World Cup 2026: Home Advantage Is Mispriced, and the 3.5 Points From an Empty Season

On 16 May 2026, I sat in front of a screen in Seoul at 20:30, waiting for the Ruhr derby. Signal Iduna Park holds 81,365 seats. That night, 213 people were inside: two starting elevens, coaching staff, referees, medical crews and a small group of accredited reporters. Borussia Dortmund beat Schalke 04 by 4-0.

The scoreline was not what kept me awake. By the 60th minute, the only sound coming from the stands was the public address system. The South Stand, where 25,000 people normally stand for the full 90 minutes every weekend, was covered in yellow and black tarpaulin. Sitting in a small flat in Sinchon, I could hear Dortmund defenders calling to each other, the ball thudding against shin pads, the referee's short whistle. With no crowd, I could hear the match breathing.

Ten days later, I began pulling the data for all 306 Bundesliga matches of the 2026-20 season plus the nine seasons before it. The first result forced me to rerun the calculation three times because I assumed I had entered the formula wrong.

Across the 81 matches after the Bundesliga restarted, the home win rate was 37.8 percent. The equivalent rate over the previous nine seasons was 41.3 percent. Average home expected goals (xG) per match fell by 0.28. Yellow cards shown to away teams dropped by roughly 0.4 per match. Penalties awarded to home teams fell by nearly half against the baseline.

I presented the report to the head of analysis at Sports Data Lab in Seoul. He read it, put the paper down and said exactly one sentence: eighty-one matches prove nothing yet. I did not argue. I spent four weeks calling analysts, supporters and representatives of several betting companies, then ran an open online workshop with 150 participants called "Football Data Without Crowds." The people in that workshop gave me the ten years of historical data I was missing and pointed out three confounding variables I had overlooked. The model was subsequently applied across the entire 2026-21 season.

In 2026, the story returns at a far larger scale. The 2026 World Cup runs from 11 June to 19 July, co-hosted by the United States, Canada and Mexico. The tournament expands to 48 teams, 104 matches, 12 groups and 16 host cities spread across multiple time zones. The opening match takes place at Estadio Azteca in Mexico City, which sits at 2,240 metres above sea level. The final is at MetLife Stadium in New Jersey.

For the first time in the tournament's history, three host nations are in the group stage without going through qualifying. It is a natural experiment that data analysts have waited years for, because it sets two components of home advantage against each other on the same pitch. Across the 22 World Cups already played, hosts have won six times, and only two hosts have been eliminated in the group stage: South Africa in 2026 and Qatar in 2026.

Three variables, not one

What the betting market prices as a single block called home advantage is in fact three things stacked on top of each other. The first is social pressure from the stands on refereeing decisions. The second is travel and recovery cost. The third is familiarity with the pitch, the dressing room, the hotel, mealtimes and daily rhythm.

The empty-stadium season of 2026 removed the first variable and weakened the third, but left the second untouched. Bayern Munich still boarded a bus to away grounds, still slept in hotels, still ate dinner later than usual. So what the empty-stadium data measured is not the power of the crowd, but the portion of advantage that disappears when social pressure on officials is removed. Three and a half percentage points is the size of that portion, and it is much smaller than the stands imagine.

Research by Dohmen and co-authors published in 2026 on Bundesliga data showed referees tended to add more stoppage time when the home team was trailing, and that tendency vanished in matches played at neutral venues. The 2026 empty-stadium season accidentally repeated that experiment league-wide. Away-team yellow cards fell, home-team penalties fell, and disputed decisions inside the box fell at almost exactly the rate the model predicted.

2026 reverses the setup. The stands will be full. The United States, Mexico and Canada will not lack crowds, scalped tickets or festival atmosphere. But away teams have never had to travel this much. Vancouver to Mexico City is nearly 4,900 kilometres. A squad based in Toronto may have to fly to Los Angeles or Houston after only four days of rest. With 104 matches in roughly 39 days, the schedule is denser than any previous World Cup.

The new format also changes the incentive structure. In the 2026 group stage, twelve groups of four send their top two plus the eight best third-placed teams into the knockout rounds. The share of teams advancing rises from 50 percent to about 66.7 percent. The safety threshold drops, and that usually pushes teams toward lower-risk football in the final group match. For a modeller, this matters more than any transfer.

One more detail deserves attention. FIFA is using dynamic ticket pricing for the 2026 World Cup, meaning prices rise with real-time demand. Technically, that is a mechanism for converting supporter emotion into a forecastable revenue line. From a data perspective it creates a new variable: crowds at group matches between lower-ranked teams may be thinner than expected, and attendance no longer correlates tightly with match importance the way it used to.

What 3.5 percentage points does not say

I have to state the limits of the very model I defended through the 2026-21 season. Eighty-one matches is a small sample. Its confidence interval is wide enough that another season could produce the opposite result without breaking any hypothesis.

The 2026-20 restart happened under conditions that cannot be reproduced: five substitutions per team, cooling breaks, players returning from a three-month pause, a schedule compressed into six weeks. Every one of those factors affects pressing intensity and duel volume, which means it affects the very metrics I used to draw conclusions. Correlation is not causation, and it took me a while to accept that my model was measuring a composite in which the crowd is only one part.

On the supporters' side, the feeling of home advantage is real. I understand that. On 27 June 2026 at Kazan Arena, South Korea beat Germany 2-0 with goals in the 90+3rd minute from Kim Young-gwon and the 90+6th from Son Heung-min. I wrote that South Korea's expected goals were only about 1.12 against Germany's 2.31, that the home side held under 40 percent possession, and that the win came from fifteen minutes of late pressing. My inbox filled with angry messages. The Seoul night of 2026 taught me that the truth can be lonely, but never wrong. That lesson applies unchanged to the home-advantage story of 2026.

So when someone tells me a host nation will go deep this year because they are playing on home soil, I want to split that sentence into three separate parts before answering. Is the benefit coming from the stands, from not travelling, or from familiarity with the venue? The three answers can differ completely, and only one of them is systemic.

I also do not want to repeat the old mistake in the opposite direction: taking the 2026 empty-stadium data and simply subtracting 3.5 percentage points from every host match at the 2026 World Cup. A World Cup crowd is not a Bundesliga crowd. Travel intensity in North America is not travel intensity in Germany. The 2,240 metres of altitude in Mexico City is a variable that has never appeared in my model. Data does not shout, it whispers — and I have learned to lean in and listen.

Signals to track

Across the first sixteen group-stage matches, watch the cards issued to away teams and the penalties awarded to home teams. If both return to the levels of the nine Bundesliga seasons before 2026, the crowd has genuinely returned to its old role and the social-pressure hypothesis is confirmed globally. If they recover only halfway, most of home advantage sits in flight legs and sleep hours, not in singing.

Next, record the xG of the three host nations in matches against lower-rated opponents. The xG differential in that subset is the cleanest available measure of the portion of advantage that cannot be attributed to squad quality.

And keep an eye on matches in Mexico City and Guadalajara. Altitude changes ball trajectory, changes how quickly players recover and changes when teams choose to accelerate. That is a physical variable, not a psychological one, and it does not care whether the stands are full or empty.

Before you trust a metric, ask where it was born.

I am not stopping you from placing a bet — I only want you to understand what you are betting on. A host nation at the 2026 World Cup is not handed a fixed advantage in the scoreline. It is handed a conditional advantage, and those conditions shift by city, by time zone, by matchday. The job of anyone working with data is to break that advantage into small parts, measure each part, and admit to the part still unmeasured.