Is the Threat of Poker Bots Real, Imagined, or Somewhere In-Between?
On their debut podcast episode, my colleague Alex Weldon and WSOP bracelet winner Andrew Barber talked about Google’s AlphaGo Go program prevailing over one of the best GO players in the world, and what it might mean for poker – Barber also believes this (Artificial Intelligence) is one of the most important issues of our time.
As the two noted, the University of Alberta’s Limit Holdem bot Cepheus is said to have solved heads-up limit Holdem. Claudius, a heads-up No Limit Holdem bot, may not be “unbeatable” (it got hammered by a group of top heads-up NLHE players) but it would certainly ravage any low or perhaps even mid-stakes games it sat in. And then there is Baby Tartanian 8, the most recent winner of the Annual Computer Poker Competition, which has studied trillions of poker hands.
Still, even the majority of bot designers and programmers admit it’s unlikely most of these poker variants anytime in the near future. Furthermore, in their current form, bots aren’t playing optimal poker (taking advantage of their opponent’s mistakes) they’re simply playing, or trying to play, unexploitable poker. To a bot, a win is a win, so they may leave value on the table a human player wouldn’t.
This article from FiveThirtyEight.com has a good breakdown of the different complexities of strategy games.
These admissions of the current shortcomings, as well as high-level bot designers emphasis on heads-up games (the simplest form of poker), should be reassuring, and it is on a micro level, but it’s sort of meaningless on a macro level.
The real problem with bots
Bots are generally seen as an online poker problem (and it is a major issue), but the progress made on the poker bot front and the proliferation of bots also has ramifications for live poker players.
As a learning tool, bots are narrowing down where poker players need to be focusing their attention on, as their trillions of hands create little doubt on previously debated poker theories. Instead of having to experiment with different pre-flop raise sizes and bluff frequencies, these aspects of the game just “are.” Bots (and humans) are increasing the level of basic knowledge, and the Internet is disseminating that information to more and more people.
Poker may not be headed towards the solved game column, but the difference between good and bad players is shrinking.
If we look at poker is five stages:
- Clueless player – Players who have never played poker
- Kitchen table player – Players who understand game-play and hand rankings
- General knowledge player – Players with basic understanding of strategy
- Seasoned player – Players with a solid understanding of poker’s strategies
- High-level player – a thinking, adaptive, winning poker player
Comparing what is thought to be general poker knowledge in 2002 and 2016 is night and day. Not only has what is considered basic knowledge grown in depth, but it’s also proliferated. In 2002 a player who knew a starting hand chart, position, and how count outs and pot odds was considered decent. In 2016 it’s taken for granted that every single bad player knows this, and most know way more than this.
The better the starting point for players (or better stated, the quicker they can get to the General Knowledge level) the harder they are to exploit. They’re still beatable, but the clueless and kitchen table poker players are all but extinct at this point. Everything a person needs to reach the General Knowledge level can be found with a Google search and a couple hours of reading.
Bots are only exacerbating this problem.
The fear of playing a high-level bot is overstated…
… but this doesn’t mean you’re unlikely to feel the impact of poker bots.
Even if programmers are unable to develop a No Limit Holdem bot (or perhaps a full ring Limit Holdem bot) that is unbeatable, these programs could still wreak havoc on the online poker world.
Unless it’s competing against the best players in the world, a bot doesn’t have to be unbeatable to beat the game.
Furthermore, as I alluded to above, a bot can be beatable and still upset the poker ecosystem. All a bot has to do in order to be detrimental to the game is fare better than the typical player that would be sitting in the seat its currently occupying.
And even if it’s a wash, the bot could still be hurting the poker ecosystem by the mere existence and people’s tendency to jump to conclusions.
The perception of bots
The very idea (real or imagined) that you could be facing some unbeatable program is going to sour a lot of would-be online poker players from taking up the game, and it offers a convenient excuse for why people lost at the online poker tables.
Worst of all, the idea of rampant bot usage (overstated), with no way for the sites to curtail it (not entirely true) is a convenient vehicle for opponents of online poker to attack the game.
Finally, on an intangible level, a bot doesn’t speak or interact with other players and plays a boring style of poker. So, even if they were beatable, they are the antithesis of the atmosphere online poker sites are trying to create as they try to reengage with disillusioned poker players – the clueless and kitchen table players I discussed above.