Industry.XML (Sub-heading: can you please upload your poker hands?)
One book that changed my life was Moneyball by Michael Lewis. The basic tenet of the book is that baseball players could be measured objectively and that neglected measures of performance (on base percentage, back in the day) could be used to pick up great players for cheap.
Charlie O’Donnell is focusing his startup, Path 101, around a similar industry.xml concept: structure resumes in a XML format and then help college kids decide which jobs they might be interested in exploring by showing what paths all the people before them have taken and the different paths they end up in 10-15 years in time.
A random idea I had a while back, while delving into poker data was the concept of poker.xml. A huge number of people convert their poker hands into a text output and discuss them on forums like Two Plus Two. Here is an example. Steven Levitt of Freakonomics fame is undergoing a similar project, called Pokernomics.
I thought what if there was a way to support people who already do this but have a backbone structure of XML. And so with that, here is a new idea I am working on at the moment: After the River.
The site is driven off a XML format that represents a poker hand. So this hand here, for example, is driven off this XML file.
The idea is that if it is successful, designers can design their own players and people can use the data in computing what strategies work and don’t work etc.
As Charlie mentions, there has to be something in it for the user to upload. Hopefully with the site, that is clear: it will detect patterns of play that are hurting you and patterns of play that are helping you. And you can share your hands on your blog, MySpace page, forum posting etc.
Being the mad skillz domainer that I am, I also registered pokerxml.org and hope to have that site up and documented soon.
So a few questions:
1. Does anybody who reads this play or has played online poker? If so, would you be able to register an account at After the River and upload your hand history files and let me know how interesting the site is (first name dot last name at gmail)?
2. Is there anybody that would find this body of data interesting and would be interested in developing applications on top of an API? (All free and open, just help us define what is useful and what is not)
3. Any poker addicts interested in working on the site? So far it is a crazy idea from me and one python developer from the Ukraine. I am also planning to document the creation of the app and time/costs/revenues etc. on this blog as well.
A very interesting idea! My initial thought is that as you build a larger and larger corpus won’t you find the winner tends towards the underlying statistical probability of a given hand winning (i.e. in the long run the best hand wins)?
Well, I am thinking that people tend to overplay certain hands. So AK for instance, they probably overbet and lose a dispropotionate amount because it is a ‘premium hand’. Whereas a small suited connector like 4-5, might pay off to a larger degree or a small pocket pair that flops trips and cracks a larger pocket pair.