How it works
Most block puzzles deal pieces at random. Sometimes that hands you a set that cannot fit anywhere, and you lose through no fault of your own. BlockFit does the opposite: every deal is checked against a full-solution search before it reaches your tray.
The guarantee
For each deal, the engine tries up to 200 candidate piece sets. For every candidate it searches all placement orders - piece 1 first, piece 2 first, and so on - simulating the line clears that happen along the way. A candidate only becomes your tray if at least one order clears all three pieces. If the board is so crowded that no provable deal exists, the round ends honestly instead of dealing a dead hand.
Piece selection adapts to the board
| Board density | Piece sizes offered |
|---|---|
| Under 30% filled | 1 to 5 cells |
| 30% to 60% filled | 1 to 3 cells |
| Over 60% filled | 1 to 2 cells |
Scoring
| Action | Points |
|---|---|
| Placing a piece | +1 per cell in the piece |
| Clearing n lines at once | +10 × n² (two lines = 40, three = 90) |
Training modes
Classic starts from an empty board. Random drops you into a lightly scattered board. Endgame drill stacks the board to roughly 55% density using only provably solvable rounds, so you practice exactly the situations where random games usually fall apart.
Credits and license
The deal-vetting and solving algorithms are a clean-room TypeScript rewrite of dffge552/block-blast (MIT License, included in this site's source). The interface, scoring, training modes and all other code in this site were built separately. BlockFit is not affiliated with 1010! or Block Blast; it borrows only the classic block-placement genre, none of their names, graphics or assets.