I wrote a program that takes tens of thousands of pro-level games, extracts the first twenty moves and assembles them in a single tree. The exact move order is relevant. If two games reach the same whole-board position with a different order of moves, they will end up in different parts of the tree.
Each move that has more than one follow-up variation also contains statistics about how often each move was chosen. And the final move of each variation contains the date and names of the players who played that opening.
Here I have chosen games from 2016 – the year that AlphaGo made its impact – to the current year so we can study what kind of openings are played in the AI age. Another way to use this tree is to load it in the Sabaki editor, go to a position that you want to explore and attach Leela Zero. This way you get to play Leela Zero but with a more varied opening repertoire.
