Sunderland vs Arsenal at the Stadium of Light: the 0.93 xG defence and a 17-match unbeaten run under the data lens

In my office in Incheon, I taped two A4 sheets to the wall. The first carried one line: 0.93. The second carried one line: 17. The first is the lowest expected goals against figure in the Premier League, held by Arsenal. The second is the number of Premier League meetings Arsenal have avoided defeat against Sunderland since 2026. Both sheets arrived attached to an identical press release in different languages, ahead of the fixture at the Stadium of Light, matchweek 4 of the 2026/2027 Premier League, kicking off at 02.00 WIB on Sunday, 13 September 2026 — 02.00 in Vietnam, 04.00 for me in Incheon. I tape them to the wall instead of writing them in a notebook, because I need to see them every time I look up. The habit dates back to 2026, when I cross-checked a Korean club's financial statements against player registration records and found a 2.3 billion won gap hiding inside the agent fee line. Sunderland arrive unbeaten in three matches across all competitions: a 1-0 win over Fulham, a 1-1 draw with Brentford, and a 1-0 win at Hull City in the Carabao Cup. That record belongs to Regis Le Bris and is repeated as a mood rather than a metric. Arsenal arrive differently: five wins from five competitive matches this season, including a 1-0 Champions League win over Napoli in midweek. In the league they have won all three games, and a result on Sunday would give them four wins from four openers for the sixth time in club history. The release also states Arsenal have conceded one goal in every competitive match this season while holding the league's lowest xG against, 0.93. Sunderland are described as defensively sound: they have allowed only five shots on target in the Premier League, one fewer than Arsenal. Arsenal are unbeaten in their last 17 league meetings with Sunderland since 2026, including a 3-0 win in February. I underlined those first four sentences not to remember them but to interrogate them. Years ago I spent a full season tracing a national team's physical data and learned that three commercial providers can produce three different xG models for the same dataset, differing by up to 0.15 per match. Spread across 38 rounds, that margin is worth a European place. For this fixture I opened three independent sources to establish whether 0.93 is a per-match average or a three-round total. They disagreed, and none of them published a definition. Numbers do not lie, but the people writing financial reports do. Matchday press officers belong to the same professional family. Football is not clean, but financial statements taught me how to find the stain line by line. I do not need a confession; I need one line that does not reconcile. The first such line is the claim that Arsenal have conceded one goal per match. A team that wins five straight while conceding in every game must score at least twice each time, meaning ten goals in five matches. Another team, conceding once across the whole run, could be described with the same sentence. Both readings leave a hole in the 0.93 column. The second line is the "one shot on target fewer" comparison. That belongs to what I call stepped data: meaningless without match count and opponent quality. Sunderland allowed five shots on target; Arsenal six. But Sunderland's league fixtures came against Fulham and Brentford plus a Carabao Cup tie with Hull City, while Arsenal faced three Premier League sides and Napoli. Weighted by opponent quality, a gap of one shot becomes a far wider gap. I did the weighting by hand, because no platform publishes its own weights. Regis Le Bris builds Sunderland around a deep block, ceding possession and countering through two quick runners. That structure manufactures attractive defensive numbers by default: deep teams always allow fewer shots on target because opponents are forced to shoot from distance and through bodies. Arsenal push high, hold the ball and squeeze. Their 0.93 xG against therefore carries far more weight than Sunderland's five shots allowed. Two "solid" defences working through two different mechanisms, flattened by any single-metric table. Watching both sides on late-night screens across the first three rounds, logging every long pass and set piece, what caught my attention was not the defence. Arsenal are winning by controlling scorelines: one-goal margins, including the 1-0 in Naples. That pattern belongs to a team that has moved from exhibition to risk management. For such a side, a low xG against has two sources: genuine defending, and the deliberate reduction of opponent attacks by not losing the ball in their own half. The spine I track at Arsenal still runs through Declan Rice and Martin Ødegaard in midfield, with William Saliba and Gabriel Magalhães at centre-back; that continuity explains most of the 0.93. The blind spot sits there. Nobody publishes ball-control indices at the moment of turnover, tactical fouls that cut counters, or dead-ball time. When those lines do not exist, the picture is filled with prettier, more sellable metrics: xG, shots on target, unbeaten runs. Hidden transfers are not in the news release; they sit in a footnote nobody turns to. The same applies here: the real match lives in the rows that never make the brief. Another variable sits at the bottom of the page: the calendar. Arsenal have just played Napoli in the Champions League, travelled, rotated, and returned to England before heading to the northeast. Sunderland also played midweek, but in the Carabao Cup, against Hull City, with a heavily rotated side. On the surface both are tired. Structurally, Arsenal have two options per position; Sunderland have one plus a fallback. Between the 60th and 85th minute — the window I always mark in red — squad depth converts into shot volume. The Stadium of Light is another variable that models usually assign a fixed coefficient. That crowd can push a deep block five metres higher, enough for the second defensive line to form after a turnover. The 02.00 WIB kickoff carries no sporting meaning; it only matters to viewers in Asia, me included. The international broadcast schedule and the competitive calendar are designed by two different departments, and the second one is rarely consulted. The 17-match unbeaten run against Sunderland is the prettiest line in the release and the most decayed. Across that span Sunderland changed owners, changed managers repeatedly, were relegated and promoted, and barely kept a single player from the first meeting. A seventeen-year streak only carries statistical value when both structures hold. Here, only the shirt colour holds. The 3-0 win in February is the data point with weight, because it belongs to last season and to this squad. The real signal lies in the first twenty minutes. Sunderland will test Arsenal with an early press; if Arsenal survive it without losing the ball in their own half, the match settles into a familiar shape: possession in three-quarters of the pitch, a two-bank block, and set-piece quality deciding the outcome. If Sunderland score first, the structure flips entirely and 0.93 stops being a shield and becomes a reminder that Arsenal have not trailed all season. On refereeing, one risk is specific. A deep side like Sunderland lives on counters launched from the halfway line, and those are precisely the situations where offside technology flags offences measured in centimetres. A goal erased in that moment never appears in any xG chart, because it never existed as data. I have spent years watching matches bent by a single line, and my position stands: when the referee becomes the match's editor, attacking instinct is taxed and defenders are handed an invisible extra man. Another layer the release omits is availability. Clubs publish injuries the way a conglomerate preparing to raise capital publishes its balance sheet: bury what needs burying in the annex, promote what sells. Doping does not start with a syringe; it starts with the silence of the dressing room. At a smaller scale, injuries do not start in the medical room but with a team sheet filed a few minutes late. I read that sheet before I read any predicted lineup. The final layer is who supplies the data. Matchday metrics are not computed by clubs; they come from commercial providers who sell the same dataset to media, bookmakers and clubs alike. When one dataset is sold to parties with competing interests, its framing always leans toward the sellable story. That does not create lies. It creates an edited truth, and edited truth is this industry's raw material. Calling Sunderland mere makeweights would be wrong. Three unbeaten matches is a real run, two clean sheets in the last three, and the 1-0 over Fulham showed they can manage a narrow game. The league needs sides like that, because the nuisance of a home deep block is something every xG model ranks low. If I argue the 17-match streak has no predictive value, Sunderland fans are entitled to ask why Arsenal have never found it easy there. The blind spot is elsewhere. The two clubs are selling different products: Arsenal sell perfection, Sunderland sell endurance. Both need data to price themselves, and both have incentives to pick flattering data. When a fixture is marketed on the away side's unbeaten streak, pressure shifts to the referee, who must run a game in which every incident is reviewed frame by frame. In that system defenders are rewarded, attackers penalised, and viewers receive a product that has passed three layers of approval. If I stake any attention at all, I stake it on the 20th minute and on the team sheet. Everything after that will be recorded in the minutes; what is kept out of the minutes is what decides the story. I want a day when the league publishes raw tracking data, unedited, so supporters can verify instead of believe. Until then, someone has to sit and tape sheets to the wall.