Information Gain
Definition
Information gain measures how much a guess is expected to shrink the set of possible answers, averaged over every result that guess could produce. It is the formal name for the intuition that a good guess 'tells you a lot' regardless of whether it happens to be correct.
Why it matters
Maximising information gain is the single most reliable way to lower your average solve count. Players who chase the answer too early sacrifice information; players who maximise gain consistently arrive at the answer with guesses to spare.
Examples
- An opener of all grey tiles still has high information gain because it eliminates five common letters.
- A guess that splits 100 candidates into many small groups has higher gain than one that leaves a single 90-candidate group.
Best practices
- Early in the puzzle, prioritise gain over chasing a hit.
- Build guesses around letters that are roughly 50/50 to appear.
- Late in the puzzle, weigh gain against the cost of not guessing the answer.
Frequently asked
Is information gain the same as entropy?
They are closely related. Entropy is the mathematical measure; information gain is the expected entropy reduction a guess produces.