How Elo ranking works

Eloranker Elo for a personal list: start 1000, K=32, scale 400, no draws.

This is the rating Eloranker uses on a list that never leaves the browser. Two items appear. You pick one. There is no draw. Each new item starts at 1000, K is 32, and the scale is 400.

Elo is a rating system. Each item starts at 1000. When A beats B, A takes points from B. The underdog takes more.

What is an Elo rating?

An Elo rating is a single number that estimates how often one item should beat another. It is not a vote count. Two items with the same rating are treated as even. After each pick, both numbers move: the winner up, the loser down.

Eloranker starts every new item at 1000.

What does a 200-point gap mean?

A 200-point gap means the higher-rated item is expected to win about 76% of the time. That follows from the scale constant of 400:

expected(A vs B) = 1 / (1 + 10^((ratingB − ratingA) / 400))

If A is 1200 and B is 1000, the exponent is −0.5, so A’s expected score is 1 / (1 + 0.316) ≈ 0.76.

What is the K factor?

K is how far a rating is allowed to jump after one match. Eloranker uses K = 32. After a pick, each rating moves by at most 32 points.

newA = oldA + 32 × (scoreA − expectedA)
scoreA = 1 if A wins, 0 if B wins

If two items at 1000 meet, the winner goes to 1016 and the loser to 984. If a 1400-rated item beats a 1000-rated item, it takes far less, because that result was expected.

Why pairwise picks instead of a numbered list?

Writing 1–10 by hand hides indecision. A pairwise pick is one comparison. After enough comparisons, the ranking is the thing you actually prefer, not the order you typed on a tired evening.

Eloranker does not need every pair. It prefers items with fewer matches, then a nearby rating, so the next question is useful.

Where does the data live?

On this site: in the browser. There is no account. Export JSON to take a list to another machine. Read the privacy note if you want that in one page.