Stats made simple for coaches

One number is all you really need to follow the rest of this series. The rest is here for the curious — no propellor hat required.

---

You don't need to understand r-values, p-values or confidence intervals to use Powercoach. The app handles that for you.

To follow the rest of this series, you only need to learn one — the r-value. That's it. The rest of this piece is optional.

The one to know: the r-value

Coaches already think in links between things: contested ball and winning, pressure and turnovers, entries and goals. The r-value puts a number on the strength of a link — on a 0-to-1 scale, where bigger means the two things move more tightly together.

• 0 — no link at all. The stat tells you nothing about who wins.
• 1 — a perfect link. Every time one goes up, so does the other.

Real footy stats sit in between, on a simple scale Powercoach also uses:

So pressure and turnovers moving together at 0.81 is "very strong" — nearly joined at the hip. Clearances and winning at 0.32 is "moderate" — a real but loose link. Don't memorise the bands; just know bigger means tighter.

(One wrinkle: r can also be negative, down to −1, meaning more of that stat goes with losing. That usually has a tactical story — like piling on repeat inside 50s because you can't convert, so the ball keeps coming back.)

That's the main one. Whenever you see an r-value in one of these articles, you now know what it means. Honestly, that's enough.

You're set. Take it to What actually predicts a winning quarter and you'll read it like a coach with a stats degree. Or stick around if you want to know what the other letters mean — no exam at the end.

---

For those who want more

Still here? Welcome. Here are the other numbers Powercoach shows you, roughly in the order they come up.

r² — how much it explains

Square the r-value and you get r², which answers a slightly different question: how much of the result does this one stat explain? An r of 0.7 gives an r² of about 0.49 — so that single stat accounts for nearly half of what separates winning from losing. It's a reality check on r: even a strong-sounding link rarely explains everything, because footy has a lot of moving parts.

An example makes it concrete. Chain efficiency runs at an r around 0.7 on its own — an r² near 0.5 — so it explains about half of what separates winners from losers. That's a dominant factor. Most stats aren't like that: they each explain a smaller slice, and the real picture is built from several of them together.

Here's the catch worth knowing: you can't just add r² values up. Clearances and centre clearances both link to winning, but they overlap heavily — two windows onto the same contested-ball strength, not two separate explanations — so stacking their r² would count the same part of the game twice. A team's true winning formula is usually a handful of overlapping factors, which is why no single number explains everything. (A stat's own r² doesn't shrink just because another stat is also strong — each stat's r² is its own; the overlap only matters when you try to combine them. More on this in There's no universal formula.)

Confidence interval — the range we're sure about

No number from a sample of games is exact. The confidence interval is the range the true value almost certainly sits inside — the honest version of a single figure.

The dot is the best estimate; the bar is the range. Narrow bar = confident. Wide bar = we need more games. The trick: if the bar crosses zero, we can't rule out "no link at all," so Powercoach marks it not yet significant and you shouldn't lean on it. The more games captured across the platform, the narrower these bars get — it's a collective effort.

This is also why some of Powercoach's insights stay locked until you've recorded enough quarters. It's not the app being coy — until the bands tighten, the numbers would be more noise than signal, and a misleading insight is worse than none. Around ten games' worth of quarters is usually enough for your team's picture to firm up. The practical takeaway: capture as many games as you can, as early as you can. Every quarter you record narrows the bands, so by the time finals arrive — when the margins matter most — your team's numbers are at their sharpest, instead of you flying half-blind on a thin sample exactly when it counts.

p — the fluke check

The p-value answers one question: could this just be random luck? It's the chance you'd see a result this strong even if there were really nothing going on. Lower is better.

Picture a coin. One head, nobody blinks; two, still normal. But if a mate tosses five heads in a row, you start eyeing the coin — that only happens about 3% of the time with a fair one. The p-value is exactly that gut feeling put on a scale: the lower it gets, the louder the result says "this isn't chance."

• Below 0.05 — less than a 1-in-20 chance it's noise — is the usual bar for "worth trusting." Five heads in a row (about 3%) sits just under that line; four in a row (about 6%) doesn't quite make it.
• The smaller it goes, the surer you are. When we found that footy goals genuinely cluster into runs, the p-value put the chance of a fluke at less than 1 in 1,000 — the equivalent of tossing about ten heads in a row. At that point you're not blaming luck; you're sure the coin is loaded.

It's the difference between "we won three in a row, might be luck" and "we've done this so consistently it can't be a coincidence."

z — how surprising it is

One quick thing first: z-scores measure distance in units of standard deviation — basically, how much a number usually bounces around its average. Two midfielders both averaging 20 disposals can be totally different players if one's a steady 18–22 and the other swings between 6 and 34. Standard deviation is the number that captures that bounce; z-scores use it to judge how unusual any one result is.

A z-score says how far a result sits from "normal," measured in those standard-deviation units.

• z near 0 — totally ordinary, right around what you'd expect.
• z around 2 — unusual; the kind of thing that happens maybe 1 game in 20.
• z of 3 or more — genuinely rare; if nothing special were going on, you'd almost never see it.

That's exactly how we tested momentum. If goals fell in a random order, you'd expect a certain number of scoring "runs." We saw far fewer — goals clumped together — and that result landed at z ≈ −3.4, way out in the tail of the chart above. Too far from normal to be luck, so the run-on is real. A big z (positive or negative) is just the number's way of saying "this would be astonishing if nothing were going on."

Cheat sheet: one word per letter

The letters are just statisticians' shorthand (p genuinely stands for probability; r and z are conventions). Here's a one-word hook for each:

Letter	One word	What it really tells you
r	Strength	How tightly two things move together (0 to 1).
r²	Explanation	How much of the result that one stat accounts for.
p	Probability	The chance it's a fluke — small p, big confidence.
z	Surprise	How far from normal a result is — big z, very surprising.

Here's the trick that makes them stick: the letter hides inside its own word — the r in st*rength, the **p* in probability, and the z you'd get spelling surpri*ze the American way. Three letters, three jobs. (r² is the odd one out — just think of it as r's bigger brother: *how much it explains.)

A few honest traps

• Correlation isn't proof of cause. These stats track with winning; they don't guarantee it. Winning more clearances doesn't make you win — teams that win clearances just tend to win. Usually there's a real mechanism, but treat the number as a strong hint, not a law. (The series' closing piece, Correlation is not causation, is the full version of this trap, with a stat that proves it.)
• Some stats overlap. Clearances and centre clearances measure nearly the same thing — not independent clues, more the same clue wearing different hats.
• Small samples lie. Early on, odd results — including negative ones — are often noise. Give it games before you rebuild your game plan around a surprise.

How this pays off — handy, not homework

To say it once more: none of this is required to use Powercoach — the app does the interpretation for you. But if you've read this far, you've got the vocabulary to read every other article in this series with confidence, and one bonus: it makes the AI Brain Chat sharper too. Ask it more pointed questions — "is our forward-half pressure significantly linked to winning, or is the sample too small?" — and read its answers with judgement — "moderate correlation, limited games: promising, not proven." You go from passenger to driver.

Glossary — with an AFL example for each

Term	In plain English	AFL example
Mean (average)	Add them up, divide by how many	A team averaging 11 goals a game
Median	The middle value when you line them up	Median weekly score — unmoved by one 80-point blowout
Mode	The most common value	The margin or quarter-score you see most often
Standard deviation	How much the numbers bounce around the average	Two 20-disposal mids: one steady (15–25), one wild (6–34)
r (correlation)	How tightly two things move together, −1 to +1	Pressure and turnovers move together at 0.81
r² (variance explained)	The share of the result one stat explains	Chain efficiency (r≈0.7) explains roughly half of wins vs losses
Confidence interval	The range the true number probably sits in	"Somewhere between 0.40 and 0.60" — narrows with more games
p-value	The chance the result is a fluke (small = trust it)	Goal-runs clustering: p < 0.001, almost certainly real
z-score	How many standard deviations from normal (big = surprising)	Momentum runs landed at z ≈ −3.4 — far too unusual to be luck
Significant	Solid enough to act on (usually p < 0.05)	A stat whose confidence interval doesn't cross zero

---

A note for the curious

These definitions mirror the ones Powercoach uses in your team's Stats Analysis, on purpose — so the words on the screen and the words here mean the same thing. The footy examples come from real captured games across local competitions. Want to see the ideas doing actual work? Every data piece in this series — from what predicts a winning quarter to whether momentum is real — runs on exactly these numbers.