Many major news outlets missed this as I’m sure they’re busy reporting on the WSOP at the moment but the University of Alberta AI program Polaris beat a group of live poker players for the first time.
The team of pros put together to match Polaris were Nick “stoxtrader” Grudzien, Matt “Hoss_TBF” Hawrilenko, and IJay “doughnutz” Palansky. Not an easy lineup to be sure but not exactly Chris Ferguson or other players who have dominated big HU matches like NBC’s HU Poker Championship.
Regardless, this is the first time that Polaris has bested human opponents of that caliber. The AI team behind Polaris made it tougher to beat by having the program change it’s play at certain points in the match. This helped plug a hole in one of the weaknesses previous opponents had exploited. They also had the program analyze how it could have played the hand differently to get a different outcome. This helped it learn from mistakes and constantly adjust to the conditions of the game.
While this will certainly stir up a debate on whether or not this was a the best of the best humans have to offer I think the more interesting point which is being overlooked is that this is developing into a very solid case that poker is a game of skill.
For all of those who consider poker a game of chance the results of tests like this and research by the Alberta AI group could be used to establish that poker outcomes have more to do with the skill of the player than it has to do with the quality of the cards.
I know noted economist Steven Levitt has been trying to acquire a large sample size of hand histories to prove that online poker is a game of skill but perhaps he could use some of the data being employed by the Alberta AI team to prove the same assertion.
By using an AI program and a group of willing participants one could run various simulations in order to demonstrate how specific changes in variables changes the outcome of the matches.
This could provide for a completely new approach to looking at the question and if the conclusions were demonstrated by noted economists and academics from AI research teams it would carry more weight than a self-conducted study by the gaming industry.
At that point how is poker any different than chess? Yes, I know the predominately a game of skill/chance argument but a key component of the skill vs. chance argument is the turn of the card factor. But if hands don’t make it to showdown because a player is pushed off a pot or you can demonstrate that a good player can play inferior hands against a poor opponent and still show superior results then you can minimize the effect of the turn of the card component of the argument.
To be sure, a lot of work needs to be done to prove the assertion but I think too much of the current debate has been focused on two players playing optimal strategy against each other. In that case two similarly skilled opponents will determine each match by the strength of the hands they are dealt. However if you demonstrate how different strategies being employed and certain weaknesses being exploited can markedly change the outcome regardless of the cards then you can reshape the argument in a way that makes it more difficult for the anti-gaming groups to make a reasoned argument against poker being a skill game.