In the Margins
What statistics are correlated with winning and losing in season 2025? And how do those correlations differ for different teams with different strengths and game styles?
This article comes with an acknowledgement and a few disclaimers. I wouldn’t have been able to do this without the incredible work of Andrew Whelan of WheeloRatings.com – having such a rich data source as a base meant I could take the time to pull together the analysis.
The disclaimer, for the purposes of this piece, is that I’ve used really simple linear regression with r2 as the basis for determining correlation. It’s not something you’d use to try to put a predictive model together, but it does enough to allow us to draw some interesting points.
Another disclaimer is that correlation is not causation, and doesn’t establish directionality. For example, West Coast’s margins are more strongly correlated to their ruck output than the rest of the league. Is that because when Bailey Williams and Matt Flynn have managed to win the battle, Harley Reid is able to go to work, or is it that an opposing ruck getting bested by them is emblematic of a team ripe to be beaten by West Coast?
It could also be that a given stat is a real non-negotiable for a team, it’s something they can be relied to win week in week out regardless of the end result – which would be reflected in a low correlation. The data can hopefully lead us to some interesting points for discussion, but can’t be definitive one way or the other.
Lastly,it is worth noting that I have used stat differentials (team minus opponent) rather than raw stats when correlating to margin, so keep that in mind.
With that out of the way, let’s get into the statistical correlations.
As you’d expect, kicking more goals than your opponent is very strongly tied to the final result. Champion Data’s rating points are also very closely correlated.
We can see that xScore has a higher correlation with victory than the pure number of shots, which we’d expect from a measure that incorporates not just the volume but the level of difficulty of shots taken.
Among score sources, Points from Turnover appear more valuable than Points from Stoppage, unsurprising as turnover is the primary scoring source. Points from forward half are a better predictor than points from defensive half.
xScore rating, that is how well the teams are executing on the shots at goal they generate, appears to be worth about as much as a gap in uncontested possessions, which is a better predictor than contested possessions or clearances.
Commit more clangers than your opponent and you’re likely to lose, however the correlation is relatively weak (to have a clanger you’ve generally got possession first).
Defensive half pressure acts is a rare example of a “positive” stat with a negative correlation to margin. If you’re racking them up, it means both that the ball is in your defensive half and the opponent has control of it.
We’ve got the league averages, so where and how does each team diverge on individual statistics?
The arrow indicates the direction a team diverges from – a red arrow to the left means that stats correlates less (or more negatively) with margin for the team than for the league at large and blue indicates stronger correlation.
Adelaide win through having a better spread of goalkickers than their opponents. They’ve had more unique goalscorers on 10 occasions for an eye-watering average margin of +62 points. Handballs are more valuable in their games than average, and kicks less so. The gap in value of points from forward half compared to defensive half expands.
They also don’t rely on a high mark inside 50 differential as much as the rest of the league. To revisit our disclaimer, this doesn’t necessarily mean they’re bad at it, just that it hasn’t correlated with winning and losing to the same degree it has for other teams. Adelaide has lost 5 games for the year – in three of them they won marks inside 50 and in a fourth they drew. They’ve lost marks inside 50 three times, and won two of those games. They’ve drawn it three times for a three point loss and two 10+ goal wins. They’ve also managed to win by 10+ goals with a +0,+0,+1, and +2 marks inside 50.
Brisbane aren’t converting xScore into wins particularly well (because they’re 4th worst in goal accuracy this year). They’re getting more value from centre clearances than most teams, and appear to not be as affected by turnovers. This is partly driven by the fact that they haven’t had a turnover differential larger than 8 in the positive or negative whereas a quarter of games league-wide have blown out past this.
Carlton don’t often lose more defensive 1 on 1s than their opponent, only on four occasions so far and never by more than two. Their biggest wins against West Coast and North saw them win the stat by 6 and 4 respectively. When they’ve been required to make more defensive half pressure acts than their opponent however they’ve got an average margin of -24 compared to +20 the other way.
Similar to Adelaide, Collingwood benefit from having a better spread of goalkickers than their opponents. All of Collingwood’s losses have come while winning the inside 50 count and three of the four came while also winning the marks inside 50 count, including a three point loss to Geelong while recording their best differential for the year (+9).
They’ve only lost the tackle count once all year, in their opening round drubbing by GWS. While they recorded a solid +21 tackles in their 1 point loss to Fremantle, the other two losses saw low differentials (for Collingwood) of 8 and 10. Three of their four biggest wins have matched up with their three biggest tackle differentials. Their pressure game also helps explain why they can lose the kick count convincingly and still come out on top.
Essendon want clean hands. Their average result is a 47 point loss when recording more ineffective handballs than their opponent, compared to just a 9 point loss when recording fewer. This is mirrored in effective disposal tallies. It’s not surprising, decimated by injury my best guess is that they just don’t have the drilled structures in place to respond to errors so when things go bad there is little damage mitigation.
It’s been a common theme of criticism that Fremantle can tend to rack up meaningless uncontested possessions. They’re 7-3 in games they win the count and 6-3 when losing it, but with a slightly better average margin. By comparison their average margin when winning contested possession is +26.2 compared to -10 when losing it.
Geelong benefit from winning the intercept game as well as tackles inside 50. When the Cats have recorded +8 tackles inside 50 or better they average a 65 point win. On the two occasions they’ve achieved -8 or worse they’ve lost by 18 and 41. They also don’t mind getting on the positive end of some xScore variance. Points from stoppage aren’t as big a predictor for them as others.
Gold Coast are towards the bottom of the league for post-clearance ground ball, but they’re 8-1 when they’ve won the stat. They boast the same record when winning crumbing possessions, but are dead average in the stat across the season.
GWS have only won points from centre bounce in 6 games this season, but they’re 6-0 with an average margin of +45 when doing so. They’re not as dependent as most teams on building an xScore advantage to win, because they outperform their opponents on xScore rating by a maddening 10+ points per game.
I’ll be back next week to step through the remaining nine teams as well as hopefully looking at which teams do or don’t have their performance captured well by Rating Points.
