Australia finished its 2026 World Cup qualifying campaign with 38 goals scored against an expected goals total of 28.5. That is a gap of roughly ten goals — about 35% above what the underlying chance quality predicted. In the European football analytics community, a number like that is almost always read as a regression-to-the-mean signal: a team has been finishing above their model output and will, sooner rather than later, score fewer than the model predicts as the variance evens out.
That reading is probably right for most teams. For Australia it is also probably incomplete. Below is why. The Socceroos under Tony Popovic — who replaced Graham Arnold in September of 2024 — have built a tactical structure that systematically undervalues itself in publicly-available xG models. The +10 gap is not entirely noise. It is partly the model failing to measure something Australia is actually good at.
Three structural reasons explain the gap, where it would regress at the World Cup anyway, and a short framework for reading Australia’s Group D matches against Türkiye, the United States and Paraguay without falling into either the “they will fall back to earth” trap or the home-fan corollary that has shown up across the Australian sports press for six months.
What the +10 xG gap actually measures
Expected goals as a public metric is built on shot-location data. The model assigns each shot a probability of being converted based on distance, angle, body part, defensive pressure and a few other inputs. The total xG for a team across a season is the sum of those probabilities. When a team finishes 35% above that sum, the standard reading is that they are getting lucky in front of goal — and across enough matches, the luck reverts.
The Australian qualifying gap is large enough to be statistically meaningful (10 goals across 16 matches sits well outside the noise band for a sample of that size). The honest question is what produced it. Three possibilities, and they are not mutually exclusive.
The first possibility is straight-line luck. The Socceroos’ finishers — Mitchell Duke, Awer Mabil, the emerging Nestory Irankunda — produced finishing percentages above their career baselines across qualifying. Some of that almost certainly is variance and will not repeat.
The second is shot selection. Australia took fewer shots than most CONCACAF and AFC qualifying sides (about 11.2 per match) but converted them at higher rates. The model penalizes them for shot volume; the underlying shot quality, by the model’s own outputs, was actually mid-pack. Where the gap appears is in the conversion rate per shot, not the chance creation.
The third is tactical structure. Popovic has built a system around set-piece variation and counter-attacking transitions that the public xG model does not perfectly capture. Set-piece xG models are notoriously weak relative to open-play xG models, and Australia scored a disproportionately high share of its qualifying goals from set pieces and counter-attacks — the two phases where public models systematically under-measure threat. Our piece on set-piece analytics covers the structural reasons this gap exists in public data.
What will regress at the World Cup, and what probably will not
| Component | 2025 qualifying | WC 2026 projection | Why |
|---|---|---|---|
| Shot conversion % | 14.1% | ~11% | Variance correction against better keepers |
| Set-piece xG efficiency | 1.41 actual vs xG | ~1.20 | Partial regression but structural advantage holds |
| Counter-attack conversion | 22% of goals | ~18% | Better opposition defensive shapes |
| Defensive xG against | 0.8/90 | ~1.1/90 | Tougher attacks; defense still strong |
| Group D xPts | — | 3.5-4.0 | Realistic path to R16 |
The version I would bet on, having tracked Popovic since the Western Sydney Wanderers years, is a Socceroos team that probably scores fewer goals per match than they did in qualifying, but not by as much as the simple regression argument suggests. The defensive structure travels well to the World Cup; the goalkeeping under Mat Ryan is the most reliable in the group. The Group D opener against Türkiye on June 14 is the most-leveraged single match, and the model-implied expected points across the group sits in the 3.5-4.0 band — enough for a realistic Round of 16 path, especially given the U.S. defensive concerns I covered in the earlier USMNT piece.
The Tony Popovic tactical identity
Popovic’s Socceroos play a 4-2-3-1 with a deliberate emphasis on defensive compactness and direct transitions. The contrast with the Graham Arnold era is structural, not cosmetic. Arnold’s teams tried to control matches through possession; Popovic’s accept that Australia is rarely going to dominate possession against tournament-tier opposition and build instead around denying space to the opponent and converting the chances the opponent gives back.
The shape, in numbers: Australia averages roughly 42% possession across competitive matches under Popovic, down from 51% under Arnold. PPDA estimate is around 11.5 — mid-block territory, not high press. The recovery time after losing the ball is slow by tournament standards, which Popovic has accepted as the cost of the defensive structure he wants.
The strength of the system is its set-piece organization. Australia has run twelve identifiable set-piece routines across qualifying, with a goal conversion rate of roughly 22% on dead-ball situations in the opposing third. That is elite. For reference, the average qualifying side converts about 12% of dead-ball situations in the opposing third. The set-piece work is where the +10 xG gap is least likely to regress, because it is structural rather than variance-driven.
Where this gets weird
Three complications keep the +10 number from settling cleanly into either the “lucky” or “underrated” boxes that public soccer analytics wants to put it in.
First, the public xG models do not separately publish set-piece xG numbers for international tournaments in real time. Australia’s tactical advantage will be invisible in the live xG graphics during World Cup broadcasts. Viewers reading “Australia 1-0 USA, xG: 0.9-1.4” will see a number that systematically understates how the goal was created. The model is not wrong; it is just optimized for open-play prediction.
Second, the +10 qualifying gap is concentrated in the AFC qualifying round, where opponent quality was relatively low. The 6.5-goal slice of that gap that came in matches against Top-50 FIFA opposition is much smaller and within typical noise bands. The aggregate gap is partly an artifact of the opponent distribution.
Third, Popovic’s tactical identity has not yet been tested in a tournament knockout context. International tournaments compress the tactical clock — substitutions matter more, fitness gaps widen across the third match, and second-half adjustments by elite opponents are harder to absorb. A mid-block system that worked in qualifying against AFC sides may struggle if Australia advances and faces a side that can pin them deep for 90 minutes.
What to watch in the Australia matches
- Set-piece organization in the opener against Türkiye. The clearest read on whether the structural xG advantage travels. Track whether Australia is winning the first ball or settling for second-phase shots.
- Defensive compactness in the second half. Australia’s mid-block has held through 70 minutes consistently in qualifying. The minute-by-minute defensive xG against in the 70-90 window is the question the World Cup will answer.
- Irankunda’s role. The 19-year-old winger is the most volatile single input on the Australian roster. If Popovic uses him as a 70th-minute substitution against a tiring opponent, the high-variance offensive ceiling expands. If he starts and loses confidence, the bench becomes the more interesting story.
- Mat Ryan’s distribution. Australia’s build-up depends on the goalkeeper. Ryan’s distribution choices in the first 20 minutes are the cleanest indicator of the game plan against each opponent.
The callback
The +10 xG gap was a regression alarm for most of the analytics community. It is partly that. It is also partly the model missing what a well-coached mid-block team that excels at set pieces does to a framework designed to measure something else. Australia will score fewer goals at the World Cup than their qualifying conversion rate suggests. They will probably score more than the naive regression argument predicts. The gap between those two numbers is where the +10 is hiding its honest accounting. The expected goals explainer covers the methodology. Australia is not a dark horse. They are a team that has been quietly outperforming public models for two seasons and is about to play the highest-leverage version of the same game in front of the country that owns most of the public xG infrastructure. The model is not built to flatter them. Group D will be the answer that the model can’t give in advance.
Qualifying xG via FBref and StatsBomb; set-piece data aggregated from public broadcast tracking.
