Toss, Dew and the Gallery: Which Layer of Home Advantage Actually Works on Bangladeshi Pitches

In the west gallery of the Sylhet International Cricket Stadium, the number I typed into my laptop has left no trace on the scorecard. An evening match; in the first ten overs the home side's spinners went at 4.2 runs per over. Dew arrived after the seventeenth over; across the next five, the same bowlers went at 9.8. Nobody changed the wrist position, nobody changed the action, but the ball had stopped gripping. The crowd was still there, same drums, same clapping. The match had already become a different sport. What I kept turning over on the way out of the ground is the real origin of this piece. The phrase 'home advantage' in cricket covers at least three different things under one name. Crowd pressure is one. Pitch and conditions are another. Market pricing is the third. Without separating them, any model measures the wrong object. The most dangerous model is not the one that measures the wrong thing; it is the one that does so confidently and files every error under 'variance'. I built the xG Chapel in Sylhet to measure belief, not to worship it. This piece is an attempt to keep the chapel door open. My method has been slow and monotonous from the start. After joining PitchData in 2026, I put my broadcasting background to exactly one use: tagging footage hour after hour. I hand-logged 3,800 Premier League shots to build my first xG model. That model refused to accept Burnley's seventh-place finish in 2026-18 as sustainable: 39 goals against 32.4 xG, and a 78.4 percent save rate against an expected 71.2 percent. The market ignored it. I tracked 12 matches and published a regression warning, and the following season Burnley won once in 12. Then, when the stadiums emptied in 2026, home advantage became a variable I could finally isolate. Across 92 Bundesliga matches played in empty grounds, home goals per match fell from 1.54 to 1.18 and the home win rate dropped from 43 percent to 33 percent. That could have been coincidence, but an adjusted model over 60 bets returned 8.4 percent ROI, and I published a technical note titled 'The Empty Stadium Is Not Neutral'. Nobody has run that experiment properly in cricket, because cricket has more variables and every one of them is wired to four others. Still, cricket needs layer separation more than football does, because in cricket almost the entire mass of home advantage sits at the venue layer. In football the crowd has a direct channel into the referee's decisions, the players' adrenaline, the visiting side's passing rhythm. In cricket the crowd's only direct channel is umpiring. LBW, caught behind, wides and over-rate pressure: those four are where a crowd's weight enters a human brain. Since DRS arrived that channel has narrowed considerably; much of the home LBW benefit of the 2010s now exits through the review system. So the crowd argument in cricket does not stand on the same foundation it does in football. The real work is done by the pitch, the air and the moisture. At Mirpur the ball stays below the height of the bat, which turns Mustafizur Rahman's cutter and Mehidy Hasan Miraz's off-spin into separate weapons in the same match. Chattogram is slower still, where a small window opens for spinners in the first 15 overs and then shuts. Sylhet is easier for batting, but as evening falls the dew overturns every calculation. Those are three different sports, yet the market calls all three 'home conditions'. On dew in evening matches my own ledger shows a pattern that people often dress up as an anti-spin story. The actual mechanism is plain: once the ball is wet, the seamers lose grip, the spinners lose revolutions, and the ball comes on truer. That makes chasing easier for the side batting second. Knowing at which over dew settles in an evening match changes the value of two decisions, the toss and the eleven. If teams fixed their spin quota around a venue-specific dew time, half the selection debates in a tournament would never be born. The second layer of home advantage shows up in selection. I treat every selection rumour as a time series with a confidence interval. Having an experienced all-rounder like Shakib Al Hasan in the side is not only a batting-bowling balance question; it is a question of how the bowling quota gets divided. Taskin Ahmed's new-ball spell and Litton Das's work behind the stumps, taken as separate variables, show that the captain's decisions depend heavily on that same dew window for which team analysts hold no data at all. The third layer is the market. The crowd is not noise; it is a hidden parameter the market keeps mispricing, but in cricket that parameter is far smaller than in football, and what replaces it is conditions. Buying a home favourite at even money in an evening match where dew settles in the second innings means buying the wrong variable. My CrowdNull adjustment was built for football; in cricket I use a simpler adjustment called MoistureNull, which takes only three inputs, the over when dew begins, the spin-economy differential, and the ratio of boundary-bound shots in the second innings. This is where my deepest doubt sits. We split home advantage neatly into crowd and conditions and build a linear story, even though the variables are collinear. The curator is local, the schedule is built to suit, travel asymmetry persists, and DRS has only existed for a few years. When variables are that entangled, causation cannot be separated. I never called the Croatia system bet a prophecy; it was a stress test of my priors. Croatia had 1.6 xG to England's 0.9, but England pressed harder, a PPDA of 8.2 against 11.4, and the lighter press conserved energy for extra time. That is exactly why I label which layer I am making a claim about. My rule for catching my own errors is public. If a venue-layer dew model fails to beat the base rate at a stated confidence interval across the next ten evening matches, I shelve that layer. Scorched earth. One more thing deserves stating: home advantage is contracting. Neutral curators, more travel-friendly logistics, better sports science, competitive tracking systems. Franchise leagues are also buying young talent from smaller leagues as satellite assets, which blurs what 'your own ground' even means for a young cricketer. The ground where he trained at fourteen may next host him in the opposition's shirt, on a hired contract. What builds a model is that discomfort. The model does not care about your narrative; that is why I feed it first. Next round I will be watching two things: at which over the ball first starts to skid in an evening match, and whether the decision to play a second spinner is driven by condition data or by social-media reaction. When dew time moves, the price of the match moves. Not when the name of the gallery changes. That is what gets measured next.