Football & Numbers #1 · Goals follow a pattern
In 1982, statistician M.J. Maher showed something simple: the number of goals a team scores in a match behaves like a Poisson distribution, the same maths used for rare, random events.
Give a team its expected goals for a match, and the model gives the chance of every score.
A team expected to score 1.5 goals:
0 goals: 22%
1 goal: 33%
2 goals: 25%
3+ goals: 19%
Most modern football models start from this idea.
Next: why 0-0 and 1-1 break the rule.
Data, not a tip · 18+
pattern90.com/dropping-odds
#football #footballstats #footballanalytics #soccerstats #poisson
In 1982, statistician M.J. Maher showed something simple: the number of goals a team scores in a match behaves like a Poisson distribution, the same maths used for rare, random events.
Give a team its expected goals for a match, and the model gives the chance of every score.
A team expected to score 1.5 goals:
0 goals: 22%
1 goal: 33%
2 goals: 25%
3+ goals: 19%
Most modern football models start from this idea.
Next: why 0-0 and 1-1 break the rule.
Data, not a tip · 18+
pattern90.com/dropping-odds
#football #footballstats #footballanalytics #soccerstats #poisson
Football & Numbers #2 · Why 0-0 and 1-1 break the rule
In #1 we saw that goals roughly follow a Poisson distribution. But in 1997, statisticians Mark Dixon and Stuart Coles found a weak spot: the simple model treats the two teams as completely independent, and real matches are not like that.
The result: low scores come out wrong.
• 0-0 and 1-1 happen more often than plain Poisson says
• 1-0 and 0-1 happen a little less often
Their fix was a small correction for low scores, plus one more idea that still matters today: recent matches should count more than old ones.
Example (home team expected 1.5 goals, away 1.1):
0-0: 7.4% → 8.7%
1-1: 12.3% → 13.5%
1-0: 11.1% → 9.9%
0-1: 8.2% → 6.9%
Next in the series: from chess to football, the Elo rating.
Data, not a tip · 18+
pattern90.com/dropping-odds
#football #footballstats #footballanalytics #soccerstats #dixoncoles
In #1 we saw that goals roughly follow a Poisson distribution. But in 1997, statisticians Mark Dixon and Stuart Coles found a weak spot: the simple model treats the two teams as completely independent, and real matches are not like that.
The result: low scores come out wrong.
• 0-0 and 1-1 happen more often than plain Poisson says
• 1-0 and 0-1 happen a little less often
Their fix was a small correction for low scores, plus one more idea that still matters today: recent matches should count more than old ones.
Example (home team expected 1.5 goals, away 1.1):
0-0: 7.4% → 8.7%
1-1: 12.3% → 13.5%
1-0: 11.1% → 9.9%
0-1: 8.2% → 6.9%
Next in the series: from chess to football, the Elo rating.
Data, not a tip · 18+
pattern90.com/dropping-odds
#football #footballstats #footballanalytics #soccerstats #dixoncoles