September 5th, 2013
Gambling machines are a familiarity on the casino floor, whether it’s land-based or online. A new kind of machine might be ready to take on the lady lucks and the poker pros. The machine is called Texas Hold’em Heads Up Poker.
Texas Hold’em Heads Up Poker is designed around artificial intelligence neural networks. Even if you’re unfamiliar with it, this machine has been in development since the 1990s. It possesses something a little more than the other popular video poker machines you're used to.
Most machines are designed to play the hand they are dealt, but as many experts have stated, poker is a game of skill, intuition, bluffing, and traps. That’s where Gregg Giuffria and Fredrick Dahl came up with the “Little Bastard.” This is a game that plays the player, not the cards. Dahl’s work with artificial intelligence and neural networks at the Norwegian Defense Research Establishment is where it all started. Although he began his quest with Backgammon, the principles were just the same.
Dahl created complex computer algorithms that developed strategies from the repetition of mathematical equations. The problem was never trying to solve any individual scenario; it was about predicting all possible scenarios in order to make a good bet. He sold his software for $250 a piece beginning in 1997. This is what led him to start his work on poker. In Backgammon, you can see another player’s position. In poker, all you have are the endless possibilities that your opponents might be holding. While this made programming difficult, Dahl eventually got his computer to bluff, play a passive game, never initiating bets, and only matching opponent’s bets. In 2003, the brain of the game adjusted to its opposition.
However, the possibility for a game to play aggressively when its opponent continually folded would never land itself on a casino floor. Casino Commissions would never allow a game to adjust to a player’s style. It would have to play a World Series of Poker Champion the same way it played a neophyte. The neural networks would have to be stripped of their ability to learn. Even though he was a pessimistic man, Dahl designed the neural networks to play a perfectly defensive game. Furthermore, he developed two other networks that would rotate when necessary. The second and third neural networks would swap in and out to keep the game from becoming predictable and when their opponent was running low on chips. This game was tested against some of the biggest names in bridge and poker, Malcolm Davis, Bob Hamman, and Justin Lall.
Hamman was the man who brought this to the attention of Giuffria, who was a rocker turned gaming developer. He claimed it was all entertainment and was so fascinated by the machine’s bluffing capabilities that he spent more than $5 million on a prototype. He saw a market of players who wanted to test their skills in a real-life casino environment without looking like an amateur at a poker table in Vegas.
After testing among some of the most well-known poker players in the World Series of Poker, Giuffria handed it over to IGT and told them to "punch holes in it." The standard time for testing is about a month or so. IGT had the game for nine months. When asked about the product playing against humans developers said, "they'd get killed." While this diagnosis was pleasing, it didn't make the game all that survivable in a casino environment. Adjustments were made to make the game "vulnerable" at times. If a very good player put in 500 hours on the game, they might average $135 per hour, but a beginner could lose $500 in seconds.
There are about 200 machines in circulation today in Nevada (Las Vegas), Mississippi, and California. Two WSP players, Johnny Chan and Phil Hellmuth, have put a patent on their own designs complete with customized audio phrases and graphics. These models are said to be released later this month.