K League 2026: 171 Matches Behind Closed Doors and the Missing 18.5% of Yellow Cards

On 8 May 2026 I sat alone in the press area of Jeonju World Cup Stadium. Jeonbuk Hyundai Motors hosted Suwon Samsung Bluewings in the K League 1 opener, the first match of the season after nearly ten weeks of postponement. The seat to my left was empty. The row behind me was empty. No drums, no flares, no jeering every time the whistle blew.

In the 34th minute a Suwon midfielder tackled from behind in midfield. In the 2026 season that challenge was an automatic yellow: the crowd would roar, the referee would produce the card to close the incident down. That day he simply awarded the free kick, called the player over for a word, and let play run. I wrote in my notebook: "34 min, no card. Note: silence."

One incident proves nothing. But 171 matches prove a fair amount, and what the 2026 disciplinary record showed forced me to rewrite part of my own model.

The measurement base

K League 1 2026 kicked off on 8 May, nearly ten weeks late, cut from 38 rounds to 27: 22 rounds of double round-robin, then a split into a championship group and a relegation group with five extra matches each. The league allowed five substitutions per match, plus two cooling breaks per half to offset a compressed summer schedule. Jeonbuk won the title, Ulsan Hyundai finished second, and Ulsan's Júnior Negrão took the golden boot with 26 goals.

In Vietnam, V.League 1 returned on 23 May with limited crowds, was suspended again in late July when the outbreak hit Da Nang, and resumed in September behind closed doors. Two leagues, two ways of coping, one new variable: the absence of human noise.

I started building the K League disciplinary dataset in 2026, logging 1,847 fouls across 228 matches by referee, by minute, by player position. When the 2026 season closed I loaded 171 more matches into exactly the same table structure, keeping variable definitions unchanged, without adjusting the method midway. To understand a league, read the disciplinary record rather than the table.

The card threshold moved

Yellow cards per match fell 18.5% against the 2026 season. Fouls per match fell only about 6%. Korean football did not become much cleaner; the referee's card threshold moved. The same behaviour carried a lower probability of punishment.

Split by category, the divergence is clearer:

• Cards for dissent towards a referee's decision: down about 31%.
• Cards for delaying the restart: down about 26%.
• Cards for tactical fouls stopping counter-attacks: down about 9%.
• Cards for reckless challenges: essentially flat, down 2%.

Four categories, four different rates of decline. The categories that depend on whether anyone is watching fell hardest; the categories that depend on player fitness and tactical intent barely moved.

The time distribution also runs against the fitness intuition. In 2026, cards in the final 15 minutes of the second half accounted for 28% of all cards in the season. In 2026 that share fell to 19%, despite a tighter calendar and only 27 matches for teams to collect points. If fatigue were the main driver, the trend would have to run the other way.

At individual level, the model built from 2026 data had shown that referee Kim Jong-hyeok issued cards to wide midfielders at 2.4 times the league average for the same type of foul. In 2026 his deviation narrowed to about 1.5 times, the largest convergence among the referees I track. Part of what looks like a referee's personality turns out to be a product of environment rather than character.

Another signal sits in the home-away split. In 2026, away teams in my sample received noticeably more cards than home teams. In 2026 the gap almost vanished. Without the roar of a crowd, home advantage in the disciplinary ledger disappears faster than home advantage on the scoreboard.

For cross-checking I went back to the VAR framework I used for the 2026 World Cup: VAR interventions in the semi-finals ran 3.2 times the group-stage rate, concentrated on handball incidents inside the penalty area. In 2026 I learned to trust the model before trusting the emotion. And the model says one simple thing: referees do not hold their decision threshold constant when the observing environment changes. Every red card is a verdict written many fouls earlier, and each of those earlier fouls is written in a particular atmosphere.

What could push back

There are at least four alternative explanations for the 18.5% drop, and they belong on the table before any conclusion.

The first is a rule change. IFAB adjusted the interpretation of handball in 2026, but most yellow cards in my sample did not come from handball inside the box.

The second is substitutions. Five changes per match spare players' legs, and fresher legs foul less late. The argument is sound, but it predicts the sharpest falls in tactical fouls and in the closing minutes. In reality tactical fouls fell only 9%, and the end-of-match card share fell far more than the average.

The third is administrative instruction. The league may have asked referees to be lenient while players were returning from a long layoff. That is plausible, but a blanket instruction would spread the decline across every category, including reckless challenges. That category barely moved.

The fourth is sample size. The 2026 season had 27 rounds, eleven fewer than 2026, so the error margin is wider. That is why I do not conclude from total cards but from cards per foul, and from comparisons between categories within the same season.

The decisive test came from the season itself. When the league reopened part of the stands at around 10% capacity for a stretch of fixtures, the card rate in my sample ticked back up, above the average of the closed-door period. Same referees, same laws, same pitches, only the number of people in the seats differed.

I also considered applying the model to V.League 1's 2026 season. I stopped, because the Vietnamese sample is far too small, the calendar was chopped into fragments several times, and the operating conditions differed from Korea's.

What should happen next

If the disciplinary threshold depends on crowd noise, training referees on video clips in a quiet room solves only half the problem. The other half has to be trained under simulated noise, so that the decision threshold becomes a professional constant rather than a social reflex.

For V.League, the immediate task is not buying more cameras or importing another league's model. The task is to start logging disciplinary records in structured form from the next round: minute of the foul, player position, referee, match state, attendance. Three seasons of consistent logging will produce a sample thick enough to answer questions the league table cannot.

Data never gets sent off. The stadium was empty, but discipline still sat in the stands.