Mirpur's 17 Percent: Where the BPL's Home Advantage Actually Comes From

One evening last season I sat in the press box at Mirpur's Sher-e-Bangla National Stadium, looking from the scoreboard to my laptop and back. The home side had made 142 in 19 overs. My shot-quality model said they had earned 158. The very next night in Chattogram, the away side scored 148; the model said 149. Same league, two matches, and the error was oceans apart.

That night I wrote a simple question at the top of my notebook: How much home advantage does the BPL actually carry? And inside that number, how much is crowd, how much is operations, and how much is plain arithmetic?

Answering it pushed me outside cricket — into German pressing data, into England's empty stadiums, and into a football model from nine years ago that taught me how hard it is to make a league see its own xG.

In 2026, aged 24, I joined Dhaka's Golpo Sports as a junior data analyst from my flat in Rajshahi. My first job was coding 1,248 shots from the football Bangladesh Premier League — who shot, from where, which foot, at what angle, under how much defensive pressure. I wrote a twelve-part series on shot quality. Abahani Limited Dhaka scored 34 goals from 27.6 xG; Sheikh Jamal Dhanmondi scored 29 from 31.2. The league's best side was finishing roughly 23 percent above league average, while the side beneath it was losing seven percent of the same work.

In Bangladesh, I taught a league to see its own xG — and after that I stopped writing the word "deserved." I wrote xG differential instead. Shot quality replaced possession in every match report, and a standard template settled in: xG, PPDA, distance covered.

In 2026 that series caught StatsBomb's eye, and I worked the Russia World Cup as a remote event-data analyst. In Germany versus Mexico, Germany's 26 shots produced just 1.3 xG; Mexico's 12 shots produced 1.1. Germany's PPDA was 6.9 — only 6.9 passes allowed per defensive action — and they conceded 18 transition chances in that chaos. I shipped the model before the final whistle. Germany finished bottom of the group.

PPDA showed me Germany, but the deeper lesson was different: a pressing metric does not describe how hurried your attack is; it describes how unsettled you are making the opponent. Back in cricket, the question became: how do I measure pressure in the powerplay?

One: Write the mapping rules first

In football, PPDA is how many passes an opponent completes per defensive action — lower means more pressure. Copying it into cricket produces nonsense, so I wrote my rules down first.

In my Powerplay Pressure Index (PPI), every delivery counts as a "pass," and a "defensive action" is any delivery that produced a wicket, an edge, a miss, or a clear mistiming. The window is the first six overs. So PPI = total powerplay deliveries ÷ false-shot-or-wicket events. I know the limits: dropped catches, run-outs and byes do not enter it, and our ball-tracking coverage is not complete across a season.

Across two seasons I coded 2,142 powerplay deliveries. League average PPI came out at 11.4. The best three bowling units sit between 8.2 and 9.1; the bottom three sit between 14.7 and 15.9. Teams holding a powerplay PPI under 9.5 won 68 percent of their matches in that phase.

I am suspicious of that number, and that suspicion is the next job: strong teams buy good bowlers, and good bowlers lower PPI. The correlation is tight; the causation is not yet unpacked.

Two: The Mirpur pitch — how true is the two-paced claim?

The polite line about Mirpur is low bounce, slow pace, two-paced surface. In my backup, expected runs per ball in the powerplay at Mirpur is 0.83, against 0.91 in Chattogram and 0.94 in Sylhet. The wicket probability runs the other way: 5.9 percent per ball at Mirpur, 4.6 percent on average elsewhere.

So runs fall at Mirpur, but the cause is uncertainty in bounce, not turn. In my coding, spinners averaged 2.4 degrees of turn in the powerplay there — only marginally above the national average. Then why do Mehidy Hasan Miraz, Rishad Hossain and Mahedi Hasan matter so much in the first six overs? Because they attack length, vary flight, and the batter — unable to time the shorter delivery — pre-selects the shot. The pressure comes from the flight of the bowling, not the spin of the wicket.

On the pace side, Taskin Ahmed's powerplay PPI is 9.4, Mustafizur Rahman's 10.1, Tanzim Hasan Sakib's 11.6. They share one trait: with the new ball their line lands in that unpleasant patch of the pitch where a batter's height never resolves normally.

Three: From 43.1 to 33.8 — what empty stands taught

In 2026, when world sport froze, I was consulting for Brentford FC. I analysed 306 behind-closed-doors matches across the Bundesliga, Championship and Serie A. Home win rate dropped from 43.1 percent to 33.8; home xG differential fell by 0.21; distance covered in the final 15 minutes fell 5.2 percent. From those numbers I built the CrowdNull adjustment. Brentford changed their set-piece routines with it, then won promotion that season.

Empty stadiums taught me that home advantage is a variable, not a law.

With that background I ran the same logic across 187 BPL matches from the 2026 to 2026 seasons. In raw terms, home teams won 56.4 percent. But once I controlled for toss outcome, start time of innings — day or night — rest days between matches and the umpire panel, the venue-linked edge collapses to just 1.1 percentage points.

In other words, the raw edge was 6.4 points and 1.1 survives: only 17 percent of raw home advantage can be written under the venue's or the crowd's name. The other 83 percent is distributed across scheduling, timing and the toss.

Four: Where the rest of it goes

At Mirpur on a night match, dew settles from the air, the ball gets wet, and spin grip fades in the second innings. In my data the chasing side wins 58.9 percent of night matches and 42.7 percent of day matches. Winning the toss and fielding is worth nearly twenty free points after dark. We mistake that for team skill.

Then there is the fixture list. Home teams get consecutive matches at their own venue and fewer travel days. Where the rest gap is two days shorter, late-innings effective pace drops by an average of 1.8 km/h. That is not courtesy; it is fatigue.

On top of that sit the curator and the practice wickets. A side training for two weeks on the main venue's practice strip keeps a much clearer defensive shot-selection — worth about 1.2 PPI points. A real slice of home advantage lives here, and it survives even with empty stands.

And there is scoring bias. At Mirpur, wides, no-balls and fielding-error entries are logged by scorers employed by the host association. The number looks small, but misclassifying two or three deliveries a match compounds across a season into decades of difference in a team's wide-conceded table. If nobody outside audits that data, a model is not clean information — it is a mirror of local habit.

Five: Where I can be wrong

The easy story is this: Mirpur spins, so the home side wins. My data does not support it. What works there is not condition but information — knowing which length on that surface breaks timing. That distinction matters commercially: investing in home players yields less; refusing to share information yields more.

But I wrote my model's limits down first, and that is the most important habit of all. Across 187 matches the 1.1-point edge emerges, yet the confidence interval is wide — the sample is small, and ball-tracking coverage has gaps in one match out of four. The two-venue comparison is stable; comparing Sylhet with Rajshahi is still risky.

I pre-register hypotheses before I touch data. Otherwise the counter-intuitive story becomes its own reward and the analysis never reaches a coach's desk. During my co-training work in the football BPL, one thing was plain: a model starts working when a coach uses it in daily decisions, like a mirror.

The takeaway: what to watch next round

Three numbers deserve attention next round: powerplay PPI, the toss-by-start-time pair, and venue-adjusted home-away splits. My request to selectors: before picking a player because he performed at home, strip out the toss and the innings timing and run the numbers again. Because a large part of what we call statistics is really scheduling.

In advanced analytics the hardest job is proving your own best discovery wrong. So my first task next season is not a new model — it is taking the scoring template away from the host side and putting it in a neutral counter. Build the pipeline first; the poetry comes later.