Average expected score of 241.82 per game, compared to 254.59 using optimum strategy.

Testing mode computes average AI score and standard deviation, as well as the average score within each category. It takes the number of games to play as a parameter; each game takes .017 seconds to process.

Displays the potential score the player would get for filling in a box after every dice roll.

Optimizing computer AI performance using decision trees

Modifying existing functions to create a testing mode used for running simulations

Fine tuning computer decisions using testing mode output

Creating a web app using JavaScript and HTML

Professor Paul Linton at the University of Kentucky assigned a basic version of Yahtzee in his Web Dev course. He provided the HTML outline used in Meowtzee and a template for the original functions

Olaf Vancurra's book Advantage Yahtzee was key in making the decision tree

Tom Verhoeff published the original research on the Yahtzee Optimal Strategy

The expected average score was of 241.82 for the computer was calculated based a sample size of 100,000 games. The average standard deviation of 62.2 implies that a smaller sample size could result in a significantly different average.

Testing mode can be accessed here: Testing Mode. Each 100 games take about 1.7 seconds. Full output can be accessed by inspecting the console.

Decision tree can be better optimized to come closer to the expected average score using the optimal strategy

Using a Markov Decision Process and graph theory to approximate the perfect score

Including a multiplayer mode

AI is currently optimized to reach the highest possible average score. It could instead focus on beating the opponent