PPDA and the 2026 World Cup Semi-Final: France Were Not Cowardly, France Were Calculating

Krestovsky Stadium, Saint Petersburg, the night of July 10, 2026. The electronic clock ticked over to minute 51. Samuel Umtiti outjumped Marouane Fellaini by a few centimetres and headed the ball into the near corner of Thibaut Courtois's goal. The red shirts in the stands fell silent for about two seconds — longer than an ordinary passage of play.

On my tracking board, one number had announced the outcome as early as minute 20: France allowed their opponents only 8.2 passes before engaging; Belgium needed 12.5. When the referee blew the final whistle, possession favoured Belgium 63% to 37%, shots 22 to 19, corners 9 to 4. Those three lines were enough for most match reports the next morning to say the same thing: Belgium deserved more, France got lucky.

I left my spreadsheet untouched and did not edit a single cell.

PPDA: define it before arguing about it

PPDA — Passes allowed Per Defensive Action — is the number of passes a team permits its opponent to make per defensive action of its own: tackles, interceptions, duels, fouls. The lower the figure, the higher and earlier that team presses. The higher the figure, the deeper the block sits and the more it concedes control in the opponent's half.

I have used this metric more than any other since the summer of 2026, because it answers precisely the question that fine phrases cannot. "Fighting spirit" is unmeasurable. "Honesty in pressing" is measurable, through PPDA. A team that sits deep is not cowardly. A team that sits deep and still counter-attacks in under five seconds of winning the ball is executing a plan, not avoiding the game.

The origin of the spreadsheet

In April 2026 I sat in a rented flat in Beijing's Chaoyang district and calculated xG for Guangzhou Evergrande against Shanghai SIPG in the Chinese Super League. My model returned 1.2 for the hosts and 2.3 for the visitors. The bookmakers still had Guangzhou Evergrande as favourites at odds of 1.85. I backed Shanghai SIPG +0.5.

A male colleague in the newsroom laughed in my face and said a sentence I remember verbatim. I showed him the spreadsheet. The match finished 2-2, I won the bet and pocketed forty thousand yuan. From that day I built a standard template for every match: xG, shot count, possession, pressure, defensive-line height, counter-attack duration. No cell was reserved for sentiment.

Evidence, layered from the top down

The first layer is structure. Didier Deschamps's France defended with a back line whose average height was roughly 38 metres from Hugo Lloris's goal — considerably lower than Belgium's 46 metres. N'Golo Kanté and Paul Pogba split the two inside channels, Raphaël Varane and Samuel Umtiti held the last line, and Kylian Mbappé was placed in what I call the "launch arrow" position: one ball over the top and he is one-on-one with a full-back.

The second layer is the evidence chain. Belgium held 63% of the ball but completed only four passes into the final 25 metres before being blocked. All nine of their corners came from wide deliveries, not from central breakthroughs. In other words, the French block was not stretched; it was forced to deal with aerial balls, which is precisely what the Varane–Umtiti pairing does best.

The third layer is chance quality. My own model returned 1.4 xG for Belgium and 1.3 for France — a gap of 0.1. But strip out shots from outside the penalty area and the ratio flips to 0.9 against 1.1. That is the crux every summary report skips: Belgium shot more, but from places where Courtois could not be beaten, while France shot less, but from places where any goalkeeper could be.

The fourth layer is time. In the first 15 minutes of the second half, France held only 31% of the ball yet produced three counter-attacks involving four or more players. That is not coincidence. It is a rehearsed pattern, and it ended in a goal on 51 minutes.

The dangerous-control index

Possession is a lagging indicator. It tells you who had the ball, not where the ball went. In 2026, while tracking Roberto Mancini's Italy at the Euros, I built an additional metric: "dangerous control" — entries into the final 25 metres per 100 possession sequences. Italy led Europe with 18.2. I wrote a piece predicting they would win the tournament at odds of 11 to 1 and made two hundred and seventy-five thousand yuan.

Apply that metric retrospectively to the 2026 semi-final and the picture sharpens further. Belgium scored 9.1. France scored 14.7. The side with less possession was the side delivering the ball into dangerous areas more often per unit of possession it held.

The counter-intuitive angle: correlation is not causation

There is a mistake fans repeat at every major tournament: reading possession statistics the way you read a scoreline. High possession and high shot counts correlate with territorial dominance, but they are not causally linked to victory. Spain held 75% possession in the 2026 World Cup round of sixteen and went home after penalties. Germany held 72% against South Korea and went home too.

The blind spot is this: we tend to call the team keeping the ball "the team playing football" and the team sitting deep "the team spoiling football". But a well-organised low block is not a denial of football. It is one of the hardest forms of football to coach. It demands positional discipline without glory, and it is only ever called cowardly when that team loses.

If Samuel Umtiti had headed wide in the 51st minute, and Belgium had won 1-0 in extra time, the next morning people would have written that France were punished for passivity. Same data, two stories. The data does not change. The storyteller does.

Numbers never lie; only the people reading them lie to themselves.

The home-advantage shock and the limits of the model

In March 2026 my data contract was cut by sixty per cent. I was forced to rebuild my forecasting model from ten years of historical data. When the Bundesliga returned in May, matches were played in empty stadiums. My model showed home advantage falling by 37% without crowds. I bet according to that model and won 12 of 15 wagers.

Then I lost four in a row.

The cause was not the data. It was that I refused to update the parameters after the first three rounds, because I trusted my model more than I trusted the fact that teams had adapted. When the stadium falls silent, we finally hear the voice of probability clearly — but probability also needs to be heard again. Since then, every analysis I write ends with a fixed section titled "Assumptions and Lag".

Assumptions and lag

My PPDA model assumes defensive actions are recorded fully and consistently across data providers. A second assumption is that the defensive block does not change structure between the first and second halves. Both assumptions can break. If France switched from a 4-4-2 block to a 4-3-3 in the second half, their PPDA would rise — not because they stopped pressing honestly, but because the structure changed. Data lag runs roughly 24 to 48 hours after a match, which means any hot take written that night is a take without numbers.

The data lever: next-round signals

Across any team's last three matches, look at two numbers first: PPDA and completed passes into the final 25 metres. If PPDA falls while final-third entries do not rise, that team is pressing out of desperation, not design. If PPDA rises while final-third entries hold steady, that team is deliberately ceding territory to attack the space behind — the signature of a side that trusts its transition game.

That is the signal I track before any headline gets written.

Takeaway

In 2026 I put xG in front of the sceptics. Nine years later, they are still arguing. But the argument is no longer about whether xG is correct — it is about who reads it slowly enough to understand. PPDA is not a measure of spirit; it is a measure of honesty in pressing. And the real question of the 2026 semi-final is not whether France were cowardly, but this: when a team executes its plan and wins, why do we still insist they must play beautifully in our way?