How to Read Football Stats: xG & PPDA Predictors

How to Read Football Stats: xG, PPDA and What Predicts Results

Most stats sites hand you 40 numbers per match. Maybe four of them tell you anything about the next one.

If you've gone looking for how to read football stats, xG and PPDA are where most people start - and where most people stop, because a definition on its own doesn't win anything. The bettor's question is narrower: which of these numbers repeat, which are noise dressed up as insight, and how do you use either before kickoff?

The short answer, up front:

• xG difference over 10+ games is the most useful team signal. Not one match. A run.
• Shot quality beats shot volume. 15 shots worth 0.06 xG each is a quiet night.
• PPDA measures pressing intent, not defensive quality - and it moves with the score.
• Possession %, pass accuracy and corners predict very little. They describe style.
• Everything here is a probability, never a guarantee.

Why Most Football Stats Don't Predict Anything

A stat only earns its place if two things are true: it repeats over time, and it connects to goals. Most popular numbers fail both tests.

Take possession percentage. It's partly a byproduct of the scoreboard - the team that's behind chases the ball. So a team can dominate the ball while losing, and pile up possession precisely because it's losing. That number describes what happened, not what's likely to happen next.

Pass accuracy is worse. A side sitting deep and knocking it sideways between centre-backs can post 90% accuracy while creating nothing. Accuracy measures risk appetite, not quality.

Corners are the classic trap. A corner is a low-value chance - most models price it around 0.02–0.03 xG, since the majority never clear the first defender. Twelve corners can mean twelve wasted deliveries. Over a full season, corner counts and league position have almost no relationship.

Most football stats explained in glossaries are descriptive. Very few are predictive. Everything below is filtered through that one question: does it repeat?

xG Explained: What Expected Goals Actually Measures

Ask what is xG and you'll get a clean one-liner: it assigns every shot a value between 0 and 1, representing the estimated probability that the shot becomes a goal. A 0.30 xG chance goes in roughly three times out of ten over many repeats. Penalties sit near 0.76–0.79 depending on the model.

That's expected goals explained in the simplest possible terms. The useful part is what happens when you add the shots up.

How a Shot Gets Its Value

A model looks at a handful of things, and it mostly cares about four:

1. Distance and angle from goal. A central shot from six yards is a different species to a shot from 25 yards.
2. Body part. Feet beat heads. A header from a cross usually lands between 0.05 and 0.10; the same position with a foot is worth far more.
3. How the chance was created. A cut-back across the six-yard box is priced far higher than a hopeful cross from wide.
4. What's in the way. Defenders between ball and goal, whether the keeper is set, whether it's a one-on-one, whether it's a rebound.

A worked total makes it concrete. Six shots:

That's 0.30 xG per shot, which is excellent. Now compare a side with 14 shots: twelve long-range attempts at 0.03 and two from 18 yards at 0.08. Total: 0.52 xG, or 0.04 per shot. One team looked busy. The other created something.

One warning: Opta, StatsBomb and Understat produce slightly different numbers for the same match. Pick one provider and stay with it - mixing models ruins every comparison you make.

xG Over One Match vs xG Over Ten

Single-match xG is noisy. A penalty, a red card or one deflection can swing the whole figure, and a 1.9 xG performance can come from four chances rather than a dominant display.

Over ten matches, patterns appear. Around 15–20 matches you have a working minimum for cautious team comparisons, and over a full season xG is one of the more stable signals in the sport.

The figure to actually track is xG difference: xG created minus xG conceded, written per game. A team running +0.8 per match over ten games is genuinely better than its results suggest. A team running −0.6 is living on borrowed points.

PPDA Explained: Measuring Pressing Intensity

PPDA stands for passes allowed per defensive action. It answers a simple question: how many passes does the opposition complete before you do something about it - a tackle, interception, challenge or foul - inside a defined zone?

Lower is more aggressive. Typical bands:

• 7–9: a high, relentless press.
• 10–12: a mid-block, pressing in spells.
• 15–18: a deep block, inviting pressure.

The critical point, and one the top explainers make clearly, is that PPDA measures intent, not outcome. A team can press beautifully at PPDA 7.5 and still concede three because the press got played through. A passive side at PPDA 16 can keep a clean sheet behind a packed box. Pressing intensity tells you how a team wants to defend, not how well it defends.

Search ppda football and you'll find that definition everywhere. Using it is the harder half - you have to ask which game state you're looking at.

Other Stats Worth Knowing

Four more names show up constantly on stats sites. Short version of each:

• Field tilt - a team's share of attacking-third touches or passes. 50% is even, 62% is heavy territory. It's a territory proxy, not a quality measure: you can hold the final third and still create nothing.
• Progressive passes - passes that move the ball a meaningful distance toward goal, usually quoted per 90. High volume suggests a team builds through the lines rather than launching it.
• Big chances - Opta's label for a clear sight of goal, such as a one-on-one or a close-range finish. Each is generally worth around 0.3 xG or more, so two big chances beat eight speculative shots.
• Per 90 - any counting stat divided by minutes played, then multiplied by 90. Raw totals flatter players who start every week; per-90 figures let you compare a starter with a substitute fairly.

Which Stats Actually Predict Results

Ranked honestly, best to worst:

1. xG difference over a run of games. The single strongest team-level signal. It separates teams whose results match their performances from teams riding finishing luck or a hot keeper. Goals themselves are surprisingly noisy over 10 games; xG settles faster.

2. Shot quality. xG per shot, big chances created, shots taken from inside the box. Volume without quality is the oldest false signal in football - 20 shots from range is not pressure.

3. PPDA - but only in context. As a style marker it's genuinely useful: it tells you whether a match is likely to be stretched or cagey. As a standalone predictor of points, it's weak, because game state rewrites it.

And what predicts poorly: possession percentage, pass accuracy, corners, and raw shot counts on their own. Shots on target is a small step up from total shots, but still well behind xG. xG difference answers the question these numbers pretend to answer.

Worked Examples: Reading a Stats Line

1. High possession, low xG

Team A: 68% possession, 614 passes, 89% accuracy, 15 shots, 12 corners - 0.9 xG, or 0.06 per shot. Team B: 32% possession, 5 shots - 1.8 xG. Final score 2–1 to B.

A owned the ball and the corner count. B owned the dangerous areas. Anyone backing A on the strength of 15 shots and 12 corners was reading style, not signal. If A get priced short at home next week on the same numbers, the value can sit with the other side.

2. Solid xG against, poor keeper

Over 10 matches, Team C allow 0.95 xG per game - top-quartile chance prevention. They concede 1.6 goals per game. Their keeper saves 58% of shots on target; the league average sits near 70%.

The defence is doing its job. The last line is leaking. Because xG against measures chances allowed and not goals conceded, expected-goals tables rate C highly while the real table doesn't. That gap is where overreactions and slow market adjustments tend to show up - a spot to examine, not a sure thing.

3. A pressing side whose PPDA rises when chasing

Team D average 8.4 PPDA across the season. Split by game state: 7.1 while level or ahead, 13.6 after the 70th minute when trailing. Their field tilt also falls from 58% to 44%.

Their press is a function of the scoreboard and their legs. Intent stays high when they're behind, but fatigue and stretched shape push the number up - opponents play through them instead of around them. So the season average tells you nothing about minute 75. Before kickoff, expect the aggressive version and a high line. In-play, expect chaos at both ends.

The Context Traps

Five things break otherwise-good stats work:

• Score effects. Trailing teams shoot more and hold more of the ball; leading teams sit deeper. Possession and shot counts are partly functions of the scoreline.
• Game state. Filter to "while level" wherever your tool allows. It's the cleanest slice of pressing and territory data.
• Opposition strength. 1.8 xG against a bottom-three side is not 1.8 xG against the leaders. Without adjustment, you're comparing different jobs.
• Small samples. Three-match hot streaks are noise. One red card or one early penalty can dominate an entire match's numbers.
• Home and away splits. Some teams show 20-point gaps in home vs away xG difference. A season average hides exactly the thing you need.

How to Build a Simple Stats Routine

Twenty minutes before kickoff, in five steps:

1. Pick One Provider and Commit

Opta-based sites, Understat, FBref—choose one data model so every comparison is like-for-like.

2. Write Down the 10-Match xG Difference per Game for Both Teams

That number, not the league table, is your starting point.

3. Slice by Game State Where You Can

Check xG and possession while the game was level, not the blended full-match figure.

4. Check the Pressing Pair

PPDA and field tilt together tell you whether a match should be open or compressed.

High field tilt with high PPDA indicates a slow, controlled game, while high field tilt with low PPDA indicates chaos.

5. Adjust for Opposition and Venue, Then Compare With the Price

Log the line you saw and the read you made. Judge your process after 20 bets, not one.