Thunder Elo

Elo is one number for how strong a team is, kept up to date one game at a time. Everyone begins at 1500 — an average team, by definition. The gap between two ratings is a prediction: a team rated 100 points above its opponent should win about two games in three, and 200 points above about three in four. After the game the winner takes points from the loser: a lot when the result was a surprise or the margin was wide, few when a heavy favourite won as expected. Nobody assigns the number. It is what is left after every game since 1900 has had its say.

So the scale is relative, not a score. Wheaton has ended a season as low as 1301 and as high as 1797, with a typical year near 1598; the rating peaked mid-season at 1809 in 2019. A rating is only ever a claim about the games played, and the exact arithmetic is at the bottom of the page.

Every team starts at 1500. Before each game the expected result is E = 1 / (1 + 10^((opp − own − home) / 400)), with home field worth 55 points. After the game the rating moves by K × m × (result − E), K = 20, where m is the margin multiplier ln(|margin| + 1) × 2.2 / (0.001 × favourite's lead + 2.2) — a bigger margin moves the rating more, a blowout by a heavy favourite less. Between seasons every rating regresses a third of the way back to 1500.

Opponents are rated the same way, but only from their games against Wheaton — the archive holds no one else's schedule. That makes the ladder Wheaton-centric: an opponent met once a year carries a thin rating, and a first meeting carries none. The Graph's schedule strength inherits the same limit and says so.

Nothing here is stored as fact. Rerun the archive and the ladder is recomputed from the game log; change a score in the record and the ladder follows.