The BPL Draft Ledger: Wage Bills, Retention Logic, and a Dataset I Coded by Hand

On the night before the 2026 BPL player draft I sat with a spreadsheet holding four seasons of events I had typed by hand. One name caught me. A death-overs specialist, released by a franchise the previous season, had conceded at 7.9 an over across his last 18 innings — well below that season's league average of 9.1. Yet the draft placed him in category B, base price 3 million taka. The very next row listed another bowler, death-overs economy 10.4, base price 5 million, category A. Put the two numbers side by side and the question asks itself: does the auction price performance, or does it price a name and a few highlight sixes? I switched off the screen that night and thought it through. Almost everything written about the BPL is about runs, strike rates and two or three sixes. But the decisions that actually build a franchise are made with the wage bill, the retention arithmetic and bench depth. That is the layer I want to examine here, using a dataset I coded by hand. In a league with no public API, knowing where every number comes from is half the analysis. I coded the Bangladesh Premier League by hand before I trusted its numbers. The BPL draft is not the English football transfer window, but the economics rhyme. There are three routes — retention, direct signing and the draft — and all three are haggling inside a budget ceiling. A franchise must stay inside a fixed wage bill for the whole season. Which means whoever you retain at a high price is taking away your freedom to build the rest of the squad. Retention is not just keeping a player; it is an opportunity cost. When I joined MatchLab as a junior analyst in 2026, there was no standardised strike-rate database for the BPL. I watched 24 matches of a season twice each, tagging shots, boundaries, dot balls and death-overs spells ball by ball. No API, no shortcut, just ninety minutes of keystrokes and a monk's patience. That does not make the data perfect; it means I know where every number came from. That transparency is the real asset here. Once you understand the wage bill, an odd picture of the field clears up. A side that retains three senior batters at high prices has less money for a death bowler and a finisher. What you see on the field is a heavy top order and a thin bench. That imbalance never shows on the front page of a scorecard. But look at the bottom half of the points table, and at the list of matches lost in the last five overs, and the arithmetic starts to speak. My dataset holds event-level records from 67 BPL matches between 2026 and 2026 — roughly three and a half thousand deliveries. Across those three seasons one pattern kept returning: franchises pour money into batting strike rate and far less into death-overs economy. Four of the five most expensive players are batters. Yet the sharpest link to match outcomes sits in the bowling of the last two overs. I ran a simple regression: powerplay strike rate correlates weakly with winning, death-overs economy correlates far more strongly. The reason is plain — matches are decided in the final overs, and that is where bowling is hardest. So why does the market walk the other way? Because batting numbers are visible. A batter's 150 strike rate is bright and loud. A bowler's 7.9 economy is dry, context-dependent, and takes more data to trust. A franchise that will not dig deep takes the easy road — it buys the highlight. Here sits a subtle trap. Strike rate is a volatile number. A strike rate of 160 across eight innings is not the same as 160 across forty. A middle-order batter's two good innings on a small sample plant a big number beside his name, with no guarantee it survives the next season. In my dataset, batters who struck above 145 in fewer than 15 innings lost, on average, 11 to 14 strike-rate points the following season. That is regression to the mean — the most ignored rule in the market. Bowling shows the reverse. Death-overs economy is a relatively stable skill once you control for context. A bowler who regularly sends down the 18th to 20th over builds his numbers across many innings, so the variance is lower. And precisely because of that, the market prices him cheaply — his numbers are not visible, and reading them takes deeper data. I ran one count. In the 2026 season, of the sides that kept their last-five-over economy under 8.5, three reached the playoffs. On the other side, of the sides whose death economy ran above 9.5, only one finished in the top half. That is not proof, but it is a signal — and at draft time a signal is the most valuable thing there is. Now caution is due. Seeing a link between death-overs economy and winning and concluding that good economy means victory would be a mistake. Correlation is not causation. If a side fields well, drops no catches and sets the right field, its bowler's economy looks good — but that work is not the bowler's alone. The reverse also holds: a superb death bowler in a poor fielding side can look ordinary. And there is the trap of sample size. If a bowler sends down 12 death-overs spells in one season, spending 30 million taka on the strength of those 12 is turning an incomplete number into a large decision. I am not saying economy lies; I am saying that reading a number without context is reading it wrong. A bowler's economy must always be filtered through the quality of the batter in front of him, the state of the pitch and the situation of the match. So I do not trust even my own model blindly. I keep one rule: a model that reaches no decision is a diary, not a weapon. In the BPL, a decision means a clean recommendation — this bowler can be kept at this price, this batter's price runs above his sample. At the next draft I will watch one thing: the sides that spend little on death-overs bowling and get much back are the ones truly exploiting the market's inefficiency. The franchise that understands now that batting highlights always cost more is the one that will win matches in the final over next season. So the question is not for draft night, but for the weeks before it — whose numbers are you trusting, and why?