Chainle Scoring Explained: The 80/20 Formula

Chainle scores every answer out of 1,000 with an 80/20 blend of average and weakest connection strength.

The formula

Every Chainle answer is scored from 0 to 1,000. The current scoring version — in effect from the daily season that began on October 3, 2026 — computes the score as:

80% of the geometric mean of the per-clue connection strengths, plus 20% of the weakest-clue strength, multiplied by ten.

The geometric mean is the part that matters. It is pulled toward the smallest value in the set, so one weak clue drags the whole answer down.

Why balanced beats bold

Compare two answers to the same two clues:

Answer	Clue 1 strength	Clue 2 strength	Geometric mean	Score
Balanced	80	80	80	800
Lopsided	100	20	≈ 44.7	≈ 400

The lopsided answer has a perfect connection to the first clue and barely registers on the second — and it scores about half of the balanced answer. This is the single most important habit the formula teaches: check the weakest link before you commit, not the strongest.

What earns the full 1,000

Three routes lead to the maximum:

Recognized answers — the featured answer plus selected alternatives — always earn the full 1,000.
Estimates for unlisted words can reach the full 1,000 in the current version. This is new: earlier rules capped estimates at 970.
In close-meaning rounds, answers are scored against reviewed alternatives and a specific dictionary sense.

The flip side: known opposites in a close-meaning round receive at most 100 points.

The specificity adjustment

In bridge rounds, the estimated strengths of indefinite words such as “something” and broad evaluations such as “good” are multiplied by 0.6 before scoring. The adjustment never applies to close-meaning rounds, and it never overrides a recognized answer. The result screen tells you when the adjustment was applied.

What a connection strength is — and is not

The “80/100”-style figures in the formula measure how strongly the scoring system links your answer to each clue. They are not player percentiles and not a leaderboard rank — an 80 says nothing about how other players answered the same round.

The machinery behind the strengths: a scoring vocabulary of 100,000 dictionary-filtered words, a compact version of the Stanford GloVe 2024 WikiGiga vectors, and WordNet 3.0 for word forms and senses. Checking runs on the official server — no outside AI service is involved.

Keep reading

The cluster pages cover the failure cases (why answers score low), the variety of accepted answers, and the version history. For turning scores into better choices, continue to the strategy cluster.

In this section

Why Did My Chainle Answer Score Low?
Most low scores trace back to a handful of causes, and the result screen usually names the culprit.
Chainle Alternative Answers: More Than One Way to Score 1,000
There is rarely a single correct Chainle answer — featured words, alternatives and estimates all live on the scoreboard.
Chainle Rule Versions: What Changed on October 2, 2026
The daily season that began on October 3, 2026 plays by new scoring rules — older dates keep the old book.