Process vs proxy: what predicts winning

We ranked the stats coaches shout about by how well they really predict winning a quarter. The clearance count comes last.

New to r-values, p-values and confidence intervals? Stats made simple for coaches explains every number in this piece in plain English — no propellor hat required.

Walk past any huddle at three-quarter time and you'll hear it: "We have to win the clearances." Centre bounce, around the ground — first hands on the footy is treated as the master key to the quarter. It feels true. The team that keeps winning the tap and the crumb looks like the team in control.

So we put it to the test. For every quarter of captured AFL football, we lined up the stats coaches care about and asked one thing of each — how reliably does winning that stat-battle predict winning the quarter? — on a scale from 0 (no link) to 1 (perfect link). (Where it helps, we'll quote the more intuitive companion number too — "win the battle and you win the quarter X% of the time" — alongside the score.)

The clearance count came last.

There are two stories in that chart. The amber bars near the top are the ones we should be a little suspicious of, and the blue bars are where the real coaching lives.

The "obvious" stats are obvious for a reason

Inside 50s, inside 25s, scores per inside 50, intercepts that turn into scores — these sit near the top, and of course they do. They're a half-step from the scoreboard. Saying "the team with more inside 50s tends to win" is barely more useful than "the team that scores more tends to win." They're real, but they don't tell a coach much they can act on. They're amber for a reason: they describe the result more than the method.

The blue bars are the process — the things you actually drill, set up and demand. That's where it gets interesting.

Finding #1: the clearance count is the weakest lever on the ground

Here's the uncomfortable one. Winning the centre-clearance count is the single weakest predictor of winning the quarter of every stat we measured (r = 0.24). Total clearances barely do better (r = 0.32). Yes, the team that wins the centre clearances still wins the quarter more often than not — about 62% of the time — but that's the smallest edge any battle on the ground gives you. Compare it to winning the chain-efficiency battle, which carries an 89% quarter-win rate.

Think about why. A clearance is first possession, not clean possession. Tap it forward into a pack, kick it to a contest, turn it over inside 30 metres — you "won the clearance" and gained nothing. The count rewards getting a hand to it; it says nothing about what happened next. And what happens next is the whole game.

This doesn't mean clearances are worthless — they're the start of everything. It means the count is a poor scoreboard. A team can lose the clearance count and win comfortably, and it happens all the time.

(One nuance worth knowing, before anyone reads this as "the contested ball doesn't matter": the contested quality of a stoppage win — was it ripped out under heavy pressure, or a free walk-up? — isn't lost in this picture. It's captured in the pressure rating, which tags every clearance by the pressure on the winner. Part of why pressure (r = 0.50) out-predicts the bare clearance count (0.32) is that it weighs how hard each win was, not just counts them. Powercoach doesn't track contested possessions as a separate stat — but it isn't ignoring contest, either; it's folding it into pressure.)

Finding #2: it's not how often you win the ball, it's what you do with it

The strongest genuine process stat — comfortably clear of the clearance counts — is chain efficiency: the share of your possession chains that end in a score (r = 0.71; win that battle and you win the quarter 89% of the time).

That's the whole game in one number. Not how many times you won the ball. Not how many entries you had. How often a possession you owned turned into a shot. Everything coaches can influence between the contest and the goal square — clean disposal, holding the ball longer, using the right option, retaining under pressure — shows up here.

Below it, the levers that matter are also process, also coachable: pressure on the ball (r = 0.50), winning it back through intercepts (0.43), and doing that high up the ground with forward-half intercepts (0.38). These beat the clearance counts — sometimes comfortably.

(One honest caveat: chain efficiency is itself close to the scoreboard — it has scoring baked into it. The purest "how you play" levers are pressure, intercepts and possession. The point stands either way: every one of them out-predicts the clearance count.)

What this means for your team

• Stop scoreboarding the clearance count at the breaks. It's the least informative number on your stat sheet. A team can lose clearances and win going away. Use it to diagnose your stoppage craft, not as a measure of who's on top.
• Coach the five seconds after the clearance, not the clearance itself. Winning first possession only matters if the next disposal keeps the chain alive. Chain efficiency — score per possession — is the number that actually moves with winning.
• The levers that travel are pressure and turnovers. Applying pressure, winning the ball back, and winning it back in your forward half all out-predict raw clearance counts. They're also things you can structure and demand every week.
• Treat inside-50 counts as outcomes, not targets. They rise when the rest is working. Chasing the entry count directly can mean bombing it long into a pack — entries up, efficiency down, scoreboard flat.

In short

• The centre-clearance count is the weakest predictor of winning a quarter we measured, and total clearances aren't far behind. First possession isn't clean possession.
• Chain efficiency — scores per possession chain — is the strongest genuine process stat. It's not how often you win the ball, it's what you do with it.
• The obvious entry stats predict winning because they're almost the scoreboard already. The coachable edges are pressure, turnovers, and what you do with each chain.

For the wider lesson underneath this analysis — why some stats correlate with winning yet tell you nothing useful — see the series' closing piece: Correlation is not causation.

A note on the data

This analysis pools every quarter of captured AFL football across local Melbourne competitions, after filtering out demo data, unfinished matches and drawn quarters. For each quarter we take each team's advantage over its opponent on a stat (e.g. "won the clearance count by 4") and correlate it with whether that team won the quarter, giving the r values plotted above. "Win that battle and you win the quarter X% of the time" is the win rate among quarters where a team out-rated its opponent on that stat.