Noah is an avid poker player and Troy recently caught the poker bug. Together they realized that there was substantial untapped potential in online poker analytics and wanted to do a fun yet challenging project to improve the decision making process for online poker players.
What it does
The tell detector uses hand history data (records of previous poker hands played) to identify predictable behavior that a given player exhibits. It especially focuses on players' bet sizing patterns. If an opponent's hand strength tends to be strongly correlated with the size of their bet, we can exploit their behavior when faced with the particular bet size in the future. The more hands the user plays with this opponent, the more precisely the tell detector will be able to pick up on these patterns.
How we built it
We modified some open source code to build a python function that scrapes hand history data from Carbon Poker (the poker site that we play on). We created a database in MySQL to effectively organize and sort the data. We then run the data through our tell detection algorithms (in python again) which identifies patterns in the data. Using Plotly's Python package, we generated graphics for players so that they can visualize the exploits of their opponents
Challenges we ran into
(1) Developing the skills to use Python, MySQl, & Matlab (2) Creating a website that Poker players can use to their own individual advantage. a. Heatmap b. Histograms (3) Since the project’s inception, Noah has managed to nearly double the money in the account that he puts forth to play poker
Accomplishments that we're proud of
(1) Implementing python and other diverse software packages as a coherent system (2) Developing a user interface takes more time and understanding than we originally could comprehend (3) Not being able to fully use our programs from the previous semester because they were in MATLAB, not Python
What we learned
(1) Three (or four) heads may be better than two. (2) Even without the greatest background knowledge of the software that we used throughout this year, integrating a systematic process/methodology to create the product was more important than just understanding how to create the product in the end. (3) Creating something that even the most basic user can understand could potentially help advance the average skill level of an online poker player. If online players develop the intuition to eventually see the patterns of play that our team sought to exploit, it could pay great dividends in the long run.
What's next for What The Tell?
(1) If we continue to pursue this project, we can adapt it to a live Heads Up Display (HUD) application rather than a website. (2) We can also seek to refine our product and monetize it as a subscription based service for online poker players.