The Mirpur Spreadsheet: Four Variables of Home Advantage and One Misread

In the 17th over of the second T20I at Mirpur's Sher-e-Bangla Stadium on December 14, Bangladesh needed 58 from 42 balls. The stands were full, the Dhaka dew had not yet fully arrived, and the ball was in spinners' hands. A number was burning red on my live-thread screen that the big scoreboard never shows: 16.4. That was the home batters' boundary percentage over the last four overs, roughly half of what the innings had produced earlier. Seven dot balls followed in the next two overs, and the chase stopped 11 runs short.

The scoreboard records runs. A spreadsheet records which bowler bowled which over, where the fielders moved, and how many deliveries a batter deliberately left. I begin with the live thread and end with a broadcast truth. This match forced me to reopen an old account, because the story forming from the first ball of the evening — Mirpur as a fortress where the crowd noise breaks visitors — shrank considerably once it was placed on a table.

Mirpur, Chattogram, the Sydney Cricket Ground and the Melbourne Cricket Ground: four venues, one template. The question is always the same — what is the home side actually benefiting from? Then the variables get written down. Pitch age and spin share, dew coefficient, toss and innings-choice tendencies, crowd density, the visiting side's travel-and-recovery index, and ambient humidity. I set the baseline against the last five years of home and away performance, apply venue-specific context coefficients, and state the finding last.

The sample is limited, and I say so upfront. Thirty-one men's T20Is for Bangladesh at Mirpur from January 2026 to December 2026, eleven at Chattogram, and twenty-seven Australian T20Is at Sydney and Melbourne for comparison. For the dew coefficient I used start and innings-break humidity, plus changes in seam movement from ball-tracking. Matches without tracking data stayed in a separate layer; I did not blend them in. Four venues, one scale — the framework travels, but it does not colonise the local truth.

The table that raised the loudest question was the direct first-innings against second-innings comparison. Over five years at Mirpur, chasing sides scored more slowly than sides batting first, and the reason is not finishing skill.

| Metric | First innings | Second innings | | Run rate (per over) | 7.62 | 7.04 | | Spin share (percent of balls) | 42.1 | 51.3 | | Dot balls (percent) | 38.4 | 44.1 | | Boundaries per 10 balls | 1.31 | 1.12 | | Sixes per innings | 5.8 | 4.1 |

Spin's share of deliveries rises by more than nine percentage points in the second innings, and the dot-ball rate climbs 5.7 points in almost exact proportion. That is a bowling-plan outcome, the physical consequence of a pitch changing. The Mirpur surface skids a little with the new ball for ten overs, then dries and begins to grip. Ball-tracking shows the grip index rising about 14 percent after the 12th over. Batters change their lines, bowlers push their lengths back, and the crowd only sees dots.

The dew coefficient offers the most interesting counter-evidence. Dhaka dew helps the ball come onto the bat, which favours the chasing side. Broken down by month, the second-innings run-rate gap flips.

| Month | Dew coefficient | Second-innings run-rate gap | | January | 0.8 | minus 0.21 | | March | 1.9 | plus 0.38 | | November | 1.4 | plus 0.22 | | December | 2.3 | plus 0.51 |

In December the chasing side scores half a run per over more than the side batting first. In January the picture reverses. Mirpur's home advantage is therefore not one number but a tug-of-war between two forces: pitch age pulls one way, dew the other. In months where dew dominates, bowling first after winning the toss is defensible. Where pitch age dominates, batting first means taking control of the match.

The third table is the crowd's. Many treat it as the biggest variable of all. Our numbers make it the smallest of the four.

| Window | Crowd | Home first-innings run rate | Visiting spinners' economy | | 2026 (full stands) | around 8,000 | 7.51 | 7.88 | | 2026 (zero or limited) | 0 to 1,000 | 7.29 | 7.41 | | 2026 to 2026 (full) | around 12,000 | 7.62 | 8.02 |

More spectators do lift the home run rate, but the swing sits between 7.29 and 7.62 — two to three tenths. Building a no-crowd coefficient for Australian football in 2026, I learned the same lesson I later carried into cricket. Empty seats taught me that home advantage is a variable, not a myth — but it is the variable that shouts the loudest.

Breaking down the variance makes the picture cleaner.

| Variable | Share of explained variance | | Pitch age and spin share | 0.21 | | Dew coefficient | 0.14 | | Visiting side travel and recovery | 0.11 | | Toss and innings choice | 0.09 | | Crowd density | 0.06 | | Unexplained | 0.39 |

Zero point three nine remains unexplained. That is my model's failure, and I do not hide it. It says a Mirpur result largely comes from the uncontrolled portion — a dropped catch, a run-out, one bowler's rhythm in one over. The spreadsheet remembers what the stadium forgets, but admitting what the spreadsheet does not know is the honest part of the job.

The real separation at Mirpur happens in the middle overs, seven to fifteen. Over five years, Bangladesh's frontline spinners have an economy of 6.42 in that phase; visiting spinners sit at 7.88. That gap of more than a run an over comes from two places — the speed at which the ball is turned, and the field setting. Home spinners can hold a slider-to-googly trajectory on this surface because they bowl on it all year. Visiting spinners cannot find their length in the first two matches and locate it by the third. In a three-match series, that delay is close to inevitable.

This is where the true engine of home advantage hides. Which pitch is used, which ball the match is played with, when play starts — all three decisions sit with the home board. A board cannot change the dew coefficient, but by moving the start time it can control dew's impact. It cannot change trajectory, but it can change the slider-to-googly ratio. This is not an advantage in talent. It is an advantage in context, and context is the one thing an analyst can reconstruct.

My experience says the framework carries into other sports. When the A-League returned to empty stadiums in 2026, I looked at 24 matches: home expected goals fell from 1.45 to 1.12 while away pressing intensity rose. I installed a no-crowd coefficient within 72 hours and adjusted set-piece routines. Apply the same method to cricket and the pitch's role grows while the crowd's shrinks. Two sports, two different ratios, one template.

Now the counter-question. How often has the decision to bowl first after winning the toss at Mirpur actually worked? In our count, toss-winning sides chose to field in 41 percent of matches, and those sides won only 47 percent of them. Sides that won the toss and batted first won 58 percent. That is not surprising, because those decisions were spread across mixed months. The real question is whether the difference belongs to the toss or the pitch. Split by month, in December and March, when the dew coefficient runs high, fielding sides won 61 percent. In January, batting first won 67 percent. The toss itself decides nothing; the month and the pitch write the decision in advance.

Broadcasters tell the crowd-noise story most often. Twelve thousand spectators surely pull something — about 0.06 of the variance in our model. More than five percent comes from ball grip and dew humidity. What is visible is numerically small; what is invisible settles the result. Correlation is not causation here. Teams win at home because the conditions were prepared in advance, and the crowd simply sits in the theatre that was arranged for them.

One of my own errors is willing to testify. On the live thread I initially read the collapse as a middle-order batting failure. Once the innings closed, ball-tracking showed the seam movement had not dropped and bounce had not changed, but the spin meter had begun climbing from the 12th over. My first hypothesis was wrong, and that kind of error happens whenever the live log is not kept separate from the final analysis. I timestamp the opening read, then revise it.

I also keep one trap open for myself. Add pitch age, spin share, dew, crowd, and at some point there is a risk of building a model that honestly produces whatever result I wanted. So I pre-register the variables before the season and do not touch them much in the second half of matches. Model output is never treated as ground truth; it is checked against ball-tracking, match reports and video before anything is written as a conclusion.

Three signals for the next round are on my board. If the spin-share differential in overs seven to fifteen exceeds 55 percent and the match is in December, bowling first after winning the toss is the statistically sound decision. On a fresh January or February pitch, with a dew coefficient under 1.2, batting first carries value. And if a visiting side's first two spinners bring their middle-over economy under seven, the equation at Mirpur could shift by the third match — that would be the series' real turning point.

The scoreboard stops when the match ends. The model keeps playing. An empty stadium loses a variable; a full one brings it back. But pitch age and the pull of dew repeat every season. So the question gets harder. Are we measuring home advantage, or have we named the right to prepare conditions at home as home advantage?