How scoring works

A round starts from a base. Penalties come off it, then the tokens the replies used stretch or shrink what is left.

1. BaseName the model, on any reply, and the base is 2,000. How many messages you used does not change it. If you never name it, the base is the region (100) and the company (300) your guesses matched. A trait the round already showed you, or that you revealed, adds nothing.
2. PenaltiesEach wrong guess costs 100. A hint costs what its trait is worth: year 100, size 100, region 400, company 600, family 600. The right name is never a wrong guess. Earlier wrong names still count. Penalties are subtracted from the base, before the token factor.
3. TokensMultiply by 1,000 divided by the tokens the replies used. Fewer tokens score higher. Anything under 100 tokens, including 0, counts as 100, so the factor tops out at 10.
4. No-hint modePlay without hints and the roster never shrinks and no hint buttons are offered. A round played that way is multiplied by 1.5. Buying a hint means you are in the other mode, so there is no multiplier.
5. RoundingThe result rounds to the nearest whole number, halves up. It never goes below 0.

Hints

Company, family, region, year, and size can each be revealed once, after the first reply. A revealed trait is known from then on, and the guess list drops models that do not match it. A trait the round already showed you has no hint.

One round

Two replies used 500 tokens. The first name was wrong. You bought the year. The second name was the model. You were not in no-hint mode, because you bought a hint.

base      = 2,000                    (named the model)
penalties = 100 + 100 = 200          (year hint, one wrong guess)
tokens    = 1,000 / 500 = 2
score     = round((2,000 − 200) × 2)
          = 3,600

Your own result page lists the lines for that round.