The Ghost Games File: Stress-Testing Cricket's Crowd Theory of Home Advantage

The Ghost Games File: Stress-Testing Cricket's Crowd Theory of Home Advantage

On 8 July 2026, at the Rose Bowl in Southampton, England's first innings against West Indies ended on 204. Their ten-year average first-innings score at that ground was 341. A single bad day cannot explain a gap that wide. I opened a blank spreadsheet in my room in Rangpur and named it The Ghost Games File. The question was simple: when the crowd disappears, how much home advantage survives — and what is the real source of whatever remains?

Context: We File Home Advantage in the Wrong Drawer

Cricket treats the phrase as self-explanatory, as if playing at home came with a six-run bonus. It is a composite variable with at least four moving parts: pitch familiarity, travel fatigue, umpiring bias, and crowd noise. The first three are structural; the fourth is psychological. The trouble is that in normal datasets all four are tangled together, so you cannot isolate a single cause.

I built my first xG model in a Rangpur bedroom in 2026, hand-logging every shot of France-Argentina 4-3. Cricket has no clean one-to-one xG equivalent — you forecast ball by ball, because a delivery's quality depends on the batter's ability, the scoreboard pressure and the fatigue of the finishing overs. But the logic transfers: a result tells you nothing unless it is weighed against a controlled change.

COVID bubbles delivered that change. Travel shrank, pitches were the same, squads were the same — only the crowd vanished. Popularly, a natural experiment. Rigorously, a confounded design, because two things changed at once. The second barely gets a mention. After 2026, the ICC made neutral umpires mandatory in Tests to curb home bias. In 2026, travel restrictions softened that rule and home umpires returned. If you look only at the crowd, you are watching half the film.

Core: A Pressure System Built From Data

I hand-assembled a scorecard notebook of roughly 740 matches across Tests, ODIs and T20Is from 2026 to 2026 — schedules, innings scores, toss decisions and, where possible, the boundary-to-dot ratio. This is scraped and partially upsampled data, not a live feed, and every claim here carries that caveat. Still, the trends are legible.

In Tests from 2026 to 2026, home teams won just above 50 per cent — a long-standing feature of the format, where draws and toss control tilt things inward. Across 2026-2026, the ghost-stadium window, that figure slid into the low forties, a drop of eight to ten points. The fall was smaller in ODIs and smallest in T20Is, because much of T20 home advantage comes from pitch profile, not the crowd. The shorter the format, the smaller the shock — the first sign that the crowd is the camera light, not the engine.

Three tracked signals tell separate stories.

First, early-innings aggression. I built a ratio of runs to balls faced for the top six overs and the last five. In empty stadiums, new-ball line and length looked more disciplined, plausibly because the fear of the boundary rope eased — dot-ball pressure is a social pressure. When the spectators leave, that social enforcement goes with them.

Second, run-rate trajectory. In ODI chases, the variance of death-over scoring (call it death-over entropy) fell. Chasing under empty stands is marginally harder to swing late, because a chasing batter needs the roar as much as the logic. The crowd here is a spur, not a tactic.

Third, the awkward one — the broken umpiring convention. When the neutral-umpire rule loosened in 2026, home officials returned. LBW and caught-behind success rates in those trials leaned slightly homeward — one or two percentage points, not dramatic, but directional. If we write only about the crowd, we are burying the umpire.

I ran a small independent-change model, combining rating-adjusted run rates with a dummy for whether the match had home spectators. The coefficient was negative, as expected, but weak — drowned in noise. The crowd effect is real, but it cannot alone account for home advantage; pitch and umpiring are at least equal partners.

Compare 2026, when home win rates were climbing, with 2026. A bowler entering a home ground in an empty stadium sledged differently. Sledging and eye contact are products of a crowd. A whole informal micro-system collapsed, and it plausibly shaped ball-by-ball calmness — perhaps the biggest, least measured variable in the set.

Contrarian Angle: Correlation Is Not Causation

Home win rates fell while the neutral-umpire convention broke. The lazy reading is that this proves a crowd effect. It does not. The ghost matches were played inside bio-secure bubbles — limited resources, broken routines, masks, testing, stress. Turn four keys at once and you will not know which one opened the door — only that it opened.

A second contrarian key is cultural. In Bangladesh we hear 'home advantage' and picture the Mirpur roar, not grass length. Yet in 2026, home advantage held steadier for sides whose T20 identity was fireproof — squads built on spin and structure, not on atmosphere. The crowd was the stage for that resource, not its source.

Third, I may be fitting the model to the wrong dataset. Tests, ODIs and T20Is are three different animals. Averaging them produces a truth that is true in none.

Fourth: perhaps home advantage did not shrink; its proxy changed. Tracking data can now measure control, dot-ball pressure, pattern-of-overs. Empty stadiums did not raise strike rates so much as shift the texture of control loss. Without a format-specific spine, no unified conversation about home advantage means anything.

Takeaway

The ghost-games file is not a closed box. Crowds have returned, but DRS, review culture and umpiring technology have reshaped the map. A model is a monastery: you enter with noise and leave with discipline. Watch the micro-stats of fielding restrictions next season, not just the league table. The real lesson of empty stadiums is procedural, not statistical: when four variables move at once, the honest answer sits quietly in front of you. The question was why home advantage leaves. The answer is that what we called home advantage was never one thing.