Poker Card Data Analysis
An important part of a poker player’s game is to critically examine their gameplay and look for areas of weakness. While reviewing hands is a time-consuming process, it is crucial to the improvement of a poker player’s game. One of the most useful tools for analyzing a hand is a poker software tool called a “hand history”. A hand history is a detailed record of a previous online poker hand, including all betting actions that took place, such as calls, raises, and folds.
There are a few different ways to record hand histories. The most common method is to use a poker software program that automatically records each hand, such as Poker Tracker or Holdem Manager. However, it is also possible to manually record hand histories with a spreadsheet or by using a pen and paper. Using hand-tracking software has several perks, such as the ability to quickly analyze and filter the results.
A hand history dataset is a set of poker hands, whose observations are used for training and testing a machine learning model. The dataset consists of 25K samples for training and 1M samples for testing. Each observation describes a five-card poker hand and has two predictive attributes (suite and rank) for a total of 10 features. The dataset is said to suffer from the Class Imbalance Problem, which makes it difficult for a classifier to achieve high accuracy. Nevertheless, a simple neural-network achieves over 90% accuracy in this dataset without any particular tuning or data preprocessing.