'Let's make a quick stop in Vegas and blame it on an unruly passenger.'

Last night we were at a dinner party where I met an honest to goodness professional poker player. I have to admit that I have never met someone who makes money off poker. Living close to Biloxi (home of several casinos) and knowing a lot of smart people I do know folks who count cards in blackjack. These people tend to be regular folks who can keep a running dialogue in their heads that goes something like +1, +1, -1, -1, +1, hit at 18. Folks who like to go to the casinos for entertainment and can count cards tend not to because it take all of the fun out of it. Folks who try to make a living by counting cards get found out and put on a “No Fly” list for gamblers mostly because the house resents them.

This professional gambler didn’t wear fancy clothes or a lot of jewelry. Turns out she could count cards and win at blackjack, but prefers poker because poker is a game of skill and the house tales a cut then lets players play (to coin a phrase). She is even better at online poker which is where my interest was piqued.

Me: So, I would think that on-line poker would be dominated by computers masquerading as real people

Poker-lady: Not so. Computers are actually not very good at poker.

Me: Really. But they are really good at chess

Poker-lady: Chess is a game where all of the elements are known. When a computer loses at chess it can go back and analyze every move, learning from its mistake and not making the same one again. In poker, there is too much uncertainty. The computer doesn’t know the opponents hole card or betting strategy. Computers don’t do well with uncertainty.

Turns out there is one bot (Cepheas) that is now able to hold its own in a type of poker (head-up limit Texas hold-em). This computer will play a single person (head up) and win money over time if the bet size is prescribed (limit). The computer is the work of AI investigators at the University of Alberta who, aside from trying to build their school’s endowment, are interested in solving the “imperfect information” game. Chess is an example of a perfect information game where all players have all information. In poker, everyone has limited information regarding their situation as well as their opponents’ situation. You know what your hole cards are but not your opponents’.

Doctoring, it turns out, is a game of imperfect information. I know what I have prescribed (medications, exercise, less calories) but have to ferret out what my opponent (patient) is holding (But doctor, when I said I only drink water, I meant to say a liter of soda pop. Does that make my sugars higher?). For years, programmers have tried to write code to do what I do and have been unsuccessful. The folks in Alberta are now working with diabetologists to create a program to help with diabetes:

“It turns out that one of the things a doctor does so well is come up with robust [recommendations] … And that’s what our poker programs have to do, they have to be robust to ‘what are the cards my opponent has, and how does my opponent play?’ ”

So, one day I might be able to use the wily computer to help me with my patient who “just can’t get controlled.” For most of my patients now, the flop is hypertension, diabetes, and hyperlipidemia, Until I get a bot, I’m assuming they have non-compliance and a lack of physical activity as their hole card. Does that make me a cardsharp?

Advertisements