A Decision Tree for Choosing Chainle Answers

Classify, generate, test, balance, commit — the same five moves every bridge round.

Last updated 2026-11-11.

The tree at a glance

The diagram compresses the whole routine into six boxes. The rest of this page explains what each step means in practice and what to do when a step fails.

Step 1: read all the clues before thinking about words

Rounds 1 to 3 show two clues, round 4 shows three, and the temptation is to start answering after the first one. Resist it. The scoring formula rewards answers that connect to every clue, so in round 4 the first clue is only a third of the requirement. Read all the clues and ask one question: what single idea could plausibly touch all of them at once?

Step 2: classify the bridge type

Before generating any candidate, decide which kind of bridge you are probably looking at. The eight-type classification gives you concrete prompts: shared material asks “what are these made of?”, part-or-function asks “what do both contain?”, second meaning asks “which word lives in both worlds?”, familiar phrase asks “what word naturally follows each clue?”, and the shared-* family asks about properties, places, actions and associations. Naming the type first narrows the search from “any word in the language” to “words of this shape”.

Step 3: list two or three candidates, not one

A single candidate is an anchor, not a search. Produce two or three, ideally drawn from different types — if one candidate comes from the material route, force yourself to also try a part-or-function or phrase candidate before committing. Holding two or three in mind keeps you honest at the next step, because there is something to compare against.

Step 4: test each candidate against every clue

This is the step most players skip. Take each candidate and check it against each clue separately: does the connection hold on its own, or does it survive only because of one clue? The geometric mean in the scoring formula is unforgiving here — a candidate that is perfect on one clue and implausible on the other scores dramatically worse than an evenly decent one.

Step 5: judge by the weakest link

Compare candidates on their worst connection, not their best. The formula is 80% geometric mean of the per-clue strengths plus 20% of the weakest clue, so a balanced 80 + 80 beats a spectacular 100 + 20 by roughly a factor of two. When two candidates feel similar overall, pick the one whose weakest link is stronger. This one habit is worth more than a large vocabulary.

Step 6: commit and read the result

Type the winner and read what the result screen tells you. It names the details that matter: when the specificity adjustment was applied, and in the close-meaning round, whether your word kept the cued sense. Every result is information for the next round — the tree gets faster each time you close the loop and remember which route worked.

When nothing fits

Round 5 breaks the pattern on purpose: it hands you a meaning cue instead of clue words, so the classify-test loop is replaced by sense-matching, which the final-round guide covers separately. When the loop keeps failing anyway, the common-mistakes page names the seven habits that usually break it.