The puzzle, played in reverse
Reverse Sokoban inverts the original: boxes start on their targets, and your job is to pull them off the targets and back to the start positions. The mechanic is symmetric — pull instead of push — but the cognition is entirely different.
In Classic Sokoban, you predict forward: "if I push this box east, can I still solve the rest?" In Reverse Sokoban, you predict backward: "to end with this box here, where must it have started?" — which is exactly the same mental skill top solvers use in Classic, but made explicit and drillable.
Why it's a training tool
Top Sokoban solvers routinely solve levels mentally by working backward from the solved state. Reverse Sokoban turns that into the actual game:
- Forward planning becomes backward planning.
- Deadlock detection becomes reachability analysis.
- Pattern recognition for "this configuration is solvable" develops faster.
Many Sokoban training apps include a Reverse mode specifically for this reason — it's the equivalent of solving chess puzzles for tactical pattern recognition.
How to play
- Same grid, same keeper, same boxes.
- Boxes start on the target tiles — the puzzle is already "solved" visually.
- The keeper can pull boxes (move away from a box while adjacent to it; the box follows).
- Pushing is still allowed, but the goal is to return every box to its starting position.
- The puzzle is solved when every box is at its designated start cell.
- Same single-box rule: only one box can be pulled at a time.
Strategy tips
- Re-derive the start position before moving. Most Reverse Sokoban levels make the intended start obvious; confirm it mentally first.
- Watch for "push-deadlocks" in reverse. A box pulled into a corner cannot be pulled out — same as Classic, mirrored.
- Plan pull-order. Some boxes block the pull paths of others; clear the outermost boxes first.
- Use the same forward-visualisation skill. Before pulling a box, ask: "could I push this box back to its current position from where I'm pulling it to?"
- Play Classic and Reverse in alternation. The two modes reinforce complementary skills.