|
PokerAI.org Research @2007-2009
|
|
Free Hand History Database |
|
Frequently Asked Questions
How many hands are in the database, and from which sites? We have over 1 billion (1,000,000,000) hands in our database, obtained from various poker sites. These are (for now) all hand histories from real money cash money games, from NL2 to NL100000. We will not reveal from PokerSites these hands were obtained, although one can (when necessary for the analysis) separate the hands by site. Why have you built own, proprietary database to store hands? Our ambitions are to extend our database in the future to reach 10 billion of hands, and more. That would require tens of terabytes of disk space if we use conventional means (RDBMS) to store these hands. Instead, our highly optimized proprietary format will be able to store 10 billion hands in under 200 gigabytes. Additionally, importing and analyzing such database is impossible to do by using conventional software (like for example Poker Tracker). Our optimized software is able to import over 50,000 hands per second, and query aggregated statistics for several million hands per second on conventional hardware. Is there better (or more complete) documentation for the Java software? This is being prepared. At the meantime, there are several self explanatory examples, from which you can learn almost anything that you need to start building queries and programs against the HHDB. Start with a simple tweaking of existing example, and we can assure you that you will be able to use the full power of the software in no time. Is the software available in other languages than Java? Alright, currently no. But this is open source/free activity. So if you want to build such reader of the proprietary DB for another language (and release it under GPL v3), then get in contact with us! Why not have a web page where users to make real-time queries against the DB? We haven’t looked into that yet, and it will take some time (several weeks to several months) until we change anything in that direction. The first thing when we start to consider this will be to look for volunteers to develop such “end-user” query language and code it for us and everyone :-) What kind of research is such database good for? If we knew all possible applications of such database, we wouldn’t be offering it to other people and researchers. We want people to be creative, and come with their own ideas of how to use this database. There are few ideas in the Examples pages, to demonstrate what one can do with this, and many more examples in the software package. How well are the hands obfuscated? We obfuscate player ids, table names, pokersite, game ids, and the time the hand was played. Having done all that, we should be honest and admit that one, having the database, can still obtain information about specific player, if he carries our the necessary work to correlate own (or otherwise obtained) hand histories with the one in the hand history database, and by this mange to “decode” the mapping for single player. Doing that for all players (and then using the database) is however impossible without having the original hand histories (in which case it won’t matter anyway). Therefore what we can say is that it is not possible to use the obfuscated database to augment real money play, but it’s impossible to prevent obtaining information about specific player, if one is determined to do so. Due to the theoretical possibility that the information for single player (s) can be decoded, we are very careful to whom we provide even the obfuscated database. For now we offer it only to published authors of poker and game theory books and papers. This, as you can imagine, are maybe just few dozen or hundreds of people in the world, so not a huge audience. Everyone else is able to benefit from this database in other ways, see the General Interest section. What for are *you* personally using the *unobfuscated* database? Are you cheating with it? If we wanted to cheat with this database, we could have done that without announcing it on a website. We use this database the same way as we offer it to other researchers, and we don’t use this database to profile poker opponents. That’s the truth. But you don’t need to believe us on this, and we don’t need to prove it, or even explain it, as it is simply irrelevant. We started this website, took the effort to publish and maintain the software, to obfuscate the database, to make it available to researchers, to verify their credibility, to answer questions and popularize it for the simple reason to support game theoretic and poker-related research, and give chance to the community of poker players have their toughest questions answered objectively (based on data). |