Sunderland vs Arsenal: The 0.93 xGA and the Real Test for Arsenal's Perfect Start

On 13 September 2026, the Stadium of Light opens its doors to Arsenal in round four of the Premier League 2026/2027. On my tracking sheet, two numbers sit side by side, and they tell a story the league table cannot. Arsenal arrive with the lowest expected goals against in the Premier League, 0.93. Sunderland, after three rounds, have allowed their opponents just five shots on target, one fewer than the visitors. Fans read balance. I read a neatly presented confusion. Two defences look equally good on paper, but they are built from two different materials. One defends by never letting the opponent touch the ball in dangerous places. The other defends by letting the opponent touch the ball in harmless places and trusting probability. Numbers never lie. Only the people reading them do. BEFORE KICK-OFF Regis Le Bris and Sunderland enter round four on a three-match unbeaten run across all competitions. They beat Fulham 1-0, drew 1-1 with Brentford, then edged Hull City 1-0 in the Carabao Cup. None of those games made anyone rise from their seat. Yet all three ended with no opponent scoring more than once at the Stadium of Light. Arsenal move in the opposite emotional direction. They have won all five matches they have played this season. Midweek they beat Napoli 1-0 in the Champions League, a result European analysts framed as a sign of maturity: winning without playing well. In the Premier League they have also opened with three straight victories, and they now stand a chance of recording four wins in their first four fixtures for the sixth time in club history. Looking at that data strip, most people reach for the word easy and type it into the draft. I sat with the spreadsheet a little longer than usual. Arsenal have not lost to Sunderland in 17 consecutive Premier League meetings since 2026, including a 3-0 win in February. That is an impressive figure, and it is also the most dangerous figure in this entire match file. I will come back to it at the end. WHAT THE 0.93 ACTUALLY SAYS Expected goals against measures the quality of chances an opponent creates, not the number of goals conceded. A team with an xG against of 0.93 means that, per match, its opponents generate just under one expected goal. At Premier League level, that is the data zone of title-winning defences. What stands out is that Arsenal reach that level without defending in a low block. They control the ball, push the line high, and turn defending into an attacking act. When they lose the ball, they win it back within seconds, usually in the opponent's half. The chances opponents create after such sequences are of very low quality: long shots, shots while turning, or crosses with no destination. PPDA is not a measure of spirit. It is a measure of honesty in pressing. When I dissected the 2026 World Cup semi-final between France and Belgium, I learned something I still use today. Belgium allowed 12.5 passes before pressing; France allowed only 8.2. Many read that number and concluded France played negative football. They asked the wrong question. France did not dodge pressing; France chose where to press. They let Belgium hold the ball at the back, then squeezed in midfield. The match ended 1-0 to France. Arsenal 2026/2027 are playing the same logic, but with a squad of higher technical quality. An xG against of 0.93 is not a reward for luck. It is the output of a system that knows what it wants when it does not have the ball. SUNDERLAND'S FIVE SHOTS ON TARGET Now comes the part of the data that made me reopen the video. Sunderland have allowed only five shots on target in the Premier League this season, one fewer than Arsenal. Read conventionally, this is evidence of an excellent defence. Read the way a meta hunter reads it, this is a variable that needs checking before it can be believed. First question: who were the opponents? Sunderland's opening three rounds were Fulham, Brentford and a Carabao Cup tie with Hull City. Those are attacks at or below the Premier League average. Arsenal, across their opening three rounds, faced better-organised attacks, plus Napoli in the Champions League. Comparing five and six directly, without adjusting for opponent quality, is a statistically meaningless exercise. Second question: in which game states did Sunderland defend? A team that regularly takes an early lead defends from a position of advantage, with the opponent forced forward and forced to accept risk. A team that falls behind defends from a position of disadvantage, and the shots on target it concedes tend to spike in the final 20 minutes. In 2026, after the pandemic froze global football and my data contract was cut by 60 percent, I built a forecasting model on ten years of history. When the Bundesliga returned that May, the data showed home advantage dropping 37 percent without crowds. I won 12 of 15 bets following the model. Then I got complacent, refused to update parameters after the first three rounds, and lost four in a row. That lesson shapes how I read Sunderland's numbers today. A pretty defensive figure across three rounds says nothing yet about its capacity to withstand an elite attack. DANGEROUS CONTROL: THE DATA LEVER FOR THIS MATCH At Euro 2026 I built an index I have used regularly ever since: dangerous control. The calculation is simple. I count how often a team moves the ball into the opponent's final 25 metres, then divide by every 100 possessions. Mancini's Italy led Europe with 18.2. They held 60 percent of the ball without ever being harmless. I wrote a piece predicting Italy to win the title at 11/1 and made 275,000 yuan. A European betting company then hired me as a data consultant, and I set up a three-step meta-detection process with three colleagues cross-checking each other. I bring that index into Sunderland against Arsenal because it cuts exactly at the difference between the two defences. Arsenal control a lot of the ball and control it dangerously. Sunderland control little of the ball but, across the last three matches, converted that small share into purposeful attacks. This is the team profile that breaks simple forecasting models: they do not win through volume, they win through position. If Sunderland hold a dangerous control rate above 15 in the first half, the match will follow a script Arsenal dislike: the ball in midfield, few gaps, and every turnover by the visitors turning into an organised counter-attack. THE CONTRARIAN ANGLE: SEVENTEEN UNBEATEN IS A TRAP Prejudice is a match played without data. I choose to bet on the number. Arsenal's 17-match unbeaten run against Sunderland in the Premier League, stretching back to 2026, is one of the longest head-to-head sequences in the division. It appears in every preview, every bulletin, every status update. And it is close to meaningless for the match on 13 September 2026. Those 17 matches span nearly two decades, several managerial eras, several squads, and most importantly, different divisions. The sequence does not measure Sunderland's current strength. It measures the financial and competitive gap between the two clubs during a specific historical period. When Sunderland were relegated and came back, the number was preserved like a relic. Correlation is not causation. Arsenal beating Sunderland 17 times in a row does not give them an extra edge in the 18th. Probability does not remember head-to-head history; it records squad quality, physical condition and match context. The same applies to the 3-0 win in February. That was a match from last season, with a different squad, different motivation, and a Sunderland that may have been in decline. Quoting that scoreline to predict this scoreline is the behaviour of someone reading results tables, not someone reading data. It holds for the statistic about four wins in the first four rounds for the sixth time in history. It is an interesting historical footnote, and its predictive power is zero. The previous five occurred in different Arsenal squads, under different managers, in different league situations. I say this after nine years of putting xG into analysis pieces in Beijing. In 2026 I calculated xG for Guangzhou Evergrande against Shanghai SIPG: 1.2 for the hosts, 2.3 for the visitors. The bookmakers priced Guangzhou as favourites at 1.85. A male colleague laughed and asked what a woman could possibly know about football. I handed him the spreadsheet. The match ended 2-2, I won the bet and pocketed 40,000 yuan. What I remember is not the money. What I remember is being treated as an eccentric for reading a match through numbers before it was played. In 2026 I put xG in front of the sceptics. Nine years later, they are still arguing. The probability sheet does not need anyone's agreement. THE BLIND SPOTS ON BOTH SIDES Sunderland have one obvious blind spot. Their defence has never been tested by a side capable of moving the ball at high speed for 90 minutes. Fulham, Brentford and Hull City all let Sunderland organise their defensive block at their own rhythm. Arsenal will not do that. Arsenal will stretch, switch flanks, and force Sunderland's defenders into decisions while fatigued. Arsenal have a blind spot too, and it is rarely discussed. They have not yet faced an opponent willing to give up the ball entirely. Five straight wins look impressive, but one of them was a 1-0 against Napoli in which they had to endure the closing minutes. When a team accustomed to control has control taken away, its response rarely appears in the data. It lives in psychological endurance. Here I must be explicit about the limits of the model. My spreadsheet holds data on xG, on shots on target, on PPDA, on dangerous control. It holds nothing on how a 22-year-old defender will react when 45,000 people at the Stadium of Light rise to their feet in the 88th minute. I do not pretend otherwise. SIGNALS TO WATCH When the stadium falls silent, we finally hear the voice of probability. At the Stadium of Light on the night of 13 September, the ground will not fall silent, so I have prepared three signals to read the match through numbers rather than emotion. I start with Sunderland's PPDA in the opening 20 minutes. If that figure goes above 15, they have accepted a deep block and will let Arsenal hold the ball high up the pitch. It is a rational choice for an underdog, but it turns the match into a patience test for the visitors. The next signal is how often Arsenal move the ball into the final 25 metres in the first half. If that figure sits below 20, their attacking system is being broken in midfield, and the match will drift toward a narrow result. The most important signal sits in the shots on target Sunderland concede. Their figure of five will not survive this match intact. The question is whether it climbs to seven, nine or fifteen. The answer will reveal whether Regis Le Bris has a real structure or merely a number generated by three weak opponents. I keep my analytical framework intact, but I update parameters after every round according to a fixed process. Every spreadsheet is a monastery. I go in to find truth, not consensus. If Arsenal win at the Stadium of Light, it will be four wins from four, and the whole of England will call it dominance. If Sunderland hold them, it will be called a shock. Both labels are convenient, and both ignore what the data said in advance: this match is decided by who controls the final 25 metres, not by a head-to-head sequence stretching across seventeen years.