Devo182
"Well Known Member" TWSS
- Joined
- Jan 25, 2015
- Messages
- 40,766
- Likes
- 140,755
Not sure what you mean by manipulating the game. Poker AI can't manipulate the game or cards that come out. They've just found near optimal betting lines, bet sizing, and probabilitistic decision making that makes it near impossible for an opponent to beat it in the long-run. Maybe I misunderstood your statement.Well I'm an AI researcher. I'm not Andrew Ng or anything but I have read the Deep Mind papers and I've read a lot of the deep RL stuff which powers what you are talking about. Poker is child's play to having an AI be one of the best in the world at DOTA for example, and all of those make an intuitive level of sense to me.
So I'm not a super genius. But I can tell you if I could predict stock prices with AI, I would just do that and not bother with a day job. Me and every other AI researcher out there.
One of the main differences between games and stocks is that the AI in games has the ability to observe the game and manipulate the game. As it turns out, if you can manipulate the market, you really don't need AI to make a **** ton of money. See the Gamestop saga. The best humans are still significantly better than the best index funds.
As far as football, the same thing kind of applies. People will use AI to predict games and bet on them well before they use AI to call plays. And the best human gamblers are still head and shoulders above pure AI. Even all of the next Gen stats people talk about now are just fluff. I don't think any serious bookies or gamblers use them.
As for the stock market, algos have absolutely dominated for at least the last decade. Individual traders have shrunk to make up a small percentage of overall trades. That said, with an ever upward market, all can profit to a degree, but algos just do it easier and without emotions. Ofc fiber optics and proximity to market CPUs are a huge part of that too. Speed is king in trading these days and you need money to get that.
I have no data on gamblers, but I do know most of the well known guys I've read about rely on mountains of data to make their decisions. This was 5 or 10 years ago. The same period most poker AI cropped up. Wouldn't be surprised if AI has cracked sports gambling, but who knows.
Would greatly disagree modern analytics are fluff, when huge amounts of past data shows many of it is basic +EV decision making. Not really even debatable imo, the biggest margins are just based on basic math and decision theory. Mostly in the areas of going for 2 and 4th downs. But they're ultimately small edges, they aren't going to win Super Bowls by themselves, not even close. But small margins make a difference.
I see a future of playcalling being one that will puts current practices to shame. I hope we're ahead of the curve taking it more seriously. Jmo