I usually train with poker academy and I am wondering how goods are the pokerbots there. What is the truth about them ? How good are they in playing poker ?
It's totaly about artificial intelligence issue. They read the poker table window as an image, and each part of the table with information specific to the hand is analyzed and recorded for computation. No matter how they were written well, no matter how they created in a good intteligence, no bots can be successful as such as a person. Because poker is not only a card game.
What is the truth about them ?
The truth about them, they play in in poker rooms more than you think. But today's technology prevents a lot of them. "I believe that poker sites should allow bots, but only on clearly defined tables. Players should know who the bots are, and have the choice if they would like to play them or not," says Barry Greenstein.
How good are they in playing poker ?
Giving an answer to this question is subjective but they are playing good. In the summer 2005, the online poker room Golden Palace hosted a promotional tournament in Las Vegas, at the old Binions, with a $100k giveaway prize. It was billed as the 2005 World Series of Poker Robots. The tournament was bots only with no entry fee. The bot developers were computer scientists from six nationalities who traveled at their own expense. The host platform was Poker Academy. The event also featured a demonstration headsup event with Phil Laak.
First of all Poker academy has shut down. Another company took over their website but they are not related to poker academy.
There are other better poker bots available tough. I use this Poker Bot, you can get a Free Trial there as well. This bot can be customized. You can make it play however you want. It uses Profiles (scripts) for the decision making. You can make a Profile yourself or purchase ones made by professional players.
How good are they in playing poker ?
As mentioned above the Bot that I use can be customized and you can get Profiles that play a professional top level game. The Holdem Bot looks for the following information every time it is your turn to act:
- Your Hole Cards
- Number of Players on the Table
- Your Stack size (number of chips on the table)
- Pot size
- Your Position
- The Board
- Exactly how many players folded, called or raised.
- The exact amount of the raises.
As you can see there is a great deal of information collected. All this information allows for very sophisticated Profiles to be created. One particular MTT profile has 5 different stages, meaning that it switches between different sections of code depending on the size or your stack. The chart below illustrates the different stages.
This is just one profile, there are a lot of profiles available each with their own strategies, some Profiles will bluff whenever it is theoretically safe to do so. The playing possibilities are endless.
Truth of the matter is, it is fairly easy to get a Bot that plays good poker but being profitable is much more important than playing good poker.
The major challenge for a Bot is to not be predictable. It is futile to play good if you are always playing the same. Your opponents will eventually get the upper hand. Keep in mind that players do take notes and there are also a few HUD type software that collect data on players. This is where I find most poker bots fall short, because most of the poker bots available cannot be customized and have their logic hard coded in the program.
I find it extremely important to change Profiles and playing strategy frequently to keep your opponents on their toes. You can change Profiles on the fly with the Holdem Bot and you can even set it to automatically change profiles when your stack reaches a certain size or when a specific number of players is left on the table or even at a specific time interval.
The truth? Depends on which form of poker you're talking about.
Heads up limit is near solved. People have successfully developed near equilibrium bots with such fine grained and sophisticated abstractions that they're unbeatable by humans. They've done this using a variety of means, the most widespread being variants of counter factual regret minimisation.
Other games with larger game spaces, such as heads up no limit or multiplayer limit are much further from being solved, although there are some bots that can beat (well, not lose in) all but the toughest games.
Put simply, the larger the game space, the further we are from beating the game. By game space I mean number of possible game states. The reason limit is so far ahead of no-limit is simply that there are a huge number of possible game states in no-limit as you can bet any amount of your stack at any time - in limit there's only a handful of possible betting sequences in comparison.
Note the bots I'm discussing here are based on getting as close to a nash equilibrium as possible - they're taught not to lose rather than to win. The thing is though, people make so many mistakes that they win "by accident" along the way as their opponents deviate from the equilibrium when playing.