Most puzzles are about shrinking a set of possibilities
A puzzle usually starts with a large set of possible answers and a set of rules that rule most of them out. Solving it means shrinking the set until one answer is left. Probability enters whenever you cannot shrink the set by pure logic and have to choose between options that are not equally likely.
Thinking this way changes how you approach a stuck position. Instead of asking “what is the answer?”, you ask “which answers are still possible, and which move tells me the most?” That question is useful in a crossword, a logic grid, a number-lock game or a mine-sweeping board.
Step one: count what is still possible
Before guessing, estimate the size of the problem. A four-position code with six colours, where colours can repeat, has 6 × 6 × 6 × 6 = 1,296 possible answers. If colours cannot repeat, there are 6 × 5 × 4 × 3 = 360. Knowing which of those you face tells you how aggressive you can afford to be.
You rarely need exact counts. A rough sense of “dozens”, “hundreds” or “thousands” is enough to decide whether to explore broadly or to start checking specific candidates. This is the same habit behind the probability basics, where listing outcomes comes before assigning chances.
Step two: prefer moves that split the possibilities
A good probe divides the remaining possibilities into groups of similar size, so that any result removes a meaningful share. The classic example is the number-guessing game. If someone picks a whole number from 1 to 100 and tells you “higher” or “lower”, guessing the midpoint each time finds the number in at most 7 guesses, because each guess roughly halves what is left and 2 to the power of 7 is 128, which covers 100. Six guesses would only cover 64.
Guessing 1, then 2, then 3 can also work, but on average it takes far longer. The lesson is general: a guess that could not possibly be right can still be the best guess if it eliminates the most candidates.
Step three: separate certain from likely
Many puzzle errors come from treating a likely deduction as a certain one. A useful habit is to mark your reasoning in two tiers: facts that follow from the rules, and hypotheses you are testing. Build on facts freely. Build on hypotheses only while you remember they may fail, and note which later moves depend on them.
Hypothesis testing works well in logic puzzles. Pick a square, assume a value, and follow the consequences. If the assumption produces a contradiction, you have learned the opposite with certainty. If it does not, you have learned nothing yet, which is why you should try the assumption that is most likely to fail quickly.
Step four: weigh a forced guess
Some puzzles, such as Minesweeper-style boards, occasionally leave no certain move. Then you are choosing between options with different risks. Suppose a pair of adjacent cells has exactly one hidden mine between them and nothing else is known: each cell is a 50 percent risk. Opening a cell elsewhere that has a lower risk, even a distant one, is usually better, because it may also reveal new information.
On a beginner board with 81 cells and 10 mines, an unconstrained random cell is a mine about 12 percent of the time (10 divided by 81). Compared with a 50 percent coin-flip, that is a sensible gamble. The reasoning is a small case of risk assessment, where you compare how likely a bad outcome is and what it costs. Outcomes are still uncertain, so even a good guess can fail; the aim is to make the choice defensible, not to guarantee it.
Step five: keep a record
Notes are not cheating in most single-player puzzles, and they prevent the most common failure, which is forgetting what you have already ruled out. Write down excluded options, tested hypotheses and the guesses you have used. A short list often reveals a pattern you were holding in your head and slowly losing.
Humans are also prone to seeing structure that is not there. If a puzzle generator is random, the next board does not “owe” you an easy layout, and a run of hard puzzles says nothing about the next one. This is the same trap discussed under random events, and it matters when you judge your own improvement.
Common mistakes
- Guessing early because the position feels stuck, instead of re-checking constraints.
- Committing to a single theory and ignoring evidence that conflicts with it.
- Optimising for a lucky win over steady elimination, which makes results harder to learn from.
- Judging a method from one puzzle. A handful of results is a small sample, and sample statistics explains why little can be concluded from it.
A simple routine
When a puzzle stalls, run through the same short sequence: list what is still possible, look for the move that splits the options best, separate what you know from what you suspect, and only then compare the risks of the remaining guesses. Over time the routine becomes automatic, and the skill of puzzle solving shifts from inspiration towards method. Variants and online formats add details, but the underlying questions stay the same.
Adapting the method to different puzzle types
The routine bends to fit the puzzle. In a logic grid, counting is easy and most of the effort goes into hypothesis testing. In a code-guessing game, the main decision is which probe to make next, so ranking guesses by how well they split the remaining candidates matters most. In a hidden-information board, the central question is which unknown cell carries the lowest risk, and whether opening it would also reveal something useful.
It also helps to match effort to the stage. Early in a puzzle, spend time on structure: what are the constraints, and which of them are tightest? Tight constraints, such as a row that is nearly full or a clue with only one way to be satisfied, are where certain moves tend to appear first. Late in a puzzle, the candidate list is short, and careful verification matters more than creativity. Re-checking a finished answer against every rule takes a minute and catches slips that would otherwise cost the whole attempt.
Finally, review after the fact. When a puzzle ends, look at the single decision that cost the most, whether it was a guess made too soon or a clue misread. One or two such reviews will teach more than playing many puzzles without reflection, and they keep your attention on the process you control rather than on results you do not.