Correlation analysis is the first thing anyone does with a sporting dataset, and the easiest thing to misuse. It answers one narrow question well: which measurements move together, and how strongly.
The mathematics, briefly
Pearson’s coefficient
- r = 1: perfect positive relationship
- r = −1: perfect negative relationship
- r = 0: no linear relationship at all
- |r| above 0.7: strong
- |r| between 0.3 and 0.7: moderate
- |r| below 0.3: weak
In football analytics the correlation between expected goals and actual goals sits around r = 0.87 — about as strong as any relationship in the sport.
Correlations that actually hold
Scored
First Deposit Sports Bonus: 100% up to €250
Ivibet
First Deposit Bonus for Sports Betting: 100% up to €150
BetRepublic
First Deposit Bonus: 100% up to €250
Slotsgem
Welcome package up to €1,450 + instant bonus round + 225 free spins
22bet
Welcome bonus up to €122 for sports betting
Slotrave
Welcome package: 450% up to €2,500 + 150 free spins + 1 bonus game
Winhero
150% welcome freebets up to €300 for the first three deposits
Playbaze
50% welcome risk-free bet
Spinbetter
Up to €500 for the first five deposits
GG.BET
Up to 200% bonus + €100 freebet
CrazyTower
First deposit bonus: 100% up to €200
18+. Advertising. Betting carries risk — gamble responsibly and never stake more than you can afford to lose.
Football
- Possession against wins: r ≈ 0.43 (moderate)
- Shots on target against goals: r ≈ 0.76 (strong)
- Passes against control of the game: r ≈ 0.69 (strong)
- Pressing against turnovers won: r ≈ 0.58 (moderate)
- Expected goals against goals: r ≈ 0.87 (very strong)
Note the first line. Possession is the statistic most often quoted as evidence of dominance, and it is the weakest predictor on the list.
Basketball
- Rebounds against wins: r ≈ 0.52
- Three-point accuracy against result: r ≈ 0.61
- Assists against offensive efficiency: r ≈ 0.73
- Turnovers against defeats: r ≈ −0.64
Tennis
- First-serve percentage against wins: r ≈ 0.68
- Unforced errors against defeats: r ≈ −0.71
- Winners against control of the match: r ≈ 0.59
- Break-point conversion against result: r ≈ 0.82
How correlation feeds a model
Feature selection
- Dropping inputs that barely relate to the outcome
- Removing duplicates that measure the same thing (multicollinearity)
- Ranking what is left by importance
- Reducing dimensionality so the model has a chance of generalising
Feature engineering
- Composite indices such as an attack-to-defence ratio
- Weighted averages that decay older matches
- Interaction terms, where two factors only matter together
- Normalised metrics that survive comparison across leagues
Validation
- Correlating predicted values with observed ones
- Examining the residuals rather than only the headline score
- Looking for systematic error in a particular league or period
- Checking whether the relationship is stable over time
A model whose predictions correlate above 0.75 with real outcomes is considered high quality in sports analytics.
Beyond Pearson
Spearman’s rank correlation
- Robust to outliers, which sport produces constantly
- Works with ordinal data such as league positions
- Detects monotonic relationships that are not straight lines
- The right default when comparing rankings
Partial correlation
- Holds a confounding variable constant
- Separates a real relationship from one borrowed from a third factor
- Essential before making any causal claim
- Standard in multi-factor models
A concrete case: shots on target correlate with winning, and so does possession. Once you control for shots, much of possession’s apparent effect disappears — it was largely a proxy for having the ball in dangerous areas.
The unavoidable warning
Correlation describes association, not mechanism. Two things it never establishes:
- Direction. Teams that are already winning tend to concede possession, so causation can run backwards from what you expect.
- Reality. Test enough pairs of variables and strong-looking correlations appear purely by chance.
The defence is unglamorous: state the hypothesis before looking, then check it against data you have not touched.
Conclusion
Scored
First Deposit Sports Bonus: 100% up to €250
Ivibet
First Deposit Bonus for Sports Betting: 100% up to €150
BetRepublic
First Deposit Bonus: 100% up to €250
Slotsgem
Welcome package up to €1,450 + instant bonus round + 225 free spins
22bet
Welcome bonus up to €122 for sports betting
Slotrave
Welcome package: 450% up to €2,500 + 150 free spins + 1 bonus game
Winhero
150% welcome freebets up to €300 for the first three deposits
Playbaze
50% welcome risk-free bet
Spinbetter
Up to €500 for the first five deposits
GG.BET
Up to 200% bonus + €100 freebet
CrazyTower
First deposit bonus: 100% up to €200
18+. Advertising. Betting carries risk — gamble responsibly and never stake more than you can afford to lose.
Correlation analysis is how a model decides what to pay attention to. It tells you that expected goals are worth modelling and that raw possession mostly is not.
Used carefully it is the sharpest cheap tool in sports analytics. Used carelessly it is the fastest route to a confident, wrong forecast.