Mirpur's Surface, Dhaka's Noise, and the Auction Price: Taking Home Advantage Apart

Article begins

In a pre-auction preparation meeting last season, one number wedged itself into my sheet. A foreign leg-spinner, economy 7.1 across the post-powerplay seven overs in his last two seasons, went for roughly three times his base price. A domestic left-arm spinner with a 6.3 economy in the same phase went unsold. I said in the room that the buyers were not being stupid; they were buying a different product. Which slice of "home advantage" are we actually pricing — the surface, or the noise?

The dataset I compile is small: 69 matches across the Bangladesh Premier League and Gulf franchise leagues, 2026 to 2026, 412 spells, split by phase economy and strike rate. N is small enough that this is an observation, not a finding. That caveat is the first condition of the piece.

I built my first xG template in 2026, then learned to distrust its clean edges. The football spreadsheet taught me that a metric telling a smoothly rounded story deserves maximum suspicion. In cricket that suspicion sharpens, because a ten-over football press and a four-over spell are different animals. PPDA can measure pressure in football; in cricket you measure pressure ball by ball, through dot balls and field geometry.

The 2026 empty stadiums turned home advantage into a natural experiment. When the Bundesliga returned, home win rate had fallen from 43.3 percent to 33.3 percent and home teams' average xG had dropped 0.24. That piece taught me two things: never publish a claim without confidence intervals, and a natural experiment is never as clean as it looks.

In cricket the experiment is dirtier still. The 2026 IPL was played entirely in the UAE, where no team had a home venue at all. It gave us a benchmark of zero home advantage, not a test of crowd effects. Anyone copying the football design into cricket imports the flaw at the level of study construction.

So I decompose home advantage myself, across four layers.

Layer one: pitch and conditions. Mirpur is slow, low and spin-friendly. First-innings average in my sample is 142, second innings 128, a 14-run gap that runs nine runs wider than neutral venues. The trap: both teams play on the same surface. The edge appears when the home squad composition matches the pitch. Home teams bowl 43 percent of their overs through spin, away teams 31 percent. The advantage is not in the soil, it is in the squad-building decision.

Layer two: umpire decision bias. Home sides receive favourable outcomes on 54 percent of LBW and caught-behind reviews, away sides 46 percent. The 95 percent interval on that eight-point gap runs from minus one to plus seventeen points. The gap may be real; my evidence does not establish it. Anyone building a narrative on 54 versus 46 is showing a number, not a proof.

Layer three: toss and scheduling. At Mirpur, the toss-winning side has fielded first 68 percent of the time. Second-innings spin economy is 6.4 against 7.2 in the first. That cause is fused to the pitch and cannot be separated, so I give it no independent weight. Analyses treating toss as an independent variable hide an overlap problem inside the regression.

Layer four: travel and familiarity. This is the least measurable and the most expensive at auction. Home bowlers post a variation index 0.09 higher than away bowlers, but that could be talent, familiarity or simply selection. My data does not answer it. Silence in the stands did not erase home advantage; it split it into parts.

After the split, an unwelcome sum appears: crowd-and-fear explains 15 to 20 percent of the total effect in my model. The rest is surface, squad, schedule and selection. The noise is packaging, not product.

Why do scouts miss this? Because they measure different variables: how a bowler looks in a big match, how he behaves under pressure, what he is like in the dressing room. That is not useless information; it is precisely the information my dataset cannot hold. The half-analysed label starts there.

Model failure deserves its own paragraph. My phase composite weights economy at 40, dot-ball rate at 25, finisher matchups at 20 and fielding at 15. Swap economy to 25 and dot-ball rate to 40 and the top-five bowler list changes, on identical data with an identical N. Weights are a decision's comfort, not a truth. Until a weight survives a sensitivity test, the number is a claim under review, not a verdict.

Now the part where I argue against myself. Scouts see release under pressure, tolerance of sledging, the speed of return from injury. Those things do predict the future — they just do not predict averages. Calling that information worthless claims that what cannot be measured does not exist. That is a posture, not a proof.

So what is the auction really buying? In my 412-spell sample the strongest price predictor is not powerplay economy but highlight rate — the count of authoritative spells over two seasons. Its correlation with price is the steepest in the file; its correlation with next-season performance is the weakest. The market pays for highlights, not repeatability.

This is where loan deals matter. A franchise borrowing a star destroys the smaller side season after season: the smaller side does not develop the player, it rents one year of form, and the contract structure leaves the sell-on with the wealthier club. Small boards pay the development cost; rich franchises take the profit. That is a distorted incentive, not cooperation.

The second invisible cost is the calendar. Three leagues on three continents in one season, bilateral series wedged between them. Fixture congestion itself is the biggest injury culprit; no medical team saves a player from two games a week. We debate pitches and crowds while leaving the body's arithmetic unwritten.

One warning for my own conscience. Correlation is not causation. In a 69-match sample, the confidence interval on my own estimate runs from negative to positive. The honest answer is that I do not know the percentage. I know it is not one hundred.

— — —

If Mirpur's scoreboard tilts the same way next January, measure the pitch core first, then the openers' spin-over ratio, then the crowd. Change the order and the explanation changes. And if a franchise pays a premium for home advantage in the next auction, the question should be: are you buying the surface, or the shouting? The shouting never shows up on the scorecard.