BPL Transfer Window: Million-Taka Prices, Zero Database

On the night of the last auction, a name went up on the screen in a Chattogram hotel conference room. The base price was 3 million taka. Thirteen minutes later the figure beside that name read 18 million. The announcer's voice was celebratory: "A proven finisher. Trust him in the death overs." I went home, opened my laptop, and pulled up the BPL dataset I had hand-coded. The strike rate of that category of batter in overs 17–20 was 118.4 — forty-four points below the league average. Eighteen million and 118.4. Put those two numbers side by side and the picture that emerges is not a story about auction night. It is a story about measurement.

I hand-coded BPL ball-by-ball data in 2026, sitting in a Chattogram startup, watching twenty-four matches twice each. There was no API, no shortcut — just ninety minutes of keystrokes and a kind of monkhood. That work taught me something I still carry: a decision without verified data behind it is not analysis — it is a guess, and when a guess carries a crore-taka price tag, nobody counts the risk.

Context: Where the money grows and the measurement does not

BPL's franchise economy has grown dramatically over the past decade. Sponsors, broadcast deals, auction pools — the flow of money now ranks among the largest in mid-tier franchise cricket. But the measurement system standing beside that economy has not grown at the same pace. Where franchises in The Hundred or the IPL run their own scouting databases, venue-specific models and opposition-analysis pipelines, most of our franchises still build decisions on three things: cable-television commentary, newspaper headlines, and a memory of what someone looked like last season.

The auction mechanism itself amplifies the problem. Retention, direct signing and the auction are three routes, and retention and direct signing tend to re-purchase an old valuation, because there is less pressure to analyse afresh. So a wrong valuation does not end in one season; it sits inside the squad's structure for three or four seasons, and the opportunity cost is never accounted for.

There is a structural issue here. BPL matches are played mainly at two venues — Sher-e-Bangla in Mirpur and Zahur Ahmed Chowdhury in Chattogram. The two wickets behave fundamentally differently. In Mirpur the ball comes slowly, spinners control the middle overs, and even a score below 140 often wins. In Chattogram the ball comes nicely onto the bat and boundaries are easier in the slog overs. Yet at the auction table this venue difference is almost never priced in. A batter's career T20 strike rate is the one number every franchise looks at; which venue suits him, and in which phase his shot selection holds up, is a layer nobody enters.

Core analysis: The three numbers that set the price, and the ones that should

In my dataset I split batting impact into four phases — powerplay (1–6), middle (7–12), build-up (13–16) and death (17–20). Over 1,200 boundary events, more than 600 ball-facing records, and a context state for every innings (wickets lost, required rate). Without that split, the problem you get translates directly into auction error.

Take a batter with an overall strike rate of 134. Looks fine. Split by phase, and his powerplay strike rate is 148 while his death strike rate is 109. In other words, the very phase the franchise wants to buy him for — death-overs finisher — is his weakest. That pattern was not rare in my dataset. In the BPL, the "finisher" tag is often the shadow of a batter's powerplay utility, not evidence of death-overs utility.

The second number that never enters the price is dot-ball pressure. A batter may strike at 135, but if 42 percent of the balls he faces are dots, his partner absorbs extra pressure every over — and what the team loses in the last five overs never shows up in the batter's scorecard. In my coding sheet I kept a separate column: "partner-cost" — how much run-rate pressure this batter's dots created. In BPL auction thinking, that column does not exist.

The third number is the venue split. For a spinner, the economy difference between Mirpur and Chattogram in my dataset ranged from 1.3 to 2.1 runs per over. A left-arm spinner like Taijul Islam is far more central in Mirpur than in Chattogram; equally, the value of Mustafizur Rahman's cutter-based death bowling shifts with venue and with the age of the ball. A bowler described as "good all-round" is often a match-winner at one venue and match-neutral at the other. Had a franchise built its squad around its home venue, the same money would have bought far more match impact.

In bowling there is another layer we almost never see at auction — the matchup. Keep a leg-spinner's economy against left-handers separate from his economy against right-handers and the gap often carries a stronger signal than the overall economy. But matchup data requires ball-by-ball batter hand, line, length and shot type — that is, more hand-coding. Whoever refuses that work forfeits the edge.

I once tracked one franchise's squad-building — five seasons of buys and sells, 47 player acquisitions in total. Of the batters paid the most, only three spent two consecutive seasons in the top five for team win-probability-added. The relationship between price and sustained match impact is weak.

The counter-intuitive angle: Correlation is not causation

Now to the place where my analytical scepticism speaks loudest.

The auction argument is usually this: "He played well last season, so buying him will make the team better." Two separate claims get merged — a player played well (individual performance), and he can win matches for a team (team impact). There is a relationship between them, but it is not causal — it is situational. A death bowler is doing well because his team's fielding setup is good, catches are being held, and he is bowling at a venue that helps his cutter and slower ball. If a franchise buys only the number and not the situation, the number falls the next season — and the player gets the blame.

Treating correlation as causation is the most expensive mistake in the BPL auction market. Because it is measurable in money — several crore per season, several badly built squads per cycle.

One more thing I have noticed that gets very little discussion: the age curve. In T20, batters usually peak between 27 and 31, bowlers between 25 and 29. But the BPL auction carries an experience premium — a 33- or 34-year-old sometimes earns more than a young player because "he has played the big matches." The dressing-room value of experience like Mushfiqur Rahim's or Mahmudullah Riyad's is not in dispute — but that value is not the same as per-ball impact on the field, and in the auction bag the two get priced as one. In my dataset the decline in fielding impact and powerplay mobility past 33 was clear, yet it was not reflected in price.

For young players like Towhid Hridoy or Nahid Rana the opposite happens — their market value still sits below their actual capacity, because the market does not trust small samples. That is BPL's biggest inefficiency: experience is overpriced and emerging talent underpriced, and neither price rests on verified data.

I once told a junior analyst, "A model that gives no decision is a diary, not a weapon." In the BPL case we do not have a model — but what we want is not a model. We want at least one verified indicator set that lets us see the gap between price and impact with our own eyes.

Takeaway: What to watch next window

Next auction season, when a franchise again buys a batter at six times his base price as a "proven finisher," ask three questions — and look for the answers in the data, not the announcement.

First: what is his death-overs strike rate, and which side of the league average does it sit? Second: what is his dot-ball percentage, and in which phase did those dots fall? Third: what is his economy or strike rate at your home venue, and what is it away?

If a franchise can keep those three answers in its own database — just three, no more — the money story at the BPL auction table changes. The question is not whether our cricket lacks talent; the question is that we measure talent with an instrument that can only read headlines.

I still hand-code BPL ball-by-ball data, because there is still no API, no shortcut — just ninety minutes of keystrokes and a kind of monkhood. The franchise that first understands that the fruit of this monkhood is in fact its cheapest player will write the next decade of the BPL.