Three Data Providers, Three Different xG Numbers for the Same Match

Every time I rewatch a match, I open three windows at once: an Opta event-data panel, a StatsBomb panel, and a sheet I build myself from player-tracking coordinates. They rarely agree on the xG figure for a single half. The gap is usually a few tenths — but that is enough to change the story: which team deserved to win, which team got lucky, which goalkeeper actually played well.

In June 2026, when the Bundesliga returned to empty stadiums, I re-ran my prediction model. Home advantage — a variable I had long priced at 0.45 goals per match — fell to 0.08 across nine matches without crowds. That number did not exist in any commercial dataset I could access. It came from cross-checking two sources, stripping out matches with early red cards, and verifying every fixture by hand. I waited three weeks before publishing. The admission I had to make was not that the model was wrong, but that I had ignored the crowd variable for years.

Numbers whisper. Anyone willing to listen will hear an entire match. But to hear it, you first have to know where a number was born, and who named it.

Expected goals is the probability that a shot becomes a goal, derived from event data. It sounds simple. To produce a single figure, a provider must decide a long series of things: which foot struck the ball, from what distance, at what angle, whether a defender blocked the line of sight, whether the assist was a cross or a through ball, and whether the phase came from open play or a set piece. Every choice pushes the probability in a different direction.

Opta released its xG model publicly in 2026. StatsBomb was founded in 2026 and took the opposite route: publishing its methodology, opening data to researchers, then selling the product to professional clubs. Wyscout was acquired by Hudl in 2026. Three systems, three datasets labelled by three different groups of people, three philosophies about what deserves to be counted. None of them is wrong. They simply answer slightly different questions.

Tracking data makes it messier still. A shot rated 0.12 by an event model may be rated 0.09 by a tracking model, purely because the tracking model can see the goalkeeper's position and two defenders that the event model cannot encode. A gap of 0.03 sounds trivial. Multiply it by thirty shots a match, then by three hundred and eighty matches a season, and it becomes a margin that can decide a title.

Before trusting a number, ask where it came from. That question is not an attack on data. It is the condition for data to mean anything.

Take one specific phase: a striker heads the ball from six yards after a cross from the right. Model A rates the header at 0.18 xG because it weighs the quality of the cross and the nearest defender. Model B rates it 0.27 because it only measures distance and angle and does not encode pressure. On television, only one of those figures appears, and rarely with the model's name attached. Viewers are handed a bare fact, when what they actually received is a convention.

The same happens with transfer fees. A transfer fee is a story; the data is the signature. Neymar moved from Barcelona to Paris Saint-Germain in August 2026 for 222 million euros — a figure that was officially announced and remains the world record. Yet in much Vietnamese coverage it becomes "about 200 million euros", or gets blended with add-ons into a completely different number. In January 2026, Enzo Fernández moved from Benfica to Chelsea for 106.8 million pounds. In August of the same year, Moisés Caicedo moved from Brighton to Chelsea for 115 million pounds, recorded at the time as a British transfer record. Those figures have contracts, sources and verification trails. Once they reach social media, the currency unit and the source are usually the first things to disappear.

VAR touches something more sensitive than money. I have long argued that millimetre offside lines are slowly killing the striker's instinct. FIFA first used semi-automated offside technology at the 2026 World Cup in Qatar. The Premier League adopted it from the 2026-25 season. The technology is faster, more consistent, and still rests on a chain of human choices: which frame counts as the moment of contact, which body part is measured, and how much tolerance is allowed. There is no "truth" in those choices. There are conventions. And conventions have to be stated out loud.

The referee is becoming the editor of the match — the person who decides which moments stay in and which get cut. When a goal is erased after four minutes, fans lose more than a goal. They lose the belief that a goal is a complete event.

In 2026 I began doing data analysis for The Football Sack, a then-new Australian football outlet. When the A-League reached round 12, I published a 3,200-word analysis of Melbourne City's pressing metrics. Using GPS data, I showed that Warren Joyce's side was pressing in the wrong direction: midfielder Luke Brattan covered 11.2 kilometres per match but produced only 1.3 successful tackles. Supporters mocked the piece for being too dry. Three weeks later Joyce changed the pressing structure, and Melbourne City won four matches in a row.

The lesson was not "data is always right". It was that data only carries weight when it is told as a story with a beginning, a conflict and a resolution. 11.2 kilometres means nothing on its own. Set beside 1.3 tackles, it becomes an indictment.

In 2026 I wrote an English-language piece predicting Croatia would reach the World Cup semi-finals, based on their xG. Luka Modrić was generating roughly 2.4 expected assists per group-stage match in my own calculations. A group of amateur coaches on Reddit called me a bookworm who did not understand football. Croatia reached the final. After the tournament, a journalist from The Athletic contacted me to ask how I calculated "defensive xG prevented" for defenders. I spent two weeks writing Python, cross-checking against StatsBomb data, and sent back a seventeen-page breakdown.

In 2026 they laughed at my xG. This year they ask me what xG is. That shift does not prove I was right. It only proves that a clearly defined measurement eventually finds its users.

What I want to say against the data crowd — now far larger than the sentiment crowd — is this: precision is not truth. A figure with four decimal places is not more trustworthy than one with two. It merely looks more trustworthy.

In this industry there is a temptation called filling the gap. When data is missing, people extrapolate. When a match has no tracking data, they substitute event data and call the two equivalent. When a player's sample is too small, they still draw conclusions. In June 2026 I turned down a commission to explain "football without crowds" because I needed three more weeks of data. The editor was not pleased. Had I written three weeks early, I would have had to invent a certainty I did not possess.

In serious analysis, when the input is empty, the only correct output is insufficient information. No player is born from a blank space. No match is built out of belief. Vietnamese sports media is learning fast how to use numbers. The harder skill — saying "I don't know" — is being learned far more slowly.

And one more thing: correlation is not causation. Croatia reaching the final did not prove my xG model right. It only proved the model was not systematically wrong. Those are two different sentences, and the distance between them is the entire difference between an analyst and someone selling predictions.

By the same logic, a team winning a match does not prove it played better. A goalkeeper keeping a clean sheet does not prove he reacted better than his counterpart. This is the biggest blind spot in modern football viewing: we use results to reverse-engineer process. Data is only useful when it runs from process to result, never the other way round.

The 2026 World Cup is the next test. The tournament runs from 11 June to 19 July 2026, with forty-eight teams, one hundred and four matches, and three co-hosts: the United States, Canada and Mexico. Semi-automated offside will be standard. Tracking data will be present in almost every match. That means more numbers than ever — and more ways than ever to misread them.

What is worth watching is not which figure is correct. It is which broadcaster dares to put the data provider's name on screen, and which journalist dares to write "insufficient information" when that is genuinely the case. A season missing detail is like a match missing stoppage time.