Auction Price vs Process Price: When the Scoreline Sets the Market in T20 Cricket

In late November, the IPL mega auction came to a close in Jeddah. Before the first day's bidding war had run out of steam, ₹53.75 crore had already been spent on two batters alone — Rishabh Pant to Lucknow Super Giants for ₹27 crore, Shreyas Iyer to Punjab Kings for ₹26.75 crore. The next morning, from Melbourne, I opened my ball-by-ball dataset because one question would not leave me alone: what exactly is the auction table paying for?

I went through three seasons of domestic and international T20 data. For middle-order batters, I looked at the relationship between strike rate in the death overs (16-20) and shot-control percentage in the powerplay (1-6). The correlation came out close to zero. The attribute the market pays most heavily for has almost no linear relationship with the attribute that actually decides matches.

So let me state the verdict plainly: the transfer and auction market still prices the scoreline, not the process. This piece is about why that happens, where budgets get destroyed, and which signals will tell us the market is starting to shift.

Context: where I learned to price process

I began in an A-League xG thread, where nobody watched and the numbers were clean. The 2026 A-League Grand Final: Sydney FC 1-1 Melbourne Victory, Sydney winning 4-2 on penalties. I counted 14 shots to 8, and a 1.2 to 0.7 xG edge. In that thread I argued the shootout was not a lottery — Sydney's set-piece xG chain had been structurally superior all match. Four hundred shares followed, and a DM from a betting syndicate.

Then Russia 2026. Germany took twenty-six shots, built 2.4 xG, held 70 percent possession, and scored zero goals against South Korea. South Korea's PPDA was 8.4 against Germany's 11.8 — a slow, sterile press. After the 70th minute, Germany's xG per shot was 0.09. I called it possession without penetration.

In cricket I began pulling the same logic across. Just as xG combines shot quality and location into a goal probability, cricket can build expected runs (xR) and expected wickets (xW) per delivery. But here a warning matters: cricket and football have different structures, so the translation is not one-to-one. Football has possession as a continuous state; cricket has discrete deliveries, bowler-dependent matchups, and pitch behaviour. Footballers attack and defend; cricketers specialise.

What is public: ball-by-ball scorecards, run-rate curves, Hawk-Eye line-and-length, field placement scores. What is not: workload data, injury detail, biomechanics reports, contract structure, insurance clauses, internal franchise-player revenue splits. That invisible half is the single biggest source of mispricing.

Also worth remembering: cricket's transfer window is not football's open market. There are auctions, retentions, Right to Match cards, drafts, and tightening NOC controls. Three major calendars — IPL, SA20 and ILT20, plus the BBL — compress against each other. One player, many bidders.

Core: seven layers of mispricing

First, the scoreline is a biased estimator. In the 2026 World Cup final, England made 241 and New Zealand made 241. The match was decided on boundary countback, 26 to 17. Identical totals, entirely different processes. The scoreline is an unbiased estimator of outcome but a high-variance estimator of ability. Forty-five off 28 balls and 15 off six both look fine on a card, but the card will not tell you which is repeatable.

Second, phase leverage. Not every over weighs the same in T20. In my win-probability model, a death-over delivery carries roughly double the weight of a middle-over (7-15) delivery. Yet strike rate is stored in most heads as a single number. A strike rate of 140 in the middle overs means something entirely different at the death. A batter who makes 45 off 28 between overs seven and twelve, and one who makes 22 off six at the death — the market pays far more for the second, even though the first batter's task was never the harder one.

Third, pricing without a matchup model is blind. A left-arm wrist spinner's economy against left-handed top-order batters and against right-handed middle-order batters describes two different bowlers. On the auction table, he is one name, one base price. Last season I found powerplay-matchup and death-matchup splits for the same bowler diverging by as much as 3.2 runs per over. Building a squad without a matchup model means buying a bowler you may never use in his best phase.

Fourth, control percentage and false-shot rate are absent from the market. In football, a defender is not judged on tackle count alone, because few tackles can mean he solved the problem earlier with positioning. Cricket inverts this: bowlers are judged on wickets. The bowler hitting yorkers at the death, giving the batter nothing to swing at, records dots. The bowler spraying a bad ball that gets caught at long-on records a wicket. Process metrics capture that gap; the scoreline does not.

Fifth, the small-sample trap. Everyone looks at the last two or three months before an auction. A middle-order batter may get only 80 to 120 death-over balls in a season, sometimes fewer. At that sample size, the standard error on strike rate is so large that deciding who is better is close to statistically meaningless. I pre-commit to a rolling window (36 months, weighted) and a minimum-ball threshold, otherwise the model starts fitting itself to the match narrative.

Sixth, injury information is an asymmetric market. Reading a franchise medical update last year, it struck me the statement was written for shareholders, not for treatment. No side strain depth, just a "niggle". Workload data — balls bowled, travel, back-to-back matches over twelve months — should sit on the auction table. Without it, the market is buying a risk it cannot price. Medical confidentiality is right; secrecy and strategic opacity are not the same thing.

Seventh, cricket's own loan-with-obligation. A small franchise or board develops a young player, proves him over three years, and a larger side then takes him through a retention-rule gap. The club that made him wins no trophy and receives no transfer fee. This structure keeps smaller sides permanently producing half-finished products for giants, with the risk parked on the small side's balance sheet.

Contrarian angle: maybe the market is not wrong, it is pricing something else

Here I want to argue against my own model. Suppose the auction price is not a price for cricket ability at all. Suppose it is a price for an asset — shirt sales, gate revenue, broadcaster pull, social reach, sponsor interest. Then calling it a mispricing is the wrong frame. The market is pricing correctly; I am measuring the wrong thing.

The second objection cuts deeper. Forecasting with process data means trusting your own model, and that is the biggest trap of all. Germany's 26 shots and 2.4 xG taught me not to trust scorelines. But if that lesson runs in the opposite direction, I arrive at a state where no result matters anymore — variance-first nihilism, where every match becomes sample noise. That is not analysis. That is surrender.

Third: sometimes the outcome really does happen. On October 23, 2026, I sat in the stands at the Melbourne Cricket Ground for India versus Pakistan. Chasing 160, Virat Kohli made 82 off 53 and produced something no model had in its expected range. Occasionally a player's skill simply outruns the model. Call it improbable; do not call it noise.

Fourth, context. When the Bundesliga restarted in 2026, Borussia Dortmund beat Schalke 4-0. Across the first 45 empty-stadium matches, home teams won only 33 percent and averaged 1.2 points, down from 1.6 with crowds. I built a Crowd Absence Adjustment. Cricket's equivalents are pitch, humidity, travel, rest and tournament pressure. But more variables do not make a better model. More parameters mean overfitting. My rule: a new variable enters only if it reduces forecast error in both the training and the hold-out window.

Fifth, a deeper discomfort — the indices themselves are franchise-dependent, not independent. When a broadcaster's graphics carry a team's strike-rate data, the variable selection leans in that team's favour. Without a public ledger of process, every side writes its own player's story.

Takeaway: what I will watch next window

Three things this cycle. One, whether any franchise voluntarily publishes workload and rest data — if so, a new price class is forming. Two, whether auction prices drift toward powerplay control and matchup splits — say, a side paying up for a bowler with a low wicket count but a high death-phase xW. Three, whether smaller franchises build any collective structure to defend themselves under retention rules.

The question is not simple. If the market is genuinely pricing an asset while I measure ability, then my whole model is an exercise in doing correct arithmetic in the wrong market. And if the reverse is true — if the market only ever sees the shadow of the scoreline — then next season's best buys will be the least-discussed names. Which side would you back?