BOAT PENTE AI 101: Training a Deep Learning AI for Boat Pente

About a year ago, a member of  a group of college students announced in pente.org's forum plans "to develop an adaptation of Google-owned DeepMind famous AlphaZero algorithm to Pente." I'm not sure whether or not anything ever came of those plans, but the OP only posted  once, so probably not. 

Now, slightly over a year later I am putting out a call seeking students interested in creating a slightly different project.

This will be an expert trained Deep Learning program for the Pente variant Boat Pente which, after decades of playing Pente I believe to be a deeper and fairer variant than the original game which nonetheless contains within it many game lines and partial game lines of the original game. Some lines remain unchanged by Boat Pente rules, some lines are slightly extended by Boat Pente rules but have the same end result - a win for the player who would win the line in regular Pente and some lines are transformed into more complex lines, some of which result in a different winning player for that line.

The database of played games of Boat Pente is very small compared to the database of Pente games, little work has so far been done on an opening book of moves suitable for Boat Pente and to my knowledge no computer program currently exists which plays this particular variant of Pente.

Therefore, I will be willing to assist in the creation of an opening book of interesting positions which an AI can begin to train with and provide assistance in identifying moves and lines which are most likely to be altered by Boat Pente's rule difference from regular Pente.

Any CS students or groups who are interested in developing "an adaptation of Google-owned DeepMind famous AlphaZero algorithm to" Boat Pente may contact me here or at pente.org, where I play as watsu.

Thank you for your interest.

Tom Cooley