“Impossible” can mean several very different things. A task may be beyond your current method, computationally expensive, impossible with a particular tool set, or logically incompatible with its rules. These are not interchangeable claims. Recreational mathematics offers a valuable way to practice the distinction: some puzzles yield to a better strategy, while others require a proof that no strategy can succeed.
The mutilated chessboard
Take an ordinary 8-by-8 chessboard and remove two diagonally opposite corner squares. Can the remaining 62 squares be covered exactly by 31 dominoes, each covering two orthogonally adjacent squares? Dominoes may be rotated, but they may not overlap, extend beyond the board, or cover only part of a square.
The count looks promising: 31 times two is 62. Yet counting area ignores a second constraint. Color the board in the standard alternating pattern. Each domino must cover one light square and one dark square. Opposite corners of an 8-by-8 board have the same color, so removing them leaves 30 squares of one color and 32 of the other.
Reveal the impossibility argument
Any placement of 31 dominoes covers exactly 31 light and 31 dark squares. The altered board has a 30-to-32 imbalance, so no allowed placement can cover it. This applies to every arrangement, including ones nobody has tried.
The coloring is an invariant-based obstruction. It does not tell you how close a particular arrangement will come to working, and it does not need to. It identifies a necessary condition for success and shows that the board violates it. However, satisfying a necessary condition is not automatically sufficient: removing one square of each color eliminates this particular objection, but a more general damaged board could still be untileable for another reason.
A bridge that only seems impossible
Four people need to cross a bridge at night. Their individual crossing times are 1, 2, 5, and 10 minutes. At most two people may cross at once, and every crossing must carry the single torch. A pair moves at the slower person's pace. The torch cannot be thrown, and there is no way around the bridge. What is the minimum total time?
A natural first plan sends the fastest person back after escorting each of the others. That produces 19 minutes. A better strategy asks a less obvious question: should the two slowest people cross together, while the two fastest manage the torch?
Find the shortest schedule
Hint: share the slow crossing
Try making the 5-minute and 10-minute people cross as a pair. Arrange for a fast person to bring the torch back afterward.
Reveal the 17-minute schedule
- 1 and 2 cross: 2 minutes.
- 1 returns: 1 minute.
- 5 and 10 cross: 10 minutes.
- 2 returns: 2 minutes.
- 1 and 2 cross: 2 minutes.
Total: 2 + 1 + 10 + 2 + 2 = 17 minutes.
A working schedule proves that 17 minutes is achievable. To establish minimality, separate two cases. If the two slowest never cross together, their forward crossings cost at least 5 + 10 minutes. At least one additional forward crossing costs at least 2 minutes, and at least two return crossings cost at least 1 minute each. The total is at least 19 minutes.
If the two slowest cross together, their crossing costs 10 minutes. The fast pair must also end on the far side, and someone must supply and recover the torch around that slow crossing. In a minimal schedule without redundant trips, the fast pair cross together twice and each returns once: at least 2 + 2 + 1 + 2 = 7 additional minutes. Sending a slow person back only increases the relevant costs. The displayed schedule reaches this 17-minute lower bound. This argument assumes the exact capacities and travel rules stated above.
Impossible with which tools?
The ancient problem of squaring the circle asks for a square with exactly the area of a given circle, constructed using only an ideal unmarked straightedge and compass in finitely many steps. The restrictions matter. Approximate drawings are easy, and a different tool set changes the question. The classical claim concerns exact constructibility under a specific mathematical model.
For a unit-radius circle, the square's side would have length √π. Straightedge-and-compass constructible lengths are algebraic over the rational numbers, with further restrictions on their field extensions. Ferdinand von Lindemann's 1882 proof that π is transcendental rules out the required length. If √π were algebraic, its square π would be algebraic too, contradicting that result.
Other classical tasks include doubling the cube and trisecting an arbitrary angle. Pierre Wantzel established key impossibility results for these constructions in 1837. Some particular angles can be trisected; the impossible task is a universal construction for every angle using the restricted tools. A statement can become false simply by dropping a word such as “arbitrary” or “exact.”
Search, then change the question
Try constructive methods first, but set a point at which repeated failure prompts a new approach. Ask what every legal move preserves, what any successful final state must satisfy, and whether those conditions conflict. Parity, coloring, and connectivity are especially useful starting tools.
Do not treat the absence of a known solution as proof that no solution exists. An open problem is not an impossibility theorem. Conversely, once a valid obstruction is established, another million attempts under the same rules will not help. The satisfying move is not surrender. It is recognizing that the most complete solution may be an explanation of why the requested object cannot exist.