Bots won't destroy online poker

[Iceman]

1. No bots that currently exist can beat even mediocre opponents in full table play. Any decent 10-20 player can destroy even the best computer programs. There are far too many variables to "solve" multiplayer poker. And even if it could be solved, an optimal player wouldn't do nearly as well against bad players as a human player who adjusts to those players' weaknesses. While a computer could play a mechanical strategy and win money in the softest low-limit holdem and Omaha-8 games (where all you have to do is play good starting hands in a straightforward manner), that shouldn't concern serious players. If the games were that easy, a human player could beat them for loads of money. If there were many such bots, they'd all be donators to someone who could adjust to their play.

2. Computers play chess well because they can map all possibilities several moves in the future, while a human expert is limited to considering a small number of possible lines of play. Computers play backgammon well because they calculate exact probabilities several moves in the future, while human experts are limited to knowing strategies that approximate what the computer can do exactly. Computers play bridge well because they have perfect card memory and can use far more complex bidding and signaling systems than a human could ever manage. None of those skills would be of much use in poker. A stud expert that counts how many spades are dead knows his approximate chance of making a flush, and that's enough - usually it's a clear call or fold, and when it's close it makes very little difference anyway. The computer knowing that it's exactly a 17.3964% chance has virtually no edge over a human expert here. And how do you interpret your opponents' plays. It would be almost impossible to create a bot that wouldn't be either too easy to bluff or too willing to call postflop (and human players who picked up on that weakness would destroy the bot).

3. The two areas where bots would potentially be strong players are (1) simplistic games like lowball, razz, and 5-stud, but those aren't really played anymore, and (2) heads-up play. Heads-up play would be much easier to model. While a computer's ability to make meaningful conclusions about how an opponent plays would be very limited in a full game (because people mix up their play and exact situations don't repeat often enough for brute force conclusions that Devastator bluffs 29.645% of the time in this spot), with only one opponent to track and with precise holdings and flops being less important than playing patterns heads-up, the computer would be much more able to adjust to that specific opponent's weaknesses.