21/12/24. Image: Jin Linwood
“If you look at the expected goals, it was 0.7 for them and 0.6 for us. It was a very tight game, they created very little, had very little number of shots on target, one more than us, that’s all,” Arsene Wenger said, following his side’s 3-1 defeat to eventual title winners Manchester City in 2017. This quote represents one of the first instances of the public gaining an insight into the statistical revolution taking place behind closed doors, to the bemusement of both fans and pundits alike. Notably, Jeff Stelling referred to xG as “the most useless stat,” following Wenger’s comments. “What does it tell you? The game’s finished 3-1, why do you show expected goals afterwards?”
xG, or Expected Goals, is a performance metric calculated by determining the likelihood of a shot being scored based on a variety of factors. These factors, taken from the moment a player shoots, vary depending on what statistical model is being used. Opta, who have one of the most commonly used xG models, considers, but is not limited to, location, angle of the shot, distance to goal, the type of shot, and the position of the goalkeeper, all of which are based on historical data of similar shots, which then allows the data analysts to assign this new shot an xG total. Penalties, which have the highest likelihood of a player scoring, are generally ranked between 0.77xG and 0.79xG, meaning the player taking the penalty would be expected to score 77-79% of the time.
So what does it tell you, exactly? Maybe Stelling is right; football isn’t a game of hypotheticals after all, it’s played in the real world. What does it matter if the outcome of Arsenal vs. City was more fairly represented as a draw, or a tight win in City’s favour? That wasn’t the outcome, and the cohort of armchair analysts, and old-school types who swear by the “eye test” have a point—more so when you consider what Nottingham Forest have managed to achieve so far this season.
Forest, remarkably, survived last season on 32 points, the lowest-ever tally to avoid relegation in the history of the Premier League, only to then storm up the table and establish themselves comfortably in the Champions league places so far this season. The narrative is irresistible: a sleeping giant, back at the top, challenging the elite against all odds.
According to Understat’s xPTS model (which calculates where teams would find themselves in the table based on expected points), Forest should be eleventh. Their actual league position? Third, a mirage of overperformance to the highest extremes. Pundits and fans would perhaps credit Forest’s transfer policy this season, or the manager’s shrewd counter-attacking tactical approach. The reality, however, paints a different picture.
Chris Wood, their top scorer, has 18 goals—five more than his xG tally of 12.32. But Forest’s luck goes far beyond one player. In a merciless 7-0 demolition of Brighton, they generated just 4.07 xG. In a 5-0 thrashing of Bournemouth, only 2.83 xG. The xG figures accrued by Forest, whilst still substantial, demonstrate overperformance. They’ve also won games despite being outplayed statistically—like beating Manchester United 3-2 at Old Trafford whilst generating just 0.65 xG to United’s 1.47. The pattern repeats against top teams: a 1-1 draw with Liverpool where Forest scored from 0.53 xG whilst Liverpool racked up 2.72.
The list goes on. They aren’t just overperforming—they’re defying probability on a weekly basis.
But what does this mean exactly? Is Stelling right? Are the traditionalists right? Is xG just another fad meant to “enhance” the understanding of football, when it means absolutely nothing? Well, not quite.
Football is a game that is predominantly luck, and over the course of a single game, anything can happen. A striker’s shot might cannon off the post instead of bulging in the back of the net. A deflection might take a shot inside the post when on another day it would’ve bounced out for a corner. VAR might make yet another contentious decision. The team’s goalkeeper might just be having the game of his life. Whatever it may be, overperformance is far from impossible, but overperformance is seldom sustainable. Eventually, teams regress to their expected level.
For Forest, this means their fairy tale is likely unsustainable. Just as Leicester’s 2015/16 title win was statistically improbable, Forest’s Champions League charge is built on sand. Both teams mirror each other in several ways. The main difference is just how poor a season every other “big 6” side had compared to their usual standards when Leicester claimed the title. Arsenal amassed the highest xPTS tally that year at 77.01, four less than the 81 points Leicester finished on to win the league – which they did from an xPTS tally of a meagre 68.94.
Beyond explaining results, xG is transforming how teams operate. Clubs like Brentford and Brighton use statistical models to identify undervalued players. Unlike top clubs that can throw money at superstars, these teams rely on efficiency.
Take two hypothetical strikers:
- Striker A scored 9 goals but had an xG of 17.4.
- Striker B scored 19 goals but had an xG of 12.5.
Striker B would win favour with traditional scouting; he’s the proven goalscorer. But a club like Brighton would sign Striker A. Yes, he scored fewer, but how much of that was down to the striker’s lack of ability, and how much of it was down to being unlucky? Striker B’s overperformance might suggest future regression. Signing the right player isn’t just about what happened; it’s about what should have happened. This is why Brighton can find gems whilst richer clubs make expensive mistakes. It’s why Brentford, with a fraction of Chelsea’s budget, remains competitive. Football isn’t just about spending — it’s about spending intelligently.
The run that Nottingham Forest have gone on this season is remarkable. Forest fans and neutrals are all hoping this fortune lasts to the end of the season. But history tells us it won’t last beyond then. When teams overperform xG this dramatically, reality catches up and eventually results even out.
This isn’t to say xG is a perfect tool—no stat is. The main critique is its inability to account for the individual skill of the player, instead relying on an average. But what it does offer is clarity when considering what’s sustainable and what isn’t. Football is played on the pitch, not on a spreadsheet, but ignoring this valuable data is like driving without a steering wheel.
Forest’s season might be thrilling, but it’s built on sand. As xG has shown time and time again, sand always shifts.