The Fortress Illusion of Mirpur: How Much of 'Home Advantage' Is Actually Home?

At the Sher-e-Bangla National Cricket Stadium in Mirpur that evening, much of the gallery stood empty. A BPL league match: the home side had batted first and put up 178 in twenty overs. During the interval, the reporter next to me said, "Home ground, the wicket will take spin, easy win." In my notebook I wrote one line: at this venue across the last three seasons, the first-innings average is 169 and the chasing side wins 54 per cent of the time. Which is to say the number was not on the home team's side at all — it was marginally on the side of the team batting second. The home side lost by 22 runs. Empty gallery, spin-friendly pitch, still a defeat. The easy story broke right there. And that breaking point is the actual job. CONTEXT: THE NUMBER EVERYONE CITES, NOBODY TESTS Home advantage is the most quoted and least audited figure in cricket. Globally, across formats, home teams have long won somewhere around half their matches, sometimes 55 to 60 per cent in particular competitions. The trouble is that this is an average, and an average never explains a cause; it only describes an outcome. When I first began hand-coding matches in Rangpur, I had exactly one question: is a venue's name the same object as the machinery inside that venue? The answer was no. I began with 44 matches, a Rangpur notebook, and a suspicion of easy numbers. In 2026 I hand-coded every match of the Bangladesh Premier League football season at Rangpur Stadium — shot location, pass direction, minute, outcome. The columns were four: event, location, time, context. When I moved to cricket the column structure did not change, only the language did — passes became runs, shot maps became line-and-length maps, minutes became overs. The first paid byline taught me that a model is only as honest as its assumptions. That lesson still forces a methodology footnote onto everything I file. WHAT I CODED, AND WHAT I DID NOT This piece rests on 297 matches. That includes matches from six BPL editions I watched in full, either in the ground or on broadcast; a slice of Dhaka Premier Division Cricket League games; National Cricket League first-class matches; and a large share of Bangladesh's home international T20Is and ODIs. For each match I coded the venue, the toss, the toss decision, the first-innings score, powerplay runs, the spin share of middle overs, death-over economy, wicket type (spin-friendly or pace-friendly), an estimated attendance, and the outcome of reviews and DRS decisions. What I did not code matters just as much. I have no ball-tracking data, no soil-moisture instruments, and no realistic way to measure an individual umpire's bias. So nothing here claims to prove causation. It claims direction of association and weight of evidence. Three assumptions stay open on the table: one, "home" does not mean geographic proximity but board-controlled advantage. Two, crowd attendance is a variable, but it is not independent of squad selection. Three, in small samples a gap of five to seven percentage points is larger than noise and smaller than meaning. CORE: HOME ADVANTAGE IS ACTUALLY FOUR SEPARATE MACHINES Cricket writers say "home advantage" as though it were one object, when internally it is at least four distinct processes. The first machine is pitch familiarity. A home bowler plays seven to nine matches a season on the same 22 yards. He knows when the surface goes low after the third innings, how much the new ball swings, how sharply spin turns after the tenth over. That is not data, it is memory — and memory walks straight into the coaching staff's notes. The second machine is travel and routine. Several BPL sides travel by road at night from Chattogram to Sylhet and play the following evening. In my coding, a side that had made a road journey longer than six hours showed a clearly reduced powerplay run rate in the next match. This is not mentality, it is the body. The third machine is the umpire and the decision environment. The fourth is the crowd. Beyond those four sits a fifth that nobody wants to call a machine: the power to prepare the pitch. The home board decides whether the grass stays and the surface favours pace, or the grass is shaved and the surface favours spin. That fifth machine can weigh more than the other four combined. This is where the clearest pattern in my coded data sits. In matches I flagged as spin-friendly, home spinners bowled at an economy around 6.8; the same bowlers on pace-friendly venues sat near 8.4. The difference does not come from skill, it comes from the surface and the match-up — the home side knows who to bring on, against whom, and when. MIRPUR'S NUMBERS, LAID OUT PLAINLY Mirpur is the heaviest venue in my dataset, and the largest sample. There, spinners delivered roughly 58 per cent of all overs; in Sylhet that share was nearer 46 per cent. Mirpur's first-innings average is 169 and the chasing win rate is 54 per cent. But that 54 is an average, and the average splits into two parts. In day matches at Mirpur, the second innings win rate in my coding falls below 50 per cent. In evening matches, especially once dew begins to settle after the second innings starts, the chasing win rate rises noticeably. Light, humidity and a ball that stops gripping — three physical events combine into a cricketing reality that we lazily call "the luck of the toss". Here is the important point. The relationship between winning the toss and winning the match is effectively zero in my data — toss winners took about 52 per cent of matches, close to the expected value of a coin. But what is decided at the toss does correlate with the result. The problem is not the toss; it is the decision after the toss. And both sides have the right to make that decision, so nobody deserves sympathy. EVERY "HOME" IS A DIFFERENT HOME The biggest error in writing about domestic cricket is putting every venue in one box. In my coding, Sylhet behaves like a different species — high-scoring, little help for seamers, hospitable to big hitting in the powerplay. Chattogram is sometimes batting-friendly and sometimes slow. Rangpur's ground is small with short boundaries — there, "home" means help with run-outs and a spinner's control of length. What follows is this: a side's home advantage can be positive, zero, or negative depending on the venue. A team that turns up with the same template across five different "homes" is not really at home anywhere. In some Khulna matches, the home sides in my coding had a lower powerplay run rate than they managed away — because the pitch was slow and the home batters could not adjust to it either. THE REVIEW ROOM: CONTROVERSY DID NOT SHRINK, IT MOVED One thing needs saying plainly, because it produces a great deal of my-side-versus-your-side argument. The idea that controversy declined once technology arrived is a comfortable fantasy. In reality controversy has simply walked off the field into the review room, from the umpire's finger to the grey zone of ball-tracking's "umpire's call". In my coded data a pattern shows up that is at least worth discussing. In editions where a review system was in operation, on-field umpires gave fewer marginal LBWs and instead left more decisions as not out. This is not a story of corruption, it is a story of incentives — the umpire knows the call will be checked, so he declines the risk. The result: the gap between the on-field decision and the final decision widens, and into that gap pours time, spectator patience and tactical delay. So when someone says technology has cleaned the game up, my question remains: cleaned up for whom? For the scorecard, or for the flow of play? AUCTION, RETAINERS AND A QUIET INEQUALITY In franchise cricket, home advantage is not built only on the pitch. It is built at the auction table. A side that can retain four core players on multi-year deals knows its spin-first plan will still function next season. A side that rebuilds its squad every year never gets the chance to understand its own home. Across the BPL matches I coded, a pattern held: sides with greater squad continuity — at least three bowlers together for two seasons — had better home death-over economy than the rest. That is not money, it is continuity. But continuity can be bought, and those who can buy it sometimes walk very close to the edge of the rules. One more thing belongs here, something the numbers never capture but the contract paper does: signing bonuses and retainers for the biggest names. Everyone sees and scrutinises the headline transfer fee; almost nobody sees the bonuses and privileges written inside the deal. And privilege is very often locked in there. CROWD VOLUME: WHAT THE EMPTY CHAIRS SAID Empty stadiums taught me that atmosphere is a variable, not decoration. In the seasons when crowds were minimal or entirely absent, three things fell together in my coded matches: the chasing win rate, the first-innings average, and confident aggression. How would a second-innings batter feel that? The scoreboard does not reveal it, because total runs do not fall evenly — they fall in specific phases: the three overs after the powerplay, and the first two overs of the death. With no mood to hear, a batter leans on his own arithmetic, and that arithmetic is a little slower. Yet part of my scepticism about the crowd lives right here. Where attendance is low, sponsor interest is low, squad rotation is higher, and broadcast scheduling is strange. The crowd variable never travels alone; it walks hand in hand with five others. In a small sample those five cannot be separated — and when someone separates them easily, I want to see their sample. A CHAIN OF NUMBERS, STRONGER THAN WORDS One more thing, because it is the lifeblood of my method. A single delivery's data no longer arrives through one system — ball-tracking, umpire's call, broadcast graphics, the scoreboard, all pass through separate suppliers. When that chain breaks, everyone says the same thing: "the data says so." Yet if a single joint in the chain is loose, the number itself shows no stain. That is exactly why, when I set hand-coded columns beside machine data, a mismatch is not a disaster to me but a gift. Every mismatch says something about an assumption, and every assumption deserves to be written on the face of the model. CONTRARIAN: HOME ADVANTAGE IS LARGELY A SELECTION EFFECT Now my least comfortable result. The most natural explanation is that home teams win because of the crowd. In my coded matches, that is not the largest variable. The largest is the selection effect. The home side knows what the wicket will do, so it picks for it; it has already adapted to the conditions. The difficulty is that this choice is often not a choice of talent but of environment. If a side plays two spinners at home and then plays the same two spinners away and they fail, the decision was pitch-specific. That decision, not the gallery, accounts for the five or six percentage points. The second angle is even less popular. In some cases home advantage is negative for a young local player. At home the expectation is heavier and an error costs more; in my coded data, home young middle-order batters had a slightly better death-over strike rate away from home. The size is small, the direction is consistent. The third and most important caveat: in domestic cricket my sample is small and structurally biased. Broadcast matches dominate my columns; matches played out of sight may carry the hand of a coach or an umpiring board. So an uncomfortable truth is worth facing — most cricket data writing is broadcast data, and broadcast never shows the cricket outside its own cameras. TAKEAWAY: WHAT I WILL TRACK NEXT SEASON Next season I will track three things. First, the gap between the first-innings and second-innings average at each venue, because the honest measure of home advantage is phase-level over data, not win rate. Second, what percentage of marginal LBWs on-field umpires give, and how often those decisions are overturned in the review room. Third, squad continuity counts — how many bowlers have held their place at the same franchise for two seasons. Home advantage will not disappear. But the more charming it looks, the greater the responsibility on those of us who write with numbers. Next time someone says "home ground, easy win," I will ask: which home? Which innings? Whose pitch? And how many people were actually there?