From a 2026 World Cup xG Spreadsheet to Morocco 2026: A Journey Through Defensive Data

On December 6, 2026, at Education City Stadium in Doha, Achraf Hakimi placed the ball on the spot for Morocco's third penalty. Yassine Bounou had already made two saves, Pablo Sarabia had hit the post, and the shootout stood at 2-0. Hakimi chipped it, a panenka, and the ball rolled slowly down the middle. Morocco were through to the quarter-finals.

I watched that match in Los Angeles, eighteen years old, with a spreadsheet open in my second browser tab, a file I had started three weeks earlier. The sheet recorded PPDA — the number of passes a team allows an opponent before each defensive action — for all 32 teams at the tournament. Morocco sat among the three lowest. I had published that assessment before the group stage closed, with one conditional sentence attached: if any team at this tournament were to break the dominance of the giants, it would be the most proactive defensive side, not the most attractive attacking one.

A month later, a tactics account with more than 200,000 followers shared that analysis. It was the first time I understood that what I was doing had stopped being a private hobby. But the real story started four years earlier, in an unremarkable Excel file.

The first spreadsheet

In the summer of 2026 I was fourteen and living in Los Angeles. When the World Cup in Russia kicked off, free xG data hardly existed in any form I could access. So I built my own. Across all 64 matches I logged more than 1,200 shots into a single Excel file, each one tagged with six variables: angle, distance, body part used, number of defenders within a two-metre radius, whether the move came from a transition or from settled possession, and whether the goalkeeper was off the goal-mouth axis. From those six variables I assigned weights and calculated my own goal-probability figure for every attempt.

That first xG spreadsheet taught me something: every goal has a hidden story. When France won the trophy, the media praised a flamboyant attack led by Kylian Mbappé, Antoine Griezmann and Paul Pogba. My spreadsheet pointed somewhere else. By my scale, France allowed opponents roughly 0.7 xG per match, the lowest figure among the semi-finalists. The champions did not win with firepower. They won with the ability to limit opponents.

From then on I built a fixed working routine: set the hypothesis first, collect data second, publish the prediction before the result happens, then go back and check where I was wrong. Most of the time I do not write to describe a match. I write to confirm or reject a model.

In 2026, when I started publishing my own analysis newsletter on Substack, I moved to a nine-dimension framework I had learned from esports analysts: patch and meta shifts, tournament format, roster and individual form, regional landscape, club finance, rules and governance, risk profile, public narrative and expectation, and finally industry-wide transmission. Football and esports differ on the surface, but the same layer of data sits underneath. A patch in League of Legends and a change to the five-substitution rule in football are both external shocks hitting a system that was running steadily, and both leave traces in match data.

The fastest thing I learned across six years of watching is this: most public models share one blind spot — they ignore defence. Goals are counted, replayed, written into the record. What happens before the goal — the depth of the defensive block, the timing of the pressing trigger, the position of the holding midfielder — never appears on the scoreboard, so nobody sells it to viewers. I chose to read the part nobody sells.

France 2026: a title built from the back

The final on July 15, 2026 at Luzhniki Stadium in Moscow ended 4-2 to France over Croatia. It was the highest-scoring World Cup final since 2026. The popular telling revolves around France's four goals and Mbappé's speed. My spreadsheet told a different story, and I want to rebuild that chain of evidence.

Throughout the tournament, France allowed opponents more shots than they took in three matches, including the round-of-16 tie against Argentina and the semi-final against Belgium. But the quality of the shots they conceded was very low. Didier Deschamps's defensive block ran on one clear principle: concede the ball in harmless areas, squeeze in dangerous ones. N'Golo Kanté enforced that principle, yet his value did not lie in tackle counts. It lay in how many times opponents were forced to change the direction of a pass.

I counted roughly four France goals at that tournament originating from set pieces or penalties, including Raphaël Varane's header against Uruguay in the quarter-final and Samuel Umtiti's header against Belgium in the semi-final. Both were moves in which the opposing defence was dragged out of shape before the ball arrived. In my model, the probability of those two headers was only average, but the delivery quality ranked among the best at the tournament.

The attack earned the attention; the defence earned the trophy. That is the conclusion I drew then and still hold today. When a team wins a short tournament, what decides it is not their ceiling but their floor. France in 2026 had a higher floor than everyone else: across seven matches, they barely had a single half in which their structure collapsed entirely.

Another detail I logged but only understood two years later: in the knockout rounds, France deliberately slowed the tempo. Their passes per minute in the second half were consistently lower than in the first, and that gap was larger than the tournament average. They were not losing control. They were choosing a slower match because in a slow match, squad quality matters more than luck.

Summer 2026: when the stands were empty

When the pandemic halted every league, I was sixteen and had a football-shaped gap in my life. I used it to do what I should have done earlier: compile data from more than 3,000 matches across the top five European leagues in the period before 2026, and isolate home advantage as a variable.

In my dataset, home teams benefited by roughly 0.38 goals per match across the full sample. I split that advantage into four components: crowd pressure on referees, crowd pressure on visiting players, familiarity with the pitch and conditions, and the effect of travel. The first three cannot be separated from one another using observational data, but the fourth can.

When the Bundesliga returned on May 16, 2026 with Borussia Dortmund against Schalke 04, every match was played in an empty stadium. I published a prediction before the first matchday: home win rates would fall, and the drop would land somewhere between one third and one half of those 0.38 goals. The first three matchdays after the restart confirmed the direction of my model, though the amplitude swung wider than I had expected.

Home advantage did not disappear. It decomposed into smaller parts, and the crowd was the largest part. That was the first time a prediction drawn from raw data I had collected myself came true within three weeks. When home is no longer home, you are forced to rewrite every assumption.

What mattered more than the figure was the method it gave me. An environmental variable — the presence of a crowd — can be isolated and measured if you have enough sample and enough patience. From that point on I began to see every match as a natural experiment, where each change to the laws, the schedule or the pitch conditions is an independent variable waiting to be tested.

Morocco 2026: defensive data spoke first

By the 2026 World Cup I had a method ready. I extracted two metrics for all 32 teams: PPDA — passes allowed per defensive action, where lower means more proactive pressing — and the average vertical length of the defensive block.

By my own measurements, Morocco sat among the three lowest PPDA teams at the tournament. Their approach was not deep defending inside the box. It was proactive defending in midfield, forcing opponents to pass in the direction they wanted, then recovering the ball where a counter-attack could begin instantly. Morocco's low possession share was not a consequence of weakness. It was a deliberate tactical choice.

Sofyan Amrabat was the centre of that system. He was not the kind of holding midfielder who specialises in last-ditch tackles inside his own box. He was the one who moved to seal gaps before the ball arrived, and his value lay in metres of off-ball movement that cameras never replay. Achraf Hakimi on the right flank did the opposite of a conventional full-back: he dropped before possession was lost, not after.

The model's results arrived step by step. On December 6, 2026, Morocco eliminated Spain on penalties after 120 goalless minutes. On December 10, 2026, they beat Portugal 1-0 through Youssef En-Nesyri's header. Morocco became the first African national team in history to reach a World Cup semi-final. On December 14 they lost 0-2 to France. On December 17 they lost 1-2 to Croatia in the third-place match.

Morocco 2026: when defensive data spoke first, the world listened afterwards.

What I want to stress here is not that the prediction was right. A correct prediction can be luck. What matters is that my model identified the mechanism, not just the outcome. It said that a team with under 40 percent possession can neutralise a team with over 60 percent, provided it controls where it wins the ball back. Once you hold the mechanism, you can apply it to other matches in other tournaments. Every dataset is a scripture, and I am a slow reader.

Euro 2026: corners, transfers and the perfection trap

In 2026, at twenty, I took an internship at a sports data analytics company in California. At the same time I handled corner data for a national team at Euro 2026 and evaluated transfer targets for a mid-table club.

The corner work taught me something standard xG models overlook: the real value of a corner is not the header that follows it, but the second ball after the defence clears. In my Euro 2026 dataset, the share of teams opting for short corners rose noticeably compared with previous tournaments, and the purpose of most short corners was not to create a direct chance but to drag the defensive line out of position and open space for the second ball.

At the same time, my transfer model flagged a target striker whose actual goals had fallen 4.5 short of expectation across a season. Most parties read that as decline. I read it as variance. His shot volume, shot locations and chance quality had not changed; only his conversion rate had. The club signed him, and he scored on his debut.

But that was also the period when I nearly lost myself. Perfectionism made me late delivering the corner report. A colleague told me something I still write down: a model that is 80 percent right and delivered on time is worth more than a perfect model delivered after the match ends. Transfer models price young potential through variance, and price dressing-room chemistry at zero. No index measures a player accepting three weeks on the bench so that the team functions better.

Where my model was wrong

A model is only right when you know exactly where it is wrong. Over the years I have re-checked my predictions, and the three weaknesses below are the ones I cannot ignore.

First, my 2026 home-advantage model has a causal gap. I attributed the drop in home win rates to the absence of crowds. But the pandemic also brought two simultaneous changes: five substitutions, and a fixture density denser than any previous period in the history of European league football. Both factors favour teams with squad depth, and big clubs usually play away as the higher-rated side. I assigned the entire drop to crowds while at least two other variables acted at the same time. That is the most basic error in data analysis: mistaking correlation for causation.

Second, my 2026 World Cup xG model had a goalkeeper blind spot. The six variables measured chance quality but not the quality of the man in goal. Hugo Lloris made saves that tournament to which my model assigned a high goal probability. Had I included goalkeeper quality as a variable, France's defensive value would have been even higher than what I published. I had inadvertently undervalued the champions inside my own model.

Third, I once applied an esports framework to football without fully checking the assumption of equivalence. Esports has clear patch cycles, public update notes and match-level server data. Football has none of that. The closest thing to a patch in football is a change to the laws, and law changes in football move so slowly that you need multiple seasons before the sample is large enough for a conclusion.

The same cluster of problems appears with VAR. When VAR arrived, the common expectation was that controversy would shrink. My data on major tournaments shows the opposite: controversy does not shrink, it relocates. On June 29, 2026, in Germany's round-of-16 match against Denmark at Euro 2026, within a single minute a Danish goal was disallowed for offside determined by semi-automated technology, and then a penalty was awarded to Germany for handball. What used to be one argument on the pitch became two arguments in the review room, and neither could be settled by the naked eye.

VAR moves controversy from the pitch into the review room and the grey zones of the law; it does not erase it. When the intervention threshold is described as clear and obvious, the definition of clear and obvious itself becomes the new point of dispute. No technology solves that, because it is not a technology problem.

As someone who works with data, I have learned that the three weaknesses above do not make my models useless. They make them honest. A model published with a confidence interval and a list of what it ignores remains more useful than a model presented as if it knows everything.

Signals for the next round

Based on my experience watching matches over the past six years, there are three signals I will track in the coming period.

The first is the distribution of PPDA by matchday, not the tournament-average PPDA. A team can press fiercely in the group stage and collapse on that metric in the knockout rounds when the legs are gone. The break point of the curve matters more than the mean.

The second is the value of the second ball after a corner. If the short-corner trend keeps rising, the transfer market will soon pay for midfielders who read second balls, not centre-backs who head them. I have already seen the first signs of this.

The third is the gap between actual and expected goals for young strikers. Players with a large negative gap in one season are usually priced low by the market, and most of them regress to the mean the following season. But only those whose shot volume and chance quality remained unchanged are worth backing.

I do not predict the future by intuition; I only read the traces that numbers leave behind. And the most reliable trace always sits on the side the audience is not looking at.

The most interesting part of next season will not be the champion. It will be the first team to publish its defensive metrics publicly, with confidence intervals, before the tournament begins. That is for anyone patient enough to wait a whole season to prove a number.