Chess invented it, football borrowed it, and betting models have used it for twenty years. The ELO rating answers one question in a single number: how strong is this team, right now?
How it works
Every team starts from the same rating, usually 1500. After each match, the winner takes points from the loser. The clever part is the amount transferred: beat a much weaker side and you gain almost nothing; beat a much stronger one and you gain a lot. A rating therefore climbs only when results are genuinely better than expected.
From two ratings you get a probability directly. A 100-point gap means roughly 64% for the stronger side; 200 points, about 76%. No opinion, no narrative — just a number that updates itself after every match.
Why it survived everything else
Three reasons. It is self-correcting: a team that declines loses points automatically, without anyone deciding it has declined. It needs no data beyond results — useful in leagues where statistics are thin. And it is hard to fool: a lucky win against a weak opponent barely moves it.
What it cannot see
ELO knows results. It does not know that the first-choice goalkeeper is injured, that the match is a derby, that the team played three days ago in another country. It reacts to what has happened, never to what is about to.
That is why it is never used alone. At PROLIFICK, ELO is one signal among several: surface-aware ratings in tennis (clay and grass are two different sports for a rating), recent form, absences, and — for football — the Poisson method for scorelines. The combined probability is then compared with the real odds to look for value.
The takeaway: ELO is the honest backbone of a model. It tells you who is strong. It is the rest of the work that tells you who will win tonight.