Rewriting the Home-Advantage Coefficient: A Data-Ledger Audit of Bangladesh Cricket

One match from last season still leaves a mark in my notebook. At the Sher-e-Bangla National Cricket Stadium in Mirpur, the home side won by seven wickets. But up to the 40th over, the home team's phase-adjusted run expectancy stood at 6.8 per over, against the visitors' 7.4. The win came from a cluster of three straight dot balls and a run-out in the final six overs, not from batting dominance. The scoreboard wrote home win; my ledger wrote home-variance. I opened the xG ledger in 2026; the 2026 World Cup wrote its own audit, and I carried the same discipline into cricket's run-expectancy. The Mirpur pitch has repeatedly corrected its own accounts in that ledger. So today's question is not vague but precise: how much home advantage does Bangladesh actually hold, and how much of it is durable rather than the noise of a small sample. Context: I have split every number in this piece into two explicit data windows. The first window runs from 2026 to 2026, matches played behind closed doors during the global hiatus, when Bangladesh also played home series without crowds. The second window runs from 2026 to now, after crowds returned in full or in part. For each match I log three core units: phase-adjusted strike rate (split into the first six overs, the middle nine, and the last five), run expectancy, and the home-advantage coefficient, meaning points per match at home minus points per match away. Without all three together, the story of home advantage stays incomplete. The Mirpur surface is slow, low and two-paced; once evening dew sets in, spinners lose grip in the second innings, so the toss matters more here than at other venues. The Sylhet International Stadium is comparatively batting-friendly, while Chattogram brings different wind and humidity, which means home is not a single thing even inside Bangladesh. Travel, logistics and settling-in time all belong in the coefficient too. I never force one venue's numbers onto another, because Bangladesh's conditions are not a copy of any global model; they are a distinct data environment. Core: The real picture emerges when phases are separated. In Mirpur, home spinners usually post lower economy in the middle overs, because the pitch begins to grip and batters need time to find their line. Yet on the same surface, home batters' strike rate in the last five overs is only marginally better than the opposition's, often roughly level. The home advantage is therefore generated mainly at the bowling end, not the batting end. That asymmetry is striking. In the empty-stadium window I saw the home-advantage coefficient fall clearly: bowlers' line-and-length discipline dropped, because reactions to a dropped catch or pressure change in a crowdless environment. Empty seats did not just change the noise; they rewrote the home-advantage coefficient. After crowds returned, the coefficient recovered partially but not fully, because squad structures had also changed. I log toss-win rate separately for every series, because dew in Mirpur strips spin grip in the second innings, and that shift shows plainly in the run-expectancy curve. The most useful part of this ledger is the venue-by-venue comparison. Mirpur's middle overs and Sylhet's middle overs are not the same; the same team against the same opponent produces different results at the two grounds. So I break home advantage into three parts rather than a single number: pitch-driven advantage (spin grip, bounce), environment-driven advantage (humidity, dew, light), and crowd-driven advantage (noise, pressure, fine umpiring margins). The first two are near-constant, the third fluctuates. In the 2026-21 window the third component nearly vanished, and that was the main reason the coefficient fell. Crowds have now returned, but by my count the coefficient has not returned to its old peak, because the first two components have also drifted: the pitch has become slightly more batting-friendly, and opponents' analysis departments have improved. As a Transfer Market Administrator, my job is not only the cricket audit but translating it into price. In franchise auctions and squad building I advise: do not look at a player's raw average, look at his phase-adjusted index at home. A spinner who holds pressure in the middle overs on Mirpur's slow pitch gains market value; a batter who scores only on Sylhet's batting deck must be valued carefully. This is where the gap between average and actual contribution becomes clear. One number has repeatedly served me: how well a player performs at home must always be read alongside his away numbers, or the squad is built on a false foundation. Contrarian: But my most urgent warning right now is to myself. Seeing a relationship between the home-advantage coefficient and win rate does not mean finding a cause. The opposition quality in Bangladesh's home series is often uneven: sometimes a strong side, sometimes a weak one. If the coefficient is computed without controlling for opposition strength, what looks like home advantage may actually be the effect of an easier opponent. The second trap is toss luck. Batting second on a spin-friendly Mirpur pitch is a disadvantage, and in a series-level sample that luck has a large effect. The third trap is small samples. Drawing durable conclusions from five or six matches of a home series is the easy road to regression to the mean. The fourth is selection bias: at home a team often chooses the wicket that suits it, so the pitch on which it won tends to be chosen again, and this loop inflates the coefficient on its own. Without stripping these away, treating the coefficient as a fixed truth is a mistake; it is a moving indicator, not an eternal law. So I divide every claim into three tiers: exploratory, gated, and audited. Only at the audited tier do I say it: in this series, at this venue, in this data window, after adjusting for opposition strength, a clear slice of home advantage remains. Below that tier I only show signals, not conclusions. This habit is what saves me from small-sample noise. My biggest lesson about Bangladesh's conditions is this: models here must be built in the language of local coaches, scorers and fans. Dropping in numbers from a foreign league directly produces error. So I fix phase definitions with local coaches, align innings splits with scorers, and attach an explicit data window to every claim, so anyone can verify it. That is the core principle of my ledger: if it is not reproducible, it is not published. Takeaway: In the next round I will watch one signal only: the home-advantage coefficient calculated separately for every venue and format, and its relationship with toss win-rate and opposition strength. If we forget the lesson learned from empty stadiums, we will again mistake the scoreboard for the ledger. The question is for you: in the next series, will you count home wins, or the home coefficient?