The World Cup is almost here again, and with it comes one of football’s strangest analytical environments.
National teams have limited time together, tactical identities are compressed into short preparation windows, and three group-stage matches can define whether a project survives or disappears. In club football, performance has months to correct itself. In tournament football, the margin is much thinner.
This paper looks at the group stage of the 2014 FIFA World Cup and asks a deceptively simple question:
Which match statistics were actually related to winning?
It does so through a model that tested 24 match statistics against match outcome, separating all games from close games. World Cup football is not only about dominating weaker opponents, it is also about surviving the matches where one goal, one transition, one yellow card, or one blocked shot can change the entire tournament.
Most World Cup previews will focus on squads, stars, managers, and narratives.
That is part of the fun. But when the tournament starts, the conversation usually becomes more basic and more brutal: who creates the better chances, who controls transitions, who avoids giving the opponent the wrong kind of opportunity, and who manages the emotional chaos of close games?
This study is useful because it gives us a compact framework for that question.
The authors analysed the 48 group-stage matches from Brazil 2014, using 24 match statistics grouped into goal-scoring, passing/organisation, and defensive variables. They then modelled how a two-standard-deviation increase in each statistic changed the probability of winning. Importantly, they also separated close games from unbalanced games, because a 4-0 win and a 1-0 win do not tell the same tactical story.
In all group games:
Nine statistics had clearly positive effects on winning probability: shots, shots on target, shots from counter-attacks, shots from inside the box, possession, short passes, average pass streak, aerial advantage, and tackles.
Four had clearly negative effects: blocked shots, crosses, dribbles, and red cards.
When the authors focused only on close games, two things changed: aerial advantage became trivial, while yellow cards became clearly negative.
That says a lot about tournament football. Some things help when the game state opens up. Some things help when one team is clearly better. But in tight World Cup matches, the value of a metric depends on whether it reflects control, quality, or risk.
And this is the point I would keep in mind as we approach 2026.
The World Cup is not always won by the team that produces the cleanest season-long profile. It is often won by the team that can repeatedly create high-quality moments while reducing the number of situations where the match becomes random.
That is exactly what appears in the paper’s strongest findings.
Shots mattered, but shots on target mattered much more. The authors found that a two-standard-deviation increase in shots was associated with a 13% higher probability of winning, while the same increase in shots on target was associated with a 48% higher probability. In close games, simply increasing total shots became unclear, while blocked shots were associated with a lower probability of winning. The message is obvious, but still often ignored: tournament football rewards shot quality more than shot volume.
That is especially relevant in a World Cup context, where game states can push teams into low-value pressure. A team chasing a result may accumulate shots, crosses, corners, and space, but that does not necessarily mean it is creating the kind of chances that decide games. In fact, one of the paper’s more useful insights is that blocked shots had a negative relationship with winning. That does not mean blocked shots cause teams to lose. It means that a shot blocked may often be a symptom of something else: poor spacing, rushed decision-making, ...
We now have xG, post-shot xG, shot pressure, freeze-frame data, goalkeeper positioning, defensive density, and tracking-derived models. We can analyse this far better than researchers could with public match statistics in 2014. But the underlying question remains the same:
Are you creating shots the opponent struggles to defend, or merely recording attempts?
