Tactical trends report · FIFA WORLD CUP 2026

The underdog does not lose.
The barren game model loses.

All 104 official FIFA Post-Match Summary Reports of the 2026 World Cup, parsed in full — 208 team-matches, 3,289 player-matches. The hypothesis I brought in was that the side which actively controls the game wins. The data supports it, and corrects it in one decisive place: what loses is not the side that sits deep but the side that creates nothing — and that applies just as much to the side holding the ball.

0 wins in 15barren reactive, knockoutsDeep block, no threat. Not one win in fifteen knockout matches.

88.9%productive reactive win rateThe same deep block, but 1.5+ xG. Nine matches, 8W-1D — the best record of any group.

2.31 vs 1.68three models vs twoSides that moved between more in-match models went further (mean round reached).

86.2 vs 76.2%xG vs possessionHow often each picks the winner. Proactivity is chance quality, not the ball.

01 / Method

What the data is, and what it can't tell you

Every number comes from FIFA's own Post-Match Summary Report for that fixture. I parsed all 104 PDFs end to end — the key-statistics block (xG, line breaks, receptions in the final third, forced turnovers, pressures, distance), the phases-of-play shares, the in-possession and defensive line-height, team-length and team-width diagrams, both starting formations, and the per-player distribution and offer-movement tables. That last layer alone is 3,289 player-match rows. Nothing here is estimated or taken from a secondary provider.

Classification comes from in-match behaviour, not reputation. Deep block = a low-block share of 30% or more. Ball-heavy = 50%+ possession. Both are then cut at 1.0 xG created. That simple 2×2 is the spine of this report.

Correlation is not causation

Low-block share is both a cause and a symptom. Weaker sides sit deep because they are being pushed back, and lose because they are weaker. When this report says barren reactive loses, it does not mean sitting deep caused the defeat. The closest thing to causal evidence is strong sides — the Netherlands, Belgium — choosing passivity themselves, and that is only a handful of matches.

One more: FIFA's phase shares do not sum to 100 because the phases overlap, so I use them only for like-for-like comparison between teams, never as a time budget. Extra time is not flagged in the PMSR either, so I infer it from distance covered (132 km+ per team).

02 / The correction that matters

The trap is not being reactive. It is being barren.

My original hypothesis was that the underdog's game model — extreme low block, defend-then-counter — is what loses. The 104 matches ask me to change one word in that sentence.

I split the 62 team-matches that sat in a deep block by the xG they created. What came back was not one group but two completely different ones.

Figure 1The fracture inside the deep block

xG <0.5n=21

win 4.8% · 1-5-15 · pts/g 0.38

xG 0.5–1.0n=25

win 12% · 3-11-11 · pts/g 0.8

xG 1.0–1.5n=7

win 14.3% · 1-2-4 · pts/g 0.71

xG 1.5+n=9

win 88.9% · 8-1-0 · pts/g 2.78

THE SYMMETRIC CUT — 79 team-matches with 50%+ possession

xG <0.5n=5

win 20% · 1-0-4 · pts/g 0.6

xG 0.5–1.0n=13

win 30.8% · 4-3-6 · pts/g 1.15

xG 1.0–1.5n=12

win 50% · 6-6-0 · pts/g 2

xG 1.5+n=49

win 75.5% · 37-8-4 · pts/g 2.43

The 62 team-matches with a 30%+ low-block share, split into four bands by xG created. Inside the same “reactive” label, the win rate runs from 4.8% to 88.9%.

Both ladders point the same way. Deep block or ball dominance, what decided the result was not the posture but whether threat came out of it. And at the top the order flips: productive reactive at 88.9% beats productive possession at 75.5%. Sitting deep while still creating was the single strongest profile of the tournament.

Figure 2The two traps — barren reactive and barren possession

posture

barren possessionn = 1631.2%

5W 2D 9L · xG 0.66 / 1.13

productive possessionn = 5869%

40W 14D 4L · xG 2.08 / 0.65

barren reactiven = 468.7%

4W 16D 26L · xG 0.52 / 1.9

productive reactiven = 1656.2%

9W 3D 4L · xG 1.64 / 1.34

under 1.0 (barren)xG created →1.0+ (productive)

ball-heavy ↑ / deep block ↓

Horizontal: 1.0 xG. Vertical: posture — deep block or ball-heavy. The dark cells are the traps. Whichever side of the axis you stand on, without threat the record collapses.

In the knockout stage the barren reactive model went 0W 3D 12L across fifteen matches. Not a single win. The nine matches that combined the same depth of block with 1.5+ xG went eight wins and a draw. Sitting deep was never the problem.

Figure 3Who was actually in each trap

PRODUCTIVE REACTIVE — 30%+ low block & 1.5+ xG

88.9%8W 1D 0L · n=9

Team	v	R	Poss	Blk	xG	Res
🇦🇷 아르헨티나	🇦🇹 오스트리아	조별	48.2%	31%	2.45	2-0 W
🇳🇴 노르웨이	🇨🇮 코트디부아르	32강	50.2%	30%	2.3	2-1 W
🇮🇷 이란	🇪🇬 이집트	조별	36.4%	30%	2.24	1-1 D
🏴󠁧󠁢󠁥󠁮󠁧󠁿 잉글랜드	🇲🇽 멕시코	16강	31.5%	49%	2.12	3-2 W
🇦🇷 아르헨티나	🇨🇭 스위스	8강	54.7%	30%	1.99	3-1 W
🇲🇽 멕시코	🇪🇨 에콰도르	32강	37.1%	38%	1.86	2-0 W
🏴󠁧󠁢󠁥󠁮󠁧󠁿 잉글랜드	🇳🇴 노르웨이	8강	47.4%	30%	1.64	2-1 W
🇦🇷 아르헨티나	🇩🇿 알제리	조별	44.6%	33%	1.58	3-0 W
🇳🇴 노르웨이	🇸🇳 세네갈	조별	40.1%	32%	1.5	3-2 W

Possession averages barely 45% here, yet the win rate is 88.9% — sitting deep was never the problem.

BARREN POSSESSION — 50%+ possession & under 1.0 xG

31.3%5W 2D 9L · n=16

Team	v	R	Poss	xG	Res
🇿🇦 남아공	🇨🇦 캐나다	32강	55.8%	0.16	0-1 L
🇺🇸 미국	🇧🇪 벨기에	16강	51.2%	0.33	1-4 L
🇩🇿 알제리	🇦🇷 아르헨티나	조별	50.8%	0.41	0-3 L
🇪🇨 에콰도르	🇲🇽 멕시코	32강	52.9%	0.47	0-2 L
🇭🇷 크로아티아	🇬🇭 가나	조별	50.8%	0.48	2-1 W
🇵🇹 포르투갈	🇨🇩 콩고DR	조별	68.9%	0.57	1-1 D
🇭🇹 아이티	🏴󠁧󠁢󠁳󠁣󠁴󠁿 스코틀랜드	조별	51.6%	0.7	0-1 L
🇩🇿 알제리	🇨🇭 스위스	32강	51.5%	0.73	0-2 L
🇰🇷 한국	🇿🇦 남아공	조별	60.9%	0.74	0-1 L
🇵🇦 파나마	🇬🇭 가나	조별	55.5%	0.77	0-1 L
🇲🇦 모로코	🏴󠁧󠁢󠁳󠁣󠁴󠁿 스코틀랜드	조별	54.8%	0.8	1-0 W
🇭🇷 크로아티아	🇵🇦 파나마	조별	52.5%	0.82	1-0 W
🇲🇦 모로코	🇨🇦 캐나다	16강	50.7%	0.85	3-0 W
🇰🇷 한국	🇲🇽 멕시코	조별	51.8%	0.87	0-1 L
🇪🇬 이집트	🇮🇷 이란	조별	53.7%	0.92	1-1 D
🇨🇮 코트디부아르	🇨🇼 쿠라사우	조별	55.2%	0.93	2-0 W

Possession averages over 55% here, yet the win rate is only 31.2%. Holding the ball guarantees nothing on its own.

Left: nine matches that produced real threat from a deep block. Right: sixteen that held the ball and produced nothing. Argentina appear three times on the left, England twice. The large number above each table is that group's overall win rate.

Figure 4The same team, both ways — holding quality constant

What this showsEach team compared only against itself. Opponent quality and talent gaps never enter the picture.

Horizontal axisPoints per game (W=3, D=1, L=0, averaged). The grey dot is the mean across that team's own matches under 1.0 xG; the green dot is the mean across its matches at 1.0 xG or above.

The connecting lineIts length is the 'cost of being barren' for that team — the longer it is, the more points that side loses specifically when it fails to create.

Right-hand numbersBarren matches / productive matches. Teams with a small sample (3–4 matches) are more exposed to noise — read them as indicative, not definitive.

that team's barren matches (xG <1.0)that team's productive matches (xG ≥1.0)

All 48 World Cup entrants. The 36 teams that played 4+ matches and produced both kinds — 162 team-matches. Comparing a side only against itself means "they were the weaker team" cannot explain the gap. The right-hand figures are barren/productive match counts.

The mean within-team gap is +1.13 points per game. 27 of 36 sides took more points in their own productive matches, 4 were level, 5 went the other way. The five that inverted — Morocco, Mexico, Croatia, Côte d'Ivoire and others — are drawn in as well. The measure is not infallible, which is precisely why there is no reason to hide them.

The 12 sides that cannot be compared

These sides played only one type of match. The absence of a comparison is itself a result — especially the two extremes below.

Spain7W 1D 0L · all productive

Türkiye1W 0D 2L · all productive

Australia1W 2D 1L · all barren

Bosnia and Herzegovina1W 1D 2L · all barren

Curaçao0W 1D 2L · all barren

Ghana1W 1D 2L · all barren

Haiti0W 0D 3L · all barren

Jordan0W 0D 3L · all barren

Panama0W 0D 3L · all barren

Paraguay1W 2D 2L · all barren

Scotland1W 0D 2L · all barren

Tunisia0W 0D 3L · all barren

Figure 5Which is why possession does not pick the winner

86.2% 69/80

Shots on target

86.3% 63/73

Possession

76.2% 61/80

Shots

74.7% 56/75

Defensive line breaks

72.7% 56/77

Completed line breaks

71.2% 57/80

Receptions in final third

66.2% 53/80

Forced turnovers

61% 47/77

Distance covered

57.5% 46/80

80 of 104 matches were decided. For each, did the team leading this metric go on to win? Ties on the metric are excluded, so denominators differ.

Figure 6The dividing line is horizontal, not vertical

One circleOne team's one match — all 208. Both sides of the same fixture each get a point; they always come in pairs.

Horizontal = possession %Left means less of the ball, right means more. The vertical dashed line marks 50%.

Vertical = xG createdExpected goals generated in that match. The horizontal dashed line marks 1.0 xG — the threshold this report uses for “threat”.

ColourGreen = won, dark grey = drew, light grey = lost. The dividing line runs top-to-bottom on xG, not left-to-right on possession — that is the whole point of this chart.

WonDrewLosthover a point

All 208 team-matches. Winners sit high regardless of where they sit left to right. England winning on 31.5% of the ball and Portugal failing to win on 68.9% are both on this chart. Hover any point for a detail card.

03 / The model to build

Neither all-out attack nor time-wasting — the switchable team

There is one way to avoid both traps: be able to choose the route by opponent. That is what mutation looks like in practice, and it is the concept this tournament supported most directly.

What I mean by being switchable is not choosing between active and passive. It is being able to move selectively between attacking and defending as the situation demands. One game model is not enough for that — opponents adapt to your team as the tournament goes on, and pre-empting that adaptation means arriving with several models already prepared. That plurality has to run in both directions: not one defensive method but several, and not one attacking method but several, because that is what it takes to face a range of opponents, and the same opponent's shifting response to you.

I classified every match as proactive, balanced or reactive from its own metrics, then counted how many distinct models each side used across the tournament. Thirty-two teams played four or more matches.

Distinct models used	Teams	Mean round reached	Points per game
2	19	1.68	1.41
3	13	2.31	1.78

Round is coded 0 = group exit, 1 = R32, 2 = R16, 3 = QF, 4 = semi-final, 5 = final.

The gap is modest but the direction is consistent. What is more interesting is the shape of individual runs. England used one formation — 4-2-3-1 — in all eight matches, and played three completely different games inside it.

Figure 7The eight quarter-finalists, match by match

wondrewlost

🇪🇸 Spain

우승

in order — opponent · score

조별 v 🇨🇻 CPV0-0 (D)

조별 v 🇸🇦 KSA4-0 (W)

조별 v 🇺🇾 URU1-0 (W)

32강 v 🇦🇹 AUT3-0 (W)

16강 v 🇵🇹 POR1-0 (W)

8강 v 🇧🇪 BEL2-1 (W)

4강 v 🇫🇷 FRA2-0 (W)

결승 v 🇦🇷 ARG1-0 (W)

🇦🇷 Argentina

준우승

in order — opponent · score

조별 v 🇩🇿 ALG3-0 (W)

조별 v 🇦🇹 AUT2-0 (W)

조별 v 🇯🇴 JOR3-1 (W)

32강 v 🇨🇻 CPV3-2 (W)

16강 v 🇪🇬 EGY3-2 (W)

8강 v 🇨🇭 SUI3-1 (W)

4강 v 🏴󠁧󠁢󠁥󠁮󠁧󠁿 ENG2-1 (W)

결승 v 🇪🇸 ESP0-1 (L)

🏴󠁧󠁢󠁥󠁮󠁧󠁿 England

3위

in order — opponent · score

조별 v 🇭🇷 CRO4-2 (W)

조별 v 🇬🇭 GHA0-0 (D)

조별 v 🇵🇦 PAN2-0 (W)

32강 v 🇨🇩 COD2-1 (W)

16강 v 🇲🇽 MEX3-2 (W)

8강 v 🇳🇴 NOR2-1 (W)

4강 v 🇦🇷 ARG1-2 (L)

3-4위 v 🇫🇷 FRA6-4 (W)

🇫🇷 France

4위

in order — opponent · score

조별 v 🇸🇳 SEN3-1 (W)

조별 v 🇮🇶 IRQ3-0 (W)

조별 v 🇳🇴 NOR4-1 (W)

32강 v 🇸🇪 SWE3-0 (W)

16강 v 🇵🇾 PAR1-0 (W)

8강 v 🇲🇦 MAR2-0 (W)

4강 v 🇪🇸 ESP0-2 (L)

3-4위 v 🏴󠁧󠁢󠁥󠁮󠁧󠁿 ENG4-6 (L)

🇲🇦 Morocco

8강 탈락

in order — opponent · score

조별 v 🇧🇷 BRA1-1 (D)

조별 v 🏴󠁧󠁢󠁳󠁣󠁴󠁿 SCO1-0 (W)

조별 v 🇭🇹 HAI4-2 (W)

32강 v 🇳🇱 NED1-1 (D)

16강 v 🇨🇦 CAN3-0 (W)

8강 v 🇫🇷 FRA0-2 (L)

🇧🇪 Belgium

8강 탈락

in order — opponent · score

조별 v 🇪🇬 EGY1-1 (D)

조별 v 🇮🇷 IRN0-0 (D)

조별 v 🇳🇿 NZL5-1 (W)

32강 v 🇸🇳 SEN3-2 (W)

16강 v 🇺🇸 USA4-1 (W)

8강 v 🇪🇸 ESP1-2 (L)

🇳🇴 Norway

8강 탈락

in order — opponent · score

조별 v 🇮🇶 IRQ4-1 (W)

조별 v 🇸🇳 SEN3-2 (W)

조별 v 🇫🇷 FRA1-4 (L)

32강 v 🇨🇮 CIV2-1 (W)

16강 v 🇧🇷 BRA2-1 (W)

8강 v 🏴󠁧󠁢󠁥󠁮󠁧󠁿 ENG1-2 (L)

🇨🇭 Switzerland

8강 탈락

in order — opponent · score

조별 v 🇶🇦 QAT1-1 (D)

조별 v 🇧🇦 BIH4-1 (W)

조별 v 🇨🇦 CAN2-1 (W)

32강 v 🇩🇿 ALG2-0 (W)

16강 v 🇨🇴 COL0-0 (D)

8강 v 🇦🇷 ARG1-3 (L)

Productive possession — the ball and the threat.

Productive reactive — threat without the ball. The best win rate in the tournament (88.9%) lives here.

Barren possession — the ball, no threat. 31.2% win rate.

Barren reactive — neither. Winless in 15 knockout matches.

horizontal = possession % / vertical = xG created

Each side's matches joined in order, with the round marked inside the point. The point with a bold outline is the match that ended the side's run at the title. The badge is final placement. Spain stay in one corner; Argentina and England cross the plane.

Spain's eight points sit in one cluster, upper right; France are similarly tight. Argentina start lower left in productive reactive, move up and right, then fall back to the lower left in the final. England sit upper right, jump hard to the left against Mexico, and come back. The four sides eliminated in the quarter-finals — Morocco, Belgium, Norway, Switzerland — all have their final point below the 1.0 xG line.

Figure 8England — one formation, eight different games

Croatia

4–2 · 4-2-3-1

poss49%

low block21%

xG3.33

Ghana

0–0 · 4-2-3-1

poss72.4%

low block6%

xG1.78

Panama

2–0 · 4-2-3-1

poss57.7%

low block7%

xG1.61

R32W

Congo DR

2–1 · 4-2-3-1

poss53.5%

low block16%

xG1.75

R16W

Mexico

3–2 · 4-2-3-1

poss31.5%

low block49%

xG2.12

QFW

Norway

2–1 · 4-2-3-1

poss47.4%

low block30%

xG1.64

SFL

Argentina

1–2 · 4-2-3-1

poss35.2%

low block42%

xG0.46

3rdW

France

6–4 · 4-1-2-3

poss52%

low block29%

xG2.34

Seven of the eight were 4-2-3-1. Possession still ranges from 31.5% to 72.4% and low-block share from 6% to 49%. The one filled card — the semi-final — is the only match under 1.0 xG, and the only defeat.

Figure 9Argentina — starting reactive, finishing with the ball

GALG3–0 Wposs 44.6%low 33%xG 1.58–0.41productive reactive

GAUT2–0 Wposs 48.2%low 31%xG 2.45–0.63productive reactive

GJOR3–1 Wposs 67.6%low 19%xG 2.52–0.96productive possession

R32CPV3–2 Wposs 59.2%low 21%xG 2.73–0.52productive possession

R16EGY3–2 Wposs 57.1%low 13%xG 2.63–0.76productive possession

QFSUI3–1 Wposs 54.7%low 30%xG 1.99–0.44productive reactive

SFENG2–1 Wposs 56.5%low 25%xG 1.47–0.46productive possession

FESP0–1 Lposs 32.5%low 33%xG 0.07–2.52barren reactive

Their first two group matches were productive reactive, sitting in a 30%+ low block. They then moved to a possession game all the way to the final — where they were forced back into a reactive shape at 0.07 xG. Barren reactive, and their only defeat of the tournament.

Each of the two sides that reached the final fell into barren reactive exactly once, and that one match was their only defeat — Argentina against Spain in the final, England against Argentina in the semi. Spain never had a barren match in eight, and never lost.

Spain themselves are the counter-example to this section. They used only two models and had one of the smallest match-to-match possession swings in the field (a standard deviation of 6.65). Execute one model at a high enough level and you win without switching at all. Switching is not a requirement; it is the alternative available to teams that are not Spain.

Figure 10How starting shapes moved

G1G2G3R32R16QFSFF

Argentina7-0-1

4-3-34-1-3-24-1-2-34-4-24-1-3-24-1-3-24-4-24-4-2

England6-1-1

4-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-1-2-3

Spain7-1-0

4-1-2-34-1-2-34-1-2-34-1-2-34-1-2-34-2-3-14-1-2-34-1-2-3

France6-0-2

4-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-1

Switzerland3-2-1

4-3-34-3-34-4-24-2-3-14-1-2-34-2-3-1

Norway4-0-2

4-1-2-34-1-2-34-1-2-34-2-1-34-1-2-34-1-2-3

Belgium3-2-1

4-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-1

Morocco3-2-1

4-2-3-14-2-3-14-2-3-14-2-3-14-2-3-14-2-3-1

Paraguay1-2-2

4-4-24-2-3-15-3-24-4-25-4-1

Brazil3-1-1

4-4-24-3-34-3-34-1-2-34-1-2-3

USA3-0-2

4-2-3-13-5-24-1-2-33-5-23-5-2

Egypt1-3-1

4-2-3-14-2-3-14-2-3-14-4-24-2-3-1

Dark cells mark a change from the previous match; the right-hand number is how many distinct shapes the side started in. Argentina reached the final in four, France the semi-final in one.

04 / Defence

Defence is structure, not effort

To be productive reactive, the block has to win the ball back. So what makes a good block — one that runs a lot, or one that deletes options?

Splitting every team-match by how many pressures it spent to win one forced turnover produces the steepest gradient in the entire dataset.

Figure 11Pressures per forced turnover

Most efficientpressures / turnover 4.28

60.4%win rate · n=53avg xG 1.76

Efficientpressures / turnover 5.66

45.1%win rate · n=51avg xG 1.37

Inefficientpressures / turnover 6.8

38.5%win rate · n=52avg xG 1.18

Least efficientpressures / turnover 9.15

9.6%win rate · n=52avg xG 0.79

The most efficient quartile pressed just 192 times a match and won 60.4%. The least efficient pressed 332 times and won 9.6%. Team quality is folded in, but the direction is unambiguous.

Figure 12What the three blocks actually measured, in metres

High block/press

line height 48.5 mlength 36.3 mwidth 40.1 m

Mid block

line height 37.6 mlength 26 mwidth 39.5 m

Low block

line height 18.2 mlength 22.5 mwidth 35.4 m

Tournament-wide averages from FIFA's defensive line-height and team-length diagrams. Line height = the deepest defensive line's distance from the team's own goal-line. Team length = deepest to highest player.

From high block to low block, line height falls 48.5 → 37.6 → 18.2 m and team length compresses 36.3 → 26.0 → 22.5 m: sides drop and shorten at the same time. The wider a side's span between those two extremes, the better it did — the same team occupying a broader range of defensive postures within one match, which is structurally the same story as section 03.

But the measurements hid something more important. Split the 62 deep-block team-matches into the sides that won and the sides that lost, then measure the block itself, and the two groups are essentially indistinguishable.

Figure 13The shape of the block does not explain the result

스페인16.6 · 23.2 · 36 mmean xG 2.18

파라과이17.4 · 20.8 · 34.2 mmean xG 0.4

깊은 블록 62경기	Won	Lost
라인 높이 (m)	17.3	18.6
팀 길이 (m)	20.5	19.8
팀 폭 (m)	34.5	35
생성 xG	1.54	0.57

Left: Spain's low block (champions) and Paraguay's (lowest xG side of the tournament) overlaid at the same scale. Right: the 62 deep-block team-matches split by result.

Line height 17.3 m against 18.6 m, team length 20.5 against 19.8, team width 34.5 against 35.0 — all three within 1.3 metres. In those same 62 matches xG created was 1.54 against 0.57, a 2.7-fold gap. Across the full tournament the three geometry measures correlate with winning between −0.13 and +0.17; xG correlates at +0.57.

The champions and the lowest-scoring side of the tournament sat in almost exactly the same shape. Where the block was set, and how tightly, did not separate winners from losers — only what came out of it did. This is the argument of the whole report restated in metres.

Figure 14Span of defensive posture, and results

Narrowesthigh–low span 26.81 m

21.2%win rate · n=52avg xG 0.94

Narrowhigh–low span 29.48 m

40.9%win rate · n=66avg xG 1.26

Widehigh–low span 31.38 m

44.4%win rate · n=45avg xG 1.4

Widesthigh–low span 34.33 m

48.9%win rate · n=45avg xG 1.57

05 / Where I was wrong

Did the longer match break the deep block? No.

One premise of my hypothesis was that matches now run past 100 minutes, giving sides more time to break a reactive model down. The data does not support it.

	Matches	Deeper side advanced	Rate	Goals/match	km per team	Goals per 90
regulation	23	9	39.1%	3	112	3
extra time	9	3	33.3%	2.67	143.9	2

In the nine knockout ties that went to extra time, the deeper-sitting side advanced 33.3% of the time — lower than the 39.1% in the 23 ties settled inside 90 minutes. And extra-time matches produced 2.00 goals per 90 against 3.00 in regulation. The extra half hour did not open the game up.

This refutation is not decisive either

Matches go to extra time because they are deadlocked. They are low-scoring first and long second, not the other way round. The honest conclusion is that this dataset cannot test the claim. And the fact that the deeper side survived two of the four shootouts — Paraguay, Switzerland — suggests that extra time and penalties may be the reactive model's escape hatch rather than its undoing.

PARAGUAY — the purest reactive run in the tournament

GUSA1–4 Lposs 28.7%low 28%xG 0.6–1.88middle

GTUR1–0 Wposs 19.2%low 46%xG 0.36–2.13barren reactive

GAUS0–0 Dposs 42.2%low 21%xG 0.29–0.7middle

R32GER1–1 Dposs 23.7%low 50%xG 0.61–1.25barren reactive

R16FRA0–1 Lposs 22.5%low 50%xG 0.13–1.93barren reactive

Paraguay averaged 0.40 xG and 27.3% possession across five matches and knocked Germany out on penalties in the round of 32 — the only success the barren reactive model had all tournament. In the next round the same model lost 0-1 to France at 0.13 xG. The escape hatch exists, but it opens once.

06 / A second correction

Was the field really levelling up? Only half of it.

In a 48-team tournament I read the talent gap as having closed, with the failure to convert it down to game model. The data supports only the first half of that.

Team	P	D	L	GF	GA	xG	xGA	Low blk
CPV	4	3	1	4	5	0.64	1.89	42.8%
CUW	3	1	2	1	9	0.56	2.62	36.7%
HAI	3	0	3	2	8	0.54	1.48	14.3%
JOR	3	0	3	3	8	0.62	2.17	28.7%
NZL	3	1	2	4	10	0.83	2.65	24%
PAN	3	0	3	0	4	0.73	1.14	30.7%
UZB	3	0	3	2	11	0.57	1.77	32%

Total 22 team-matches — 0W 5D 17L · 0.64 xG created · 1.96 conceded

The seven debutant and smallest-federation sides did not win once in 22 matches (0W-5D-17L). Cabo Verde's three group-stage draws and their extra-time tie with Argentina were striking, but across those four matches they created 0.64 xG and conceded 1.89. Even against Argentina the xG ran 0.52 to 2.73. This is not a picture of the gap closing; it is a picture of barren reactive football being good enough to draw group-stage matches.

GROUP-STAGE WINNING MARGINS, 72 MATCHES

draw

1 goal

2 goal

3 goal

4 goal

5 goal

6 goal

The matches themselves were still tight: 38 of 72 group games (53%) finished within a goal, and 20 were drawn. Both facts hold at once. Games got closer; converting closeness into points still belonged to the sides that could create.

07 / Variation theory, scored

How much of my framework actually showed up

Five concepts, each with a quantitative proxy from the 104 reports and a qualitative case from FIFA's own Technical Study Group.

Variable🟢 well supported

Actions that multiply the defender's options converted into real threat.

Quant — top quartile for step-ins won 46.2% at 1.71 xG against 29.4% at 1.00 for the bottom. Shot-load spread: 50.0% and 1.95 xG at the top, 31.5% and 0.78 at the bottom.

Qual — Argentina's four-step frame around Messi: midfield rotation → wide overload → switch → forward penetration (FIFA TSG).

Team variable · Lure and Shock🟢 qualitative / 🟡 untested

Designed lures clearly existed. Aggregate data simply cannot see their timing.

Quant — a limit, not a result. A lure is a time structure inside a sequence; match totals have no time axis.

Qual — Mexico built the trap on deliberately conceding the first pass (TSG, Aron Winter). Japan pulled Sweden left, switched right, and scored through the centre that opened during the second switch.

Individual variable🟢 well supported

Repeated individual threat forced structural change from the opponent.

Quant — the four semi-finalists averaged 19.8 take-ons and 10.2 step-ins per match against 17.9 and 7.3 for the rest of the field.

Qual — Itō Hiroki's repeated advances forced Sweden's half-time switch from 1-5-2-3 to 1-5-3-2. Bellingham's 'magnet effect' (TSG).

Mutation🟢 supported — the central finding

Sides able to change model mid-tournament went further. The most direct support in this report.

Quant — the 13 teams that used three distinct models averaged round 2.31 and 1.78 points per game, against 1.68 and 1.41 for the 19 that used two. Argentina reached the final having started in four shapes.

Qual — Mexico alternated 1-4-4-2 diamond out of possession and 1-4-2-3-1 in possession on every transition (TSG).

Multi-positionality🟢 qualitative / 🔴 proxy failed

The phenomenon was obvious. My way of measuring it was wrong.

Quant — failed. I used defenders' share of ball progressions as the proxy and it ran backwards (top quartile won 23.1%). It was measuring blocked progression, not role fluidity.

Qual — goal kicks taken directly by the keeper fell from 91% in 2022 to 52% in 2026 (TSG).

Two-phase defence🟢 well supported

Pressing was structure, not volume.

Quant — win rate by pressures-per-turnover quartile: 60.4 → 45.1 → 38.5 → 9.6%. The most efficient group pressed the least.

Qual — counter-pressing was the defining group-stage trend and its platform was the '3+2 rest-defence' (TSG) — effectively my own 3-2 notation.

Figure 15Two quantitative traces of the variable

STEP-INS — stepping inside the opposition line

Mostper match 13.35

46.2%win rate · n=52avg xG 1.71

Moreper match 8.76

45.2%win rate · n=42avg xG 1.27

Fewerper match 6.52

37%win rate · n=46avg xG 1.2

Fewestper match 3.63

29.4%win rate · n=68avg xG 1

SHOT SPREAD — how evenly the shooting load was shared

Most spreadnorm. entropy 0.78

50%win rate · n=52avg xG 1.95

Spreadnorm. entropy 0.69

36.5%win rate · n=52avg xG 1.29

Concentratednorm. entropy 0.61

36%win rate · n=50avg xG 1.09

Most concentratednorm. entropy 0.45

31.5%win rate · n=54avg xG 0.78

08 / Honest accounting

Three of my proxies failed

Reporting only the measures that worked would be the easiest way to make this framework look better than it is.

Defenders' share of ball progressions → multi-positionalitytop quartile won 23.1% (bottom 53.8%)It measured blocked progression, not role fluidity. The more a side is pushed back, the more its centre-backs recycle.

Offer-type spread → variable (options)top quartile won 30.8% (bottom 35.2%)Dominant sides concentrate their off-ball offers. An even spread read closer to aimlessness than to variety.

Pass-volume entropy → network dispersionpicked the winner 38.0% of the time — it ran backwardsSpread of pass volume is not the betweenness dispersion I specified. Dominant teams pile passes onto deep nodes.

All three were “more variety is better” measures, and all three inverted once dominance entered the mix. Variety is not a virtue on its own — only variety carrying threat is. The sentence that runs through this whole report shows up once more here: barren variety produces nothing, for exactly the reason barren reactive football loses.

09 / Recovering a failure

Pass-network dispersion was wrong. The raw data underneath it was not.

Section 08 reported that pass-volume entropy picked the winner backwards. That measure was an approximation built from aggregate totals. Going back to the raw 17x17 pass matrices turned up a much stronger signal sitting right next to the one that failed.

Every player-pair connection in the final, for both sides, rebuilt from pass counts and ranked strongest to weakest. Spain's curve sits entirely above Argentina's — not because two players passed unusually often, but because the whole team's web of connections ran stronger, top to bottom.

Figure 16The final — every connection, ranked strongest to weakest

🇪🇸 스페인🇦🇷 아르헨티나

Every player-pair connection for Spain (770 passes total, 22 connections at 10+) and Argentina (352 total, 8 at 10+). The dashed line marks the 10-pass threshold.

Figure 17The same data, by name

🇪🇸 스페인

top 10 connections

쿠바르시–포로60

포로–야말41

쿠바르시–로드리40

쿠바르시–라포르트39

라포르트–쿠쿠레야36

로드리–라포르트29

로드리–파비안27

로드리–포로26

쿠바르시–시몬23

라포르트–시몬22

🇦🇷 아르헨티나

top 10 connections

로메로–L.마르티네스13

로메로–데파울13

L.마르티네스–E.마르티네스13

파레데스–오타멘디11

로메로–엔소11

엔소–메시11

엔소–데파울11

파레데스–몰리나10

L.마르티네스–오타멘디9

메시–몰리나9

Each side's top 10 connections. Black bars mark 10+ passes. Spain's strongest link (Cubarsí–Fermín, 60 passes) is 4.6 times Argentina's strongest (13).

Extended to all 104 matches: the number of 10+-pass connections a team has (e10) alone picks the winner 80.3% of the time (61/76). The 5-pass version (e5) gets 69.6%. The dispersion measures dropped in section 08 — entropy, betweenness — sat in the low 50s, effectively a coin flip.

Figure 18More strong connections, more wins

Fewestmean e10 4.9

15.9%win rate · n=63avg xG 0.78

Fewermean e10 11.1

31.2%win rate · n=48avg xG 1.17

Moremean e10 15.8

50%win rate · n=54avg xG 1.45

Mostmean e10 21.9

65.1%win rate · n=43avg xG 1.9

All 208 team-matches, split into quartiles by e10 (count of 10+-pass connections). Density was the signal, not dispersion.

This is why the failure in section 08 was left in, not quietly dropped. Discarding a badly designed measure is not the same as discarding the phenomenon it was pointing at. The pass network was never signal-free — it was just being measured the wrong way.

10 / The champion

Spain — the side that never fell into either trap

Eight matches, seven wins and a draw, one goal conceded, zero barren matches.

GCPV0–0 Dposs 65.7%low 2%xG 2.26–0.13productive possession

GKSA4–0 Wposs 63.2%low 12%xG 3.2–0.11productive possession

GURU1–0 Wposs 57.1%low 17%xG 1.12–0.22productive possession

R32AUT3–0 Wposs 59.2%low 9%xG 2.25–0.25productive possession

R16POR1–0 Wposs 50.6%low 21%xG 1.72–0.91productive possession

QFBEL2–1 Wposs 61.6%low 10%xG 2.2–0.34productive possession

SFFRA2–0 Wposs 45.8%low 20%xG 2.21–0.48middle

FARG1–0 Wposs 60.2%low 8%xG 2.52–0.07productive possession

The most important line here is not the final but the semi. Against France, Spain had 45.8% of the ball — their lowest of the tournament — won the xG 2.21 to 0.48, and won the match 2-0. The champions' least possession-heavy match was one of their most complete performances.

In the final Argentina created 0.07 xG — the lowest figure of the entire knockout stage and the second lowest of all 208 team-matches. Spain recorded 22 defensive line breaks and 251 receptions in the final third; Argentina managed 5 and 52. That is the gap a 1-0 concealed. Three of the six lowest xG performances of the tournament belong to sides that were playing Spain: Saudi Arabia at 0.11, Cabo Verde at 0.13, and Argentina in the final.

THE FOUR SEMI-FINALISTS vs THE REST (per team-match)

Metric	Semi-finalists	Rest
Possession %	52.72	44.25
xG created	2.02	1.14
xG conceded	0.85	1.35
Defensive line breaks	14.41	9.24
Receptions in final third	172.53	113.53
Pressures per turnover	5.96	6.56
High press share %	7.03	4.71
Low block share %	18.66	23.99
Step-ins	10.16	7.3
High–low span (m)	31.44	30.06

Spain's eight matches: CPV 0-0 · KSA 4-0 · URU 1-0 · AUT 3-0 · POR 1-0 · BEL 2-1 · FRA 2-0 · ARG 1-0

11 / Conclusions

Five readings I take forward

What loses is not the reactive model — it is the barren one

Split the 62 team-matches that sat in a 30%+ low block by the xG they created and the win rate separates 4.8 → 12.0 → 14.3 → 88.9%. The nine that combined a deep block with 1.5+ xG went 8-1-0 — a better record than any other group in the tournament. Barren reactive, meanwhile, won none of its 15 knockout matches. The problem was never the height of the block; it was whether the block produced threat.

Barren possession loses just as surely

The sixteen team-matches with a majority of the ball and under 1.0 xG won just 31.2%; tighten the cut to 0.8 xG and the twelve that remain went 1W-2D-9L. The trap is symmetrical: no ball and no threat, or the ball and no threat. Both routes lead to an early exit.

The model to build is the switchable one

The 13 sides that moved between three or more in-match models averaged round 2.31; the 19 that stayed inside two averaged 1.68. England used the same 4-2-3-1 all tournament, playing productive possession in the group and then a 31.5%-possession productive reactive game against Mexico. Choosing the route by opponent was the common trait of the sides that went deepest.

Pressing is structure, not effort

The quartile that pressed hardest won least (9.6%). Pressures spent per turnover won produces the steepest win-rate gradient in the entire dataset.

The shape of the block explains nothing

Split the 62 deep-block team-matches by result and measure the block: line height, team length and team width all land within 1.3 metres of each other. xG created in those same matches was 1.54 against 0.57. Spain, the champions, and Paraguay, the lowest-xG side of the tournament, sat in almost identical low blocks — the argument restated in metres.

But the longer match did not break the deep block

In the nine knockout ties that went to extra time, the deeper-sitting side advanced 33.3% of the time — slightly worse than the 39.1% in ties settled inside 90 minutes. Extra-time matches are deadlocked matches by definition, so the selection bias is severe. The honest conclusion is that this dataset cannot test the claim that added time breaks reactive sides.

In practical terms it reduces to one sentence. What has to be avoided is not the low block or the ball but barrenness, and the target is a side that can produce quality chances through dominance or through the counter — and choose between them by opponent.

Open questions

The most important thing still open is timing. Lure and shock is a time structure inside a sequence, and match totals have no time axis. I know from the TSG's account that Mexico's conceded first pass was designed, but nothing here tells me how many repetitions it took to pay. For the same reason, model switching is visible match to match but not quarter to quarter.

The next step is clear: event-level logs, measuring the interval between line breaks and the time from break to final-third entry. At that point “depth of variables” stops being a metaphor and becomes a number.

Source — all 104 official FIFA Post-Match Summary Reports of the 2026 FIFA World Cup (group stage M1 through the final, M104). Qualitative evidence — FIFA Training Centre Technical Study Group round reviews and team analyses. Tournament totals: 308 goals, 1.28 average xG per team-match, 45.6% average possession, 10 defensive line breaks, 122 receptions in the final third.