Drop7 Research

Drop7 is widely considered one of the great puzzle games of all time 1 . The goal of this research is to find an autonomous strategy for playing the game as well as a human can.

An experienced human player can score millions of points in this game with the right long-term strategic thinking. The strongest research reference so far ( fair depth-4 expectimax ) averages about 309k points, so the research problem is very much still open.

Watch a game, or hover the board and click a column to drop the next disc.

Learn the rules and play the game right here in your browser, and see for yourself why this is such a captivating research prize.

All of the game engines, research approaches, scripted-round benchmarks, and this website are all open-source.

Learn the rules, and play the game right here in your browser.

Chance, look-ahead, evaluating a board, the sibling trap, and what compute can and cannot buy.

Twelve strategy families have been developed so far, and each has a hierarchy of records into specific approaches that were attempted.

Browse the theories and contribute your own, to guide the large-scale direction of the research.

There are monthly competitions where you can compete with the computer strategies and get yourself onto the same leaderboard.

Both leaders play gauntlet-01 , with the same visible discs and hidden values.