ELO Ratings Explained: How We Measure Team Strength

At the heart of our prediction system is ELO - a rating system originally developed for chess that we've adapted for Premier League football.

How ELO Works

The Basics

• Every team starts at 1500 points (league average)
• After each match, points transfer from loser to winner
• The amount transferred depends on the expected vs actual result

The Formula

When Team A plays Team B:

1. Calculate expected score based on rating difference
2. Compare to actual result (win = 1, draw = 0.5, loss = 0)
3. Adjust ratings proportionally to the surprise factor

If Manchester City (1650) beats Luton (1350), that's expected - small rating change. If Luton beats Manchester City, that's a massive upset - large rating change.

Our Enhancements

We've customised ELO for football:

• K-factor = 30: How much ratings change after each match
• Home advantage = 90: Home team gets +90 ELO equivalent
• Progressive building: Ratings reset and rebuild for each prediction date

Current Approximate ELO Rankings

Based on our model (as of January 2025):

Team	Approx. ELO	Tier
Manchester City	1680+	Elite
Arsenal	1650+	Elite
Liverpool	1640+	Elite
Chelsea	1580+	Strong
Tottenham	1560+	Strong
Newcastle	1550+	Strong
Aston Villa	1540+	Mid-table+
Manchester United	1520+	Mid-table+
Brighton	1510+	Mid-table
West Ham	1500	Average
...	...	...
Promoted teams	1350-1400	Below average

Note: These are illustrative - actual ratings update after every match.

Why Progressive ELO Matters

Many ELO systems use "current" ratings. We do something different:

When predicting Gameweek 10, we rebuild ELO from scratch using only Gameweeks 1-9 data.

Why? Because in real life, you don't know future results when making predictions. Our accuracy figures (52.4%) reflect this honest approach.

ELO to Expected Goals

We convert ELO ratings to expected goals using:

Home expected goals = (Home_ELO / Away_ELO) × Home_advantage × League_average
Away expected goals = (Away_ELO / Home_ELO) × League_average

This gives us the λ (lambda) values for our Dixon-Coles probability calculations.

Limitations of ELO

ELO doesn't capture everything:

• Injuries and suspensions
• Manager changes
• Player transfers
• Tactical matchups
• Motivation differences

But it does capture the most important factor: how strong is this team based on their actual results?

That's why ELO-based predictions consistently outperform gut feeling.

See ELO in action - check out this week's Premier League predictions powered by our hybrid ELO + Dixon-Coles model.