038]

If your playing with a six deck shoe and the casino is cutting at deck and a half, how much of an edge can you really get? Even if your the only player.

[Zarath]

Depends on your betting spread and table rules. But at a deck in a half you're probably looking at around 0.75% for a pretty standard ruleset [1]

The casino I usually play at cuts at under a deck and also has a side bet (Over Under 13) with even better odds than the blackjack game if you use a different count. It can also be gamed many other ways (they give $5 and $2.5 chips, but don't pay quarters, so if you bet $7.50 for your hand, they pay $11.50, essentially improving your odds even further).

[1] https://www.qfit.com/book/ModernBlackjackPage309.htm

[cortesoft]

Those people are dumb... other people at the table hitting too often actually helps a card counter, because you see more cards per hand.

[zxcvbn4038]

A good question might be why they are not all rich if card counting is such an easy win. I’m pretty sure the best way to get rich on Wall Street is to teach people how to get rich on Wall Street. Maybe the same thing holds true for blackjack. If anyone was winning big on a consistent basis I’m sure the casinos would change things up to stop that ie table limits, six deck shoes, continuous shuffle machines.

[Zanni]

Card counting is an "easy win" only in that it's relatively easy to master. With a bit of practice, you can gain a consistent (but small) edge. The problem is, the variance is enormous. You need a large bankroll to sustain the wins, and you need to put in a lot of hours to get into the long run. That's a major reason most successful counters play in teams.

Of course you're right in that once a casino notices you winning, you're a target for closer observation. A six-deck shoe isn't much of an obstacle, though. In Atlantic City, they'll flat-bet you (adjust the table limits so you can't vary your bet). In Las Vegas, they'll just outright ban you from playing.

[cortesoft]

It has been done: https://en.wikipedia.org/wiki/MIT_Blackjack_Team

But if you are that skilled, funded, and organized, there are probably better ways to make money.