Elo Ratings Explained: How Chess Maths Ranks Football Teams (and Helps You Bet)

On 7 October 2026, Spain climbed to 2287 in the World Football Elo Ratings. No national team has ever been rated higher in that system. The record they took belonged to Hungary, who reached 2230 on 30 June 1954. Germany's 2223 from July 2014 and Brazil's 2195 from November 2022 sit further back.

There's a lovely twist to it. The system that has just crowned Spain the strongest international side on record was dreamt up by a Hungarian-born chess player who spent his working life teaching physics in Milwaukee. He never meant to rate footballers. He wanted a fair way to rank chess players, and the maths happened to travel.

It travelled all the way to the betting shop. Give Elo two ratings and a venue and it hands you a fair price for Arsenal v Manchester City in one line of maths. Hold that up against the bookies' odds and you've got the start of a proper betting edge, as long as you know exactly what the number can't see.

Arpad Elo was born in 1903 in Egyházaskesző, Hungary, and taught physics at Marquette University in Milwaukee from 1926 until he retired in 1969. He could play, too: eight times Wisconsin State Champion. The US Chess Federation had been using the Harkness system, devised by Kenneth Harkness in 1950, and Elo replaced it with something on a properly statistical footing. The USCF took up his method around 1960, FIDE adopted it in 1970, and Elo ran the international calculations himself until the mid-1980s. He set the whole thing out in The Rating of Chess Players, Past and Present (1978).

The idea at its heart is simple. Every competitor has a true strength, and on any given day they perform somewhere around it, a bit better or a bit worse. You never see that strength directly. You only see results. So the rating shifts after every result, by an amount that depends on how surprising that result was.

That makes the system self-correcting. Rate a team too highly and they'll keep falling short of their billing, shedding points until the number fits. Rate them too low and they'll claw it back.

Football caught on in 1997, when Bob Runyan adapted Elo for international matches. His system, now kept by Kirill Bulygin at eloratings.net, inspired ClubElo (clubelo.com) for club sides. One thing to remember: there's no single official football Elo. Each site tunes its own version, so 1900 on one isn't 1900 on another. Compare teams within one system, never across them.

Strip it right down and Elo is three steps. Work out what should happen, see what did happen, and adjust by the difference.

Step one: the expected score

dr        = rating_home + 100 - rating_away      # +100 home advantage; 0 at a neutral ground
expected  = 1 / (1 + 10 ** (-dr / 400))

The expected score lands between 0 and 1. A win counts as 1, a draw as 0.5 and a defeat as 0. Two evenly rated teams on neutral ground both expect 0.5. The 100-point home bonus is World Football Elo's figure for home advantage, and it's a hefty one, as you'll see shortly.

Steps two and three: the update

new_rating = old_rating + K * G * (result - expected)

G = 1                 if the winning margin is 0 or 1 goals
G = 1.5               if the margin is 2 goals
G = (11 + margin) / 8 if the margin is 3 or more    # 1.75, 1.875, 2.0 ...

K is how much the match matters. In World Football Elo it's 60 for World Cup finals matches, 50 for continental championship finals, 40 for World Cup and continental qualifiers and major tournaments, 30 for other tournaments and 20 for friendlies. So a World Cup finals game moves a rating three times as far as a friendly.

G rewards a thumping. A 4-0 counts 1.875 times as much as a 1-0, but the multiplier flattens out, so piling on goals brings diminishing returns. A tie settled on penalties goes down as a draw.

Whatever one team gains, the other loses. The system is zero-sum, so the average rating across the pool stays put. New teams count as provisional until they've played around 30 matches.

This is why bettors should love Elo: a gap in points turns straight into a percentage. Here's the 400-scale curve that chess and World Football Elo both use:

Rating gap	Expected score	Rough feel
0	50.0%	Coin toss
50	57.1%	Slight edge
100	64.0%	Clear favourite
200	76.0%	Strong favourite
300	84.9%	Heavy favourite
400	90.9%	Mismatch
800	99.0%	Walkover

Look at what home advantage does. That flat 100 points is worth 14 percentage points of expected score between two otherwise equal sides. If you've ever wondered what the crowd, the travel and the familiar pitch add up to, Elo gives you a working answer, and it sits neatly alongside our guide to home advantage versus team quality.

Now for the bit that catches people out. An expected score is not a win probability. It's the win chance plus half the draw chance. A 64% expected score could be a 54% win chance with 20% for the draw, or 44% and 40%. Chess players can live with that. Anyone pricing a three-way football market can't, and we'll sort it out in the worked example.

Elo's biggest endorsement came from FIFA. The men's world ranking switched to an Elo-based method from 16 August 2018, the first release after the World Cup in Russia. The women's ranking had been running on a modified Elo formula since 2003.

It made sense. A 2009 comparison of eight rating methods found World Football Elo had the best predictive record of the lot, while FIFA's own 2006-2018 method did poorly. The formula FIFA moved to will look familiar:

P_new     = P_before + I * (W - We)
We        = 1 / (1 + 10 ** (-dr / 600))

The differences are all in the tuning. FIFA uses a 600 scale instead of 400, which makes the system less jumpy about rating gaps: 300 points on FIFA's scale is worth the same 76% as 200 points on the chess scale. There's no home advantage and no goal-difference multiplier. A shoot-out win counts as 0.75 and a shoot-out defeat as 0.5, and teams don't drop points for knockout defeats at final tournaments.

Importance weights run from 5 for a friendly outside the international windows up to 60 for World Cup matches from the quarter-finals on. Nations League group games carry 15, World Cup qualifiers 25.

For betting, we'd take eloratings.net over FIFA's list every time. Home advantage and winning margin are real, and a rating that ignores both is throwing information away.

Seven weeks or so into the 2026/27 European season, the top of ClubElo's ratings on 9 October 2026 looks like this:

Rank	Club	ClubElo rating
1	Bayern Munich	2046
2	Arsenal	2040
3	Manchester City	2028
4	Barcelona	2024
5	Real Madrid	1972
6	Paris Saint-Germain	1956
7	Inter	1946
8	Liverpool	1937

Six points between Bayern and Arsenal is nothing in Elo terms: a 50.9% expected score on a neutral pitch. The real story is the drop after Barcelona. Real Madrid are 52 points below them, and Liverpool sit 103 points behind Arsenal.

Further down, the Premier League's chasing pack is jammed together. Aston Villa (1895), Brighton (1891), Chelsea (1881), Newcastle (1879) and Manchester United (1879) are split by just 16 points. In that group, Elo sees home advantage as the deciding factor in most of their meetings with each other.

Internationally, Spain's 2287 puts them 112 points clear of Argentina (2175), with England third on 2146 and France fourth on 2104.

Put the world's second- and third-best club sides, on ClubElo's numbers, at the Emirates. We'll borrow World Football Elo's +100 home advantage for the exercise.

Arsenal 2040, Man City 2028
dr        = 2040 + 100 - 2028 = 112
expected  = 1 / (1 + 10 ** (-112 / 400)) = 0.656

Arsenal's expected score comes out at 65.6%. Strip out the home bonus and it falls to 51.7%, which tells you how much of this fixture's lean is down to the venue rather than the squads.

Splitting out the draw

A three-way price needs a draw estimate. Say we pencil in 26% for a tight game between two elite sides. That's our assumption, and you can swap in your own. Half the draw chance belongs to each side's expected score, so:

Arsenal win = 0.656 - 0.13 = 0.526   ->  fair odds 1.90
Draw        = 0.260                  ->  fair odds 3.85
City win    = 0.344 - 0.13 = 0.214   ->  fair odds 4.67

What the result would do to the ratings

Take a K of 40 just to show the mechanics. A 2-0 Arsenal win carries G = 1.5; a draw or a one-goal City win carries G = 1:

Arsenal win 2-0:   40 x 1.5 x (1   - 0.656) = +20.6 Arsenal, -20.6 City
Draw:              40 x 1   x (0.5 - 0.656) =  -6.2 Arsenal,  +6.2 City
City win by one:   40 x 1   x (0   - 0.656) = -26.2 Arsenal, +26.2 City

A draw costs Arsenal points because they were expected to do better, and a narrow City win hits them harder than a comfortable home win rewards them. That lopsidedness is Elo's whole philosophy in a sentence. Favourites gain little for doing what they should and lose plenty when they don't.

Here's where it pays. Elo gives you a fair price and the bookmaker gives you a real one. Your job is to compare them properly, which means taking the margin out of the bookmaker's odds first.

Picture a book pricing our Arsenal v City match at 2.10 / 3.50 / 3.60. These are illustrative numbers, not a live market:

Implied: 1/2.10 + 1/3.50 + 1/3.60 = 47.6% + 28.6% + 27.8% = 104.0%
Fair (divide each by 1.040):        45.8%   27.5%   26.7%

That 4% overround is the bookmaker's cut. With it removed, the market has Arsenal at 45.8% and our Elo model has them at 52.6%. A 6.8-point disagreement deserves a proper look. If the mechanics need a refresh, our complete guide to odds, probability and EV goes through them in depth.

The expected value check is then one line:

EV per £1 = (your probability × decimal odds) - 1
          = 0.526 × 2.10 - 1 = +£0.10

Demand a clear gap before you bet. Elo is a blunt tool, and small differences between your number and the market are just noise. We want three to five percentage points of daylight at the very least, and even then the next move isn't to bet. It's to work out why the market disagrees.

The cleanest way to find out whether your Elo prices are any good is closing line value. Log the odds you took and the price at kick-off. If the market keeps drifting towards your number, the model is seeing something real. If it drifts away, the market knows something Elo doesn't.

Elo needs nothing but results. That's its great strength and its biggest weakness.

Team news

Elo has no idea who's playing. A season-ending injury to your best striker, a rotated XI before a Champions League night, a record signing: none of it touches the rating until results come in. The market prices team news within minutes.

Shocks and newcomers

A sacking, a takeover or a summer rebuild all reach the number slowly, one result at a time. Newly promoted sides are harder still. Their rating was earned against a weaker division, and leagues only connect through cup games, European ties and promotion and relegation.

Results, not performances

A 1-0 smash-and-grab where you were battered on chances updates exactly like a 1-0 you dominated. The goal-difference multiplier helps a little, but it knows nothing about shot quality. Pair Elo with xG and each covers the other's gaps.

The early weeks

In August and September the ratings mostly reflect last season. That's when Elo is most exposed, and when the market, which has already priced in every summer signing and managerial change, is at its sharpest compared with a pure results model.

None of that makes Elo useless. It makes it a baseline, which is exactly what you need when you start building your own betting model. Even the cleverest approaches, the ones in our piece on whether machine learning can beat the bookmakers, usually have a strength rating like this sitting underneath.

Our verdict is simple: Elo is the best free baseline a bettor can own. It's transparent, it corrects itself, it's on two free websites and it turns any fixture into a fair price in seconds.

• Use Elo as your first price, never your last. Work out the fair odds, compare them with the de-margined market, then go looking for the team news that explains any gap.
• Big disagreements in the first six weeks favour the market. Summer signings and new managers aren't in the ratings yet, so be slow to back Elo against a market move in August and September.
• Home advantage is where Elo earns its keep. In tight bunches like Villa, Brighton, Chelsea, Newcastle and United, the venue is often the whole story. Check the market is giving the home side enough credit.
• Don't compare numbers across sites. ClubElo, eloratings.net and FIFA all run different versions. Pick one system and stick with it.
• Be wary of promoted sides' ratings. They were earned in a weaker league and are the likeliest to be out of step with reality early on.
• Back Elo when the market overreacts to one result. A single shock defeat can shove prices around. Elo moves calmly, and when a gap opens after one freak scoreline, the calmer number is often the better one.
• Judge it over a big sample. A few winning Elo bets prove nothing, as our guide to telling skill from luck in a betting record shows. Football is also a high-variance sport, which is why luck plays such a big part in results.

It's the same data-first mindset that turned Brentford and Brighton into Premier League success stories, only pointed at the betting market rather than the transfer market.

Whatever your edge, stake a fixed, sensible unit and let the numbers play out over a long run of bets.

• Probability in sports betting: the complete guide to odds, EV and value
• How to build a sports betting model: the complete guide
• Closing line value (CLV) explained
• Home advantage vs team quality: an advanced football betting guide
• AI in sports betting: can machine learning beat the bookmakers?
• Moneyball football: how data built Brentford and Brighton

18+. Please gamble responsibly. For free, confidential support visit BeGambleAware.org.