The Spreadsheet Beat FIFA: Why a Simple Elo Rating Out-Predicts the Official World Ranking
We tested the official FIFA world ranking against a plain Elo rating built from results across 26,603 international football matches since 1993. The spreadsheet won. Discover why a simple, transparent algorithm out-predicts the sport's official scoreboard and where FIFA gets it wrong.
The FIFA world ranking is the sport's official scoreboard of nations — it seeds World Cup draws, settles bar arguments and lets federations boast. But a ranking has one real job: to tell you who's likely to win. So we built the simplest honest competitor — a plain Elo rating computed only from match results — and ran the two head-to-head on every men's international since 1993, the year the FIFA ranking began. 26,603 matches, both ratings known before each kickoff.
What Elo actually is.
It's the chess rating, borrowed. Every nation carries a single number; when two teams meet, the gap between their numbers sets a pre-match expectation, and the result nudges both — winner up, loser down by the same amount. Beat a team you were supposed to beat and you gain almost nothing; beat one rated far above you and you gain a lot.
Ours starts every nation at 1500 and updates after each game. Three things decide how far a result moves the number: how big the upset was, how big the win was (a 4-0 counts for more than a 1-0), and how much the match mattered — a World Cup result is worth three times a friendly. Home advantage is worth about 65 points, and nothing on neutral ground. No reputation, no seeding, no committee: just who beat whom, by how much, in what.
Start with the cleanest possible test. Ignore the matches where the two rankings agree — they tell you nothing. Look only at the games where they disagree on which side is better, and ask: when they point at opposite teams, whose pick actually wins?
We Built a Better FIFA Ranking
FIFA's official world ranking is the sport's headline number. We pitted it against the plain strength rating we built from match results — head-to-head, on 26,603 internationals since 1993. One number predicts football better. It isn't FIFA's.
Each line is predicted win-chance vs what actually happened (the diagonal = perfect). Both rankings are well-calibrated; our model's edge is that it separates teams better, so it's right more often.
Method. Every men's international 1993→2024-07 with both ratings (n=26,603). Accuracy = share of decisive matches won by each ranking's favourite (home edge fit equally for both). Brier on a fair per-ranking logistic fit. FIFA points and Elo both pre-match. FIFA ranking data: Kaggle (CC0), ends 2024-06; Elo computed from 49k results (martj42).
The spreadsheet picks more winners
Across all 26,603 internationals, our Elo's favourite won 77.3% of decisive matches; the FIFA ranking's favourite won 74.1%. Elo's probability forecasts are better calibrated too (a lower Brier score, 0.1307 vs 0.146). Flip the toggle and the edge survives everything — major tournaments (72.8% vs 71.5%), and every era. It is not a fluke of one decade.
FIFA fixed its ranking in 2018 — by copying Elo
For years the old ranking was a punchline, gameable by picking soft friendlies. In 2018 FIFA scrapped it for a points-exchange system — which is to say, Elo. It genuinely helped: the accuracy gap shrank from +3.5pp before 2018 to +1.9pp after. But "we rebuilt our ranking to be more like Elo, and it's still behind Elo" is not the endorsement FIFA might hope.
Where the ranking gets it wrong
The misses are vivid. FIFA had the Netherlands 4th going into Euro 2012; they lost all three. It had reigning champions Greece 8th at Euro 2008; they went home with zero points. And it routinely fails to see a riser coming — it had Spain only 8th before they won Euro 2024, and buried African sides 100-plus places below teams they'd go on to beat. Sort the ranking's best and worst calls below.
FIFA's Best and Worst Calls
Every team's pre-tournament FIFA rank against how it actually played (points per game). The gap is where the ranking got it wrong — the contenders it flattered, and the teams it never saw coming.
How it's measured. Across every World Cup, Euro, Copa América and Africa Cup of Nations since 1993, each team's tournament points-per-game vs the points-per-game its pre-tournament FIFA rank would predict (pooled regression). Bigger bar = further from what the ranking expected. Rank is ordinal, so low-ranked over-performers show the largest gaps — that's the ranking's blind spot, shown with each team's actual rank.
None of this means rankings are useless — Elo and the modern FIFA formula agree far more than they differ. It means the official number isn't the best available, and a transparent rating you could build in an afternoon quietly out-predicts it. The most-quoted number in international football is not the most accurate one.
Method. Every men's international 1993→2024-07 with both ratings (n=26,603; 1,694 at major tournaments). Elo is computed from results only (martj42, ~49k matches), updated through each match; FIFA ranking points/rank are the most recent before each match (Kaggle, CC0, data ends 2024-06). Accuracy = share of decisive matches won by each ranking's favourite, with a home-advantage term fit equally for both. Brier from a fair per-ranking logistic fit. Tournament “best/worst calls” = each team's points-per-game vs what its pre-tournament FIFA rank predicts. — Fastmaster Intelligence.
