A game of logic, risk, and equilibrium.
Two players build rival constellations on one circular board. Each turn you both plan in secret, then reveal at the same time. There are three moves: Place, Secure and Capture.
Build the biggest connected group of stones and complete rays. Capturing a stone hands one of yours to your opponent. A Secure is wasted unless you guessed the attack. The player who reads the other best wins.
Everything is in view: the board, both supplies, the Box. The only thing you cannot see is your opponent's move this turn. Every risk is a calculated one.
Four levels. Easy, Medium and Hard are one neural network that learned Constello by playing against itself. Hard plays it as trained; Medium captures less; Easy also makes many random moves, so it is the one to learn against.
Expert adds a search: it thinks two turns ahead for both players and varies its choices, so it is hard to read. In every game against the AI the board is turned or mirrored at random from the AI's side, so it does not open the same way twice.
Constello combines the board sense of a territory game with the tension of moving at the same time. You choose between building, attacking and guarding without knowing what your opponent chose, so reading your opponent matters as much as reading the board.
The game is in playtesting. Every game played here is logged to help refine it, and your feedback shapes the design.
Constello Unplugged is the table edition, now in development: a handcrafted board, weighted stones, and mini-boards for planning in secret. Same rules; about 20 minutes a game. Join the waitlist to hear when it is available and to help shape the final design.
Leave your email to join the waitlist. One note when the table edition is ready, nothing else.
Played a few games? Found something confusing, broken or brilliant? I would love to hear about it. Bug reports, strategy notes and first impressions are all welcome.
Constello is an independent game design project.