Just noticed this article on an Italian poker website which provides a pretty good summary of my poker bot work.
Here’s a rough translation courtesy of Google Translate:
Recently the problem has come back into favor the use of bots, automated programs that are included in the Poker Room to play, we propose in this regard the story of a player, Matt Mazor, who between 2007 and 2008 he developed and used a of these programs to play on PokerStars.
Matt Mazur was discovered in 2008 by Pokerstars and your account blocked because it used a SNG bot, the player said to have used for about 2 years in the low stakes of the room and then Matt Mazur has described in his blog all the details of ‘experience passing through the actual development of software.
His bot has played SNG from $ 2 to $ 11, Matt does not provide financial details of its program, but speaks only examples of gaps and partial information, the last month of practice states that it had a profit of $ 480 Heads-Up on 500, also states that the results have been improving during the development and upgrades made to his Bot.
Mazur present its findings by analyzing what your program has done from a technical point of view / information and presents a series of graphs which illustrates the behavior of the program over time depending on the actual size of the stack, for example, we can see the chart on the first phase of the sit when the stacks are between 50 and 75 times the BB (You can view graphs of the behavior of the bot to follow link ).
Mazur has never forced the use of his bot, taking it to multi-table up to three tables at a time, for two main reasons: the first is being able to personally monitor and analyze performance data in real time and the second to intervene in the chat when the words were mentioned and computer bots or you can write one of the first alarm bells for the bots and what not engaged in chat discussions. (Matt had developed a mechanism by which the software indicated the presence of words in chat bots, or computer).
Bot developed contrary to what one can not imagine using the SAGE system based on the balance of Nash and commonly recognized as valid in the low stakes, but an EV based system and other logic conditions developed by Mazur.
In ‘October 2008 his PokerStars account has been locked, Matt received an email from the room warned him that his account was blocked for using a poker bot, but Matt took it philosophically and unlike any other, has also gave some advice to the technical room for improving the system of defense against Bot: Matt has shown what he believes points to work on to improve the research techniques of bots on the network (You can see the correspondence he had with the Pokerstars link below ).
Matt Mazur has continued to work on programs that could improve the performance in the game starting to deal with online tracking and odds calculators.
Matt Mazur 
Hi Matt,
my name is Marcos, i’m from Brazil. I developed a bot that plays Cash Limit Hold’em at PokerStars, unfortunately (maybe not), about 6 months after I started the development PokerStars kicked me out. The only thing that made me a little sad was that the code was huge and working very well, not the strategy, but all the acquisitions and controls of PS software. The sofware was cabaple of play 20 tables simultaneously. My strategy was based in some statics of each player using HoldemManager database and Monte Carlo for some statistics.
Now I started code a new bot for FT. But I don’t know if worth spend so much time developing a bot. So I’d like your opinion, does it worth?
Thanks
Hello,
I made a living running bots for years undetected (not an easy task) on PartyParty up until the US ban hit in 06 or 07. It then became a felony in the state of WA to do any kind of online gambling. So after 3 years very difficult and time consuming work I decided to give up on the project and get a real job to the delight of my wife. It played .50/1 and 1/2 limit games. Anybody can get a bot running and possibly even break even but writing a winning bot is a very difficult task and not many people are up to the task. The first step is finding a winning online player to correct all the mistakes. Luckily for me I knew a guy that made 7 figures a year playing online and he took an interest in the project. He was cool enough to go thru the 300-400k hands the bot played over about a years time and correct the mistakes. This was a very tedious programming task. Getting a bot to play like a human is not easy. One by one the bot stopped making mistakes and eventually it began to win money to the tune of about 2-4 BB an hour per table. This sounds like alot and it is but most of the good loose games are at night and during the day I was lucky to have the bot playing in 4-6 games. At night that could swell to 10-15. Table selection was very important and when the bot found these games it did quite well. Put a bot in a tight table and its going to be a serious grind. I think it would be lucky to win 1BB an hour if that in those types of games.
Towards the end after the hard core work was finished I began working on a player profiling system to increase its edge. It worked well but it was very cpu intensive and I needed to figure out a faster alternative in order for it to be effective in real time on many tables. I was just beginning on figuring that out using a separate computer to crunch the numbers when the ban hit.
Its worth the time to do a bot project but have money in the bank to live off of for awhile. Do not expect to win a dime (in fact you will lose for quite awhile) for a year or two.