The Number That Buys Cricketers Is the Least Stable One in the Room

Inside the closed auction room, seven franchise representatives sit in front of a screen with eight columns. Nobody in the room cares about the first seven: matches, overs, wickets, dot balls, fitness report, age, base price. The eighth column changes the volume of the conversation. Death-over economy. I was the data analyst in that room, so my question was simple. How many balls sit underneath that number, and what share of them were bowled to top-four batters? At the break I opened my own ball-by-ball log: 63 T20 matches, 2026 to 2026, domestic and franchise cricket combined. The compilation is mine, personal, and not an official database. One figure surfaced. For bowlers with fewer than 200 balls in the death role, season-on-season correlation of economy sits close to nothing: r ≈ 0.24, with a 95 percent confidence interval of 0.05 to 0.41. The room was buying noise with conviction. Talent was being purchased, and that part is right. What was missing was a grammar for the price. Asia's franchise circuit is now a connected market. The BPL, the Lanka Premier League, the Pakistan Super League, ILT20, SA20 and the IPL — players such as Shakib Al Hasan and Mustafizur Rahman wear more than one shirt in a calendar year, and that is the norm rather than the exception. The architecture rests on four pillars: an annual auction or draft, retention, direct signings, and the national board's no-objection certificate. In Bangladesh, a central contract and fitness clearance sit on top. The paperwork is sound. The problem is not in the paperwork, it is in the information. In the IPL, release point, spin revolution and pitch mapping are produced per ball and sold to partner franchises. Most of Asia's domestic leagues do not have that layer. What exists is the scorecard: runs, wickets, overs, economy. So the market can only price what is visible, and death-over economy is the most visible thing in the game, because every ball there is memorable — the six, the wide, the catch, the run-out. The calendar then adds load. In my log, bowlers who played three leagues plus bilateral cricket in one year showed a small but persistent decline in pace and line-length accuracy in the second half of the franchise season. That sample is thin, so I call it an observation, not a finding. But there is no workload column on the auction screen. There is an injury history. The largest risk in the room is the cheapest one on sale. There is a second, layered asymmetry. Smaller leagues give the debut, absorb the mistakes, make failure survivable. Bigger leagues buy the finished article. The same player is priced twice in two markets across three years: once as potential, once as proven. Nobody accounts for the link between the two markets, even though the benefit runs entirely one way. I would add three columns to the auction sheet. The first is adjusted death economy. Raw economy counts outcomes; the adjusted version weights every ball by the quality of the batter, the nature of the venue and the state of the match. In my log, two bowlers carry an identical raw death economy of 8.4. One bowled 62 percent of his deliveries to top-four batters; the other bowled 38 percent. After adjustment, they separate by 1.1 runs per over. Same price, different product. The second is a pressure-forcing index, the bowling equivalent of PPDA. Counting dot balls is not enough. The question is which dots forced a genuine decision from the batter and which were his own premeditated leave. In my log, a middle-overs spinner carries a high dot percentage, but more than half of those dots came from the batter's own calculation rather than the quality of the ball. Volume and control are not the same thing. The third is repeatability. What a franchise wants is the stability of an action, not the best night of a season. Domestic data cannot measure release point or the share of slower balls, so a proxy is needed: year-on-year stability of economy, performance across different venues, and the willingness to take the ball in the pressure over. The number earns its keep most clearly in regression to the mean. Bowlers in the top quartile of raw death economy in my log gave back roughly 0.9 runs per over the following season, drifting toward the league average. The bottom quartile moved the other way. That entire spread is also fully consistent with pure noise. When signal and noise cannot be separated, paying a price means buying a guess at a price. The batting sheet carries the same disease. An 18-ball 40 in the last six overs of a lost chase is not the same pressure as 14 balls for 30 with two wickets in hand. Leverage-adjusted strike rate captures that difference, and in my log, next-season repeatability correlates better with the adjusted figure than with raw strike rate. Franchises usually buy raw strike rate, because that is the number printed in large type. I built my first xG template in 2026, then learned to distrust its clean edges. Any composite — impact score, value index — produces a polished output, and that polish is the reason for suspicion. Two years ago I built a bowling score with four weights summing to one. A small change to a single weight rearranged nearly half of the top ten. A ranking that sensitive to its own weights cannot anchor a decision. It is a comment, not a measurement. Now the case for the scouts, stated properly. The people in that room are not fools. They watch elbow height in the warm-up, the rhythm of the run-up, the face of a bowler handed the 18th over. None of that appears on a scorecard. My log contains a small subsample in which scout ratings correlated better with next-season adjusted death economy than raw economy did. That sample is 31 bowlers, so I will not call it evidence. It is enough, though, to give the number permission to question my own model. The deeper problem is not the weights. It is the size of the market. Six or seven teams, thirty to forty-six matches a season, and perhaps one hundred to two hundred death balls per bowler per year. No market prices efficiently at that scale. That is a limit of arithmetic, not of will. The most honest decision available may be to decline to price at all where the sample cannot support a number. That caution is old for me. The empty stadiums of 2026 turned home advantage into a natural experiment. Across the first five rounds, home win rate fell from 43.3 percent to 33.3 percent and home teams' average xG dropped by 0.24. The picture only became clear after splitting the effect into four channels. Silence in the stands did not erase home advantage; it split it into parts. Pitch and conditions, umpire decision bias, toss and scheduling, travel and familiarity each hold a share, and misjudging the share ruins the conclusion. Bubbles, format changes and absent players are confounders that no regression on that dataset can wash out. I have written that down from the beginning. The same lesson applies to franchise bowling. At the Qatar World Cup, Morocco pressed selectively. That was the whole trick. Anyone reading only PPDA would have said they do not press, and would have measured volume rather than intent. Bowling works the same way. Control cannot be counted in dot balls; it has to be read in the trigger. In Bangladesh, nobody measures the trigger yet, because the instruments have not entered the domestic system. So what should you watch in the next auction cycle? The retention lists. Whether any franchise prices a three-season adjusted number instead of a single-season one. Whether a workload column finally finds space on the screen. The franchise that installs all three first may have a worse first season; by the second, the market will copy it. The question is simply this: can an auction room pay for one season of noise, or does it have the patience to wait for three seasons of signal?