The Arzani File and the Raw-Data Test for the A-League Transfer Window

The Arzani File and the Raw-Data Test for the A-League Transfer Window

Over the final twelve rounds of the 2026-17 A-League season, an eighteen-year-old at Melbourne City averaged 4.6 completed dribbles per match. The league average that season was 2.2. He had never appeared on a single "young talent to watch" list, had never given a national television interview, and in the press room after the Melbourne derby, not one reporter asked about the number.

I found it on an April evening while working through the GPS dataset Melbourne City had sent the newsroom under our data-sharing agreement. Twelve pages, split by round, split by pitch zone. In the column headed "dribbles completed in the final third," the name Daniel Arzani appeared forty-one times. The player ranked second had nineteen.

Fifteen months later, in his competitive debut for Celtic, Arzani tore his anterior cruciate ligament. A contract decided largely on eight minutes of highlights and one A-League season.

The transfer window is where data gets distorted more than anywhere else in football's ecosystem. That is why the Arzani file still sits in the first drawer of my cabinet.

Context: the market speaks in contracts, I read in movement data

I entered the profession in 2026 at Sports Illustrated as a fact-checker. The first lesson a checker learns is not to trust the press release. Twenty-seven years later the principle is unchanged; only the tools have. Instead of calling to verify a quotation, I download the movement file and rebuild the match from raw numbers.

The A-League has three features that make it a perfect laboratory. First, the salary cap and designated-player mechanism force clubs to split budgets into two tiers, so every major signing is a measurable gamble. Second, the season runs only twenty-seven rounds plus finals, meaning the minutes sample is tiny and every metric is amplified by a small denominator. Third, the loan pipeline between Australia and Europe creates a constant flow of players, and every move fractures the data record.

In the first three weeks of the current window I received seventeen player announcements from various sources. Fourteen contained at least one line describing the player as a "leading attacker in the league." Not one included cumulative load data. Not one included actual minutes over the past two seasons. Not one included accelerations above 25 km/h per ninety minutes.

That is the starting point of every mistake in this market.

The evidence chain: from forty-one dribbles to a torn ligament

When I called Melbourne City's coaching staff to request Arzani's full movement data across twelve rounds, they hesitated. I told them plainly: send the file, or I write that the club is withholding the physical profile of an eighteen-year-old from the market. Two days later the file was in my inbox.

Four columns stood out. Completed dribbles per ninety: 4.6, double the league average. Accelerations above 25 km/h per ninety: 18.4, among the league's highest for a wide player. Actions on the ball in the final third: 41 across twelve rounds. And xT per ninety in the left channel: 0.31, triple the positional average.

What I did not have in April 2026 was a fifth column: tolerance for load in a high-collision environment. The A-League gave Arzani wide running lanes, controlled minute progression, and long gaps between rounds. Europe gave him a game every three days, flank collisions with defenders fifteen kilos heavier, and a calendar with no room to learn.

When the world zooms in on the goal, I rewind thirty seconds and zoom in on the off-ball run. But I also have to admit: some off-ball runs no GPS dataset measures, and that gap is exactly what swallowed a nineteen-year-old's career.

A small discovery in the 2026 A-League sounded like a whisper, but three years later it became a roar at the World Cup. On the strength of the Arzani series I was sent to Russia for the 2026 World Cup. There, while the press room wrote about Luka Modrić's technique, I spent four days calculating Croatia's PPDA across their opening group matches. Before the Argentina game, Croatia's PPDA was 7.9, meaning they allowed fewer than eight passes before contesting.

PPDA does not decode Croatia. It decodes the football Croatia hides inside its patient shell. Their deep midfield block was not defending; it was a trap set beat by beat, and PPDA only measured the distance between trap and prey. Weeks after the final, UEFA's analysis unit confirmed the data series I had published.

By 2026, when the pandemic suspended the A-League, I lost all pitch access. While colleagues pivoted to social commentary, I launched a project collecting data from thirty-seven rescheduled matches played in empty stadiums. The result: home win rate fell from 49.2 per cent to 41.3 per cent. I published the conclusion that crowds are a data variable, not an emotional one. Melbourne Victory cut off contact for six weeks. Football Australia's communications director called to offer me an unpaid data consultancy. I took it immediately — it was a foothold.

The empty stadiums of 2026 did not make players weaker. They exposed the artificial metrics that crowds had been shielding. The same logic applies to the transfer market: strip away the noise of rumour and what remains is the real quality of a signing.

Summer 2026 took me to another column. I partnered with a researcher at Victoria University to build a match-load tracking system, and Pedri was the perfect subject: fifty-one matches by the end of the European Championship. His average distance covered at the Euros was 11.2 km per match. At the Tokyo Olympics, weeks later, it fell to 9.4 km. A drop of 1.8 km in the same player in the same summer is not a form issue. It is exhaustion.

I do not need to see how many matches they play. I need to see how many metres they run in a situation nobody notices. That is why every subsequent piece of mine carries a public raw-data section with a download link, and why I refuse qualitative interview work without at least one quantitative metric.

Applying those four files to the current window, I built a five-column filter for every proposed signing. Column one: minutes trajectory across the last three seasons, not the best season. Column two: accelerations above 25 km/h per ninety, benchmarked against the destination league's positional average. Column three: xT by pitch zone, isolated from goals and assists. Column four: contract structure, including release clause, remaining term, and wage share of the total payroll. Column five: twelve-month cumulative load, the sum of minutes played plus high-intensity distance.

A player can top column three and sit last in column five. The market only asks about column three.

The counter-intuitive angle: correlation is not causation, and I am the first who must remember it

When the empty-stadium data produced a 7.9 percentage-point drop in home win rate, the most seductive conclusion was that crowds create home advantage. But those thirty-seven matches were not played under identical conditions. They were compressed into congested schedules, many on neutral grounds, many requiring long travel under quarantine rules, and several teams entered that stretch after a three-month shutdown. Part of the 7.9 points came from the crowd. Part came from the calendar.

I published that caveat alongside the finding, because a data journalist is not permitted to hide his own noise. Before publishing anything, I run a reverse test: find a metric that could overturn the conclusion I have just written. If I cannot find one, I state the limitation plainly to the reader.

The same test applies to xG. xG has been abused to the point of explaining things it was never built to explain: referee decisions, a player's mental form, transfer value. A team can post 2.4 xG and lose 1-0, and people call it statistical injustice. Data never lies — but it took me ten years to learn when it tells half the truth. The half-truth is this: xG accumulates shot by shot, and the shot is the final output of a sequence that never gets recorded.

PPDA has the same limit. It counts passes the opponent completes before being contested; it does not measure the quality of those passes. A team that lets opponents pass sideways across the back line will post a low PPDA without applying real pressure. I found this while re-checking Croatia's own numbers: two of their four knockout matches had PPDA below 8, yet their final-third duels were half their group-stage volume. In some matches that team simply stood and waited, and PPDA still looked handsome.

Applied to the transfer window, the largest limitation of any data filter is the small denominator. A player with forty-one completed dribbles across twelve rounds may have only thirty minutes at a higher level. Arzani is living proof: his A-League movement data was real, and so was his ACL tear. The two truths do not exclude each other. A filter that reads only column three will always mis-buy column five.

I once watched a signing built on a high xT collapse within six months — not because the player was poor, but because the club bought a pitch-zone profile while its squad had nobody capable of creating those zones. Metrics measure the player. Squad structure decides whether the player has anywhere to run.

The final blind spot: agents sell highlight reels, not data

In the past three weeks I have taken seven calls from intermediary sources. Six opened with the phrase "this player is in the best form of his career." None opened with minutes played.

That is not a coincidence. An eight-minute highlight reel can be cut from three good matches in a twenty-seven-round season. Load data cannot be cut from three matches, because it accumulates minute by minute. Release-clause structure cannot be cut either, because it lives in the contract document, not on the grass.

When an A-League club pays designated-player wages for an import from Europe, it is not buying form. It is buying a minutes sequence over the next three years, a high-intensity distance sequence over the first twelve months, and a soft-tissue injury probability. Not one of the clubs I have worked with publishes that probability to its own supporters. That is why I still call to request the raw file, and why I still accept losing a source every time I publish what they did not want published.

In 2026, my rigidity cost me a source. I accepted it. The one principle I have kept across twenty-seven years is never to cite a metric whose longitudinal data chain I have not traced myself.

The signal for the next cycle

Two moments matter in the next three months. The first is transfer deadline week, when clubs must close deals before the window shuts and medical files become the least-read documents in the building. The second is round six of the season, when the minutes sample is large enough for load data to separate from technical data.

I will track one specific thing: the gap between each new signing's high-intensity distance in round three and in round six. A decline under 10 per cent is normal adaptation. A decline above 20 per cent signals overload, and it appears roughly four to six weeks before the first soft-tissue injury.

If this season repeats 2026, we will again see an eighteen-year-old sold on one season, and again see an ACL rupture on debut. The only thing that can break the loop is load data entering the transfer file as a mandatory, publicly visible column for both negotiating parties, in the same format the league already uses to administer its salary cap.

Over the final twelve rounds of 2026-17, forty-one completed dribbles told me an eighteen-year-old was good enough for Europe. No dataset told me his tendons were. Nine years on, the question is still on the table: will a football economy publish column five, or keep buying eight minutes of highlights?