Drop7 Research

About Drop7 Research

Drop7 Research is an open-source project investigating optimal strategies for Drop7, widely considered the greatest puzzle game of all time. The research explores fourteen different approaches to build autonomous players, ranging from expectimax search and heuristic evaluation to neural networks and evolutionary optimization. Users can play Drop7 in their browser, watch research progress, and explore the theoretical foundations behind each strategy. The project maintains a detailed log of experiments and results, with the current best candidate achieving an average of 516,155 points over 512 games, while experienced human players can score over a million points.

• ai-research
• game strategy
• search algorithms
• open-source
• puzzle game
• browser game
• machine-learning

FAQ

How does Drop7 work?

Drop numbered discs into the grid. A disc clears when its number matches the count of connected discs in its row or column. Falling discs can trigger another clear, building a chain reaction. Clears beside covered discs crack them, then reveal their numbers.

Is there an account required to play?

No account is needed to play. You can turn haptic feedback on or off in the app menu, and the app has no ads, in-app purchases, or subscriptions.

How is game history stored?

Game history is stored on your device where you played. There is no account-based history sync between devices. Open Game history from the game menu and select a completed game to watch its replay.

What data does the app collect?

The app sends interaction events and completed game records to the Drop7 Research project. Game records include the discs and moves needed to replay a game. These records do not include an account or device identifier.

What happens if the app stops responding?

Close and reopen the app, then check the App Store for an update. If Compete cannot load or a leaderboard submission fails, check your internet connection and try again.