Artificial Intelligence Takes On 6-Max NLHE
We’ve got some big news at the intersection of poker and technology. As AI has become more sophisticated, we’ve seen it take down pro players in games like Chess and Go. Researchers even found a way to build a machine that could beat elite humans in heads-up poker.
For the longest time, however, AI struggled to play full table poker because having to simultaneously compete against so many opponents brought an incredible level of complexity along with it.
That all recently changed when an AI ‘superhuman’ named Pluribus crushed a 6-max NLHE table full of poker pros. This computer was even able to rake in an hourly win-rate of $1,000 per hour!
While this news is noteworthy and can certainly lead to some interesting reflections on technological advances, this is a poker blog, so we’re here to talk about poker.
What can we learn from a supercomputer poker player?
The first major takeaway is that, at its core, poker is a massive math problem. While psychology is certainly a factor, numbers are poker’s true essence.
This confirms a mantra by which every winning poker player lives by: studying is the best way to improve your game. Sure, we’ll never have the computing power of high-tech AI, but we can still train our brains in the calculations necessary for making +EV plays.
The second takeaway is that new poker strategies may be on the horizon. Since this AI’s win-rate is so high, it must be using an improved strategy that goes beyond anything that we’ve come up with so far.
Pluribus uses strategies that are known to work, like balance and refusing to limp, but it also employed some unorthodox strategies to great effect.
For example, it often made donk bets, which is a post-flop play where the caller bets into the pre-flop raiser. Finding spots to effectively donk bet is hard, even for seasoned professionals. Hopefully, by observing how the AI employs this strategy, we can learn more about it.
Pluribus also generally made larger bets than most humans would. This allowed it to force more folds, extract more value from made hands, and generally put its opponents in tough spots.
If there’s one thing that we can learn from this, it’s that we need to mean it when we decide to bet. We should either fully commit to the bet or just fold instead.