Phase-Adjusted Valuation: Why the IPL Auction Misprices Death-Overs Bowling

February 2026, Bengaluru

An eighteen-year-old Afghan leg-spinner sat on the auction board at a base price of fifty lakh rupees. That afternoon Sunrisers Hyderabad bid four crore for Rashid Khan, and nobody in the room looked up properly. By the end of the season he had returned seventeen wickets at an economy of 6.62. The senior scout who had dismissed the model as "spreadsheet cricket" came asking for the file.

I repeat this anecdote for a reason that has nothing to do with the model being right. The model was not right that day — the market was wrong. That distinction is the axis of this entire piece, and it is why the four-crore figure still looks to me less like a price and more like a measurement of missing information.

I phase-adjusted the auction board before anyone called the first bid. That sentence sounds like vanity, but behind it was ordinary labour: reading a four-crore bid through the arithmetic of over-blocks rather than through a spinner's reputation. What follows is an audit of that arithmetic — sample size, error bars, and a list of the places where it breaks.

How the board prices, and where the arithmetic gives up

Treating the IPL auction as a talent-discovery market is the most common error. It is a capital-allocation mechanism: each franchise holds a fixed purse, a salary cap, a first claim on some players through the right-to-match card, and a separate bucket for uncapped Indian players. The direct consequence of those four rules is that the price set in the room measures not a player's ability but the information available about that ability.

The model rested on three pillars, and each term needs defining here because the whole argument leans on them.

Powerplay strike rate — runs per ball in the first six overs, restricted to deliveries under fielding restrictions. Do not read it as pure batting ability; it is a function of field settings.

Death-over economy — runs conceded per over from the seventeenth to the twentieth. Its sample-size problem needs separate treatment, because it is the central technical issue of this piece.

Wicket equity — the contribution of a single wicket in any given over, translated into runs, to win probability at that moment. My desk coined the term because cricket has no standard name for it, and without a name the debate always stalls on "pressure" and "intent" — words that cannot be measured.

The third pillar was weighted by the auction purse. If two bowlers have identical raw output, the dearer one is the one whose role is scarcer in the team's purse structure. That weighting step is precisely where market sentiment and model output diverge.

The market prices reputation. That is not a moral failing but an inevitability of information economics. A scout in the room has three hundred deliveries of footage on one player and two thousand on his rival. The two cannot carry the same price even if the ability is identical.

The sample-size crisis of death-over economy

Over several seasons I have watched death overs with the scorebook open beside me — from Chinnaswamy's short straight boundaries to the sea breeze at Wankhede. Watching from the ground makes one thing obvious that television commentary never says: in the death overs a bowler's real test is not runs but options. How many distinct deliveries he owns, and how many of them his captain trusts.

That ability registers in economy in a way that keeps its sample tiny. Across a season a specialist death bowler sends down thirty to forty death overs. That is two hundred to two hundred and fifty deliveries. The standard deviation of economy on that many balls is large enough that a single bad over — twenty-five runs — drags a whole season's economy up by nearly a run per over.

The implication is blunt: the number on which crores are decided at auction is, on one season's sample, close to statistically undecided. That was my first argument: the baseline was not neutral; it was itself a position. A bowler with four seasons behind him has a sample that testifies on his behalf; an eighteen-year-old out of Associate cricket has no sample at all. The auction room calls that absence "risk" and demands a discount on it.

To size the gap, my desk ran a simple exercise: confidence intervals for death economy at a fixed sample size of two hundred balls. The result was that a difference between seven and nine runs per over, on a sample that small, cannot be claimed with confidence to be distinguishable from zero. The "cheap death bowler" and the "expensive death bowler" may be the same person in the data.

A wicket's value changes with the over

Another layer sits on top, and this is where the distance between the broadcast box and the data desk shows most plainly. A wicket is not equally valuable in every over.

A wicket in the powerplay resets a batting plan — new ball, fast runs, and the next batter arrives circumspect. Its value in win probability is moderate, because twenty-six overs remain and anything can happen in twenty-six overs. A wicket in the middle overs is the least valuable of the lot; run rate is already set there and a new batter settles in comfortably.

A wicket between the seventeenth and twentieth is the most expensive, because win probability moves on every ball. The dismissal of a set batter at that stage usually carries more weight than the six runs saved across that entire over.

The conclusion is that death-over economy — which the market treats as the cleanest measure — is in fact the most incomplete, because it ignores wicket equity outright. A bowler conceding eight an over who removes a set batter once every four games may well be worth more than one conceding seven.

Across ball-by-ball coding of several IPL seasons I have seen this pattern repeatedly: the most discussed death bowler is often not the one with the best economy but the one whose overs move win probability furthest, fastest — runs and wickets combined. Television commentary reports economy. It never reports wicket equity, because no easy graphic exists for it.

Where eight runs become twelve points of win probability

Here is the arithmetic I find most solid. Suppose a side needs forty-two from the last four overs. If the bowler concedes thirty-five, the seven required look routine. If he concedes forty-one, the one required looks nearly impossible. On the surface that is a gap of eight runs — the value of one good piece of fielding.

In the language of win probability it is not eight runs. Eight runs across the last four overs, in normal conditions, is worth twelve to fifteen percentage points of win probability. A one-run-per-over improvement in death bowling therefore moves the match outcome as much as a four-run-per-over improvement across the first four overs does not.

This is where the auction's price structure fails. The market pays most for powerplay batting, because its number is cleanest and its sample longest. It pays a middling price for death bowling, because its number is noisy.

That gap is what held Rashid Khan's bid at four crore. Nobody in the room said he was not a good bowler. The room said information about him was thin, so the price would be thin. By the end of the season the price of that information was seventeen wickets at 6.62.

The 2026 xG sheet: what does not transfer to cricket

In 2026 I ran a small desk for a multi-sport data firm at the Russia World Cup. I audited all sixty-four matches on xG and PPDA and produced an internal truth sheet. France beat Croatia 4-2 in the final, but the sheet recorded plainly that the win stood on 1.9 xG and a PPDA of 9.8, not on possession. A broadcaster called me "the xG girl" on air; I answered with a six-page methodology appendix.

That sheet taught me the most important caution in this piece. The xG and PPDA sheet did not explain the match; it questioned the match's story. That is the job of phase-adjusted valuation in cricket.

There is a trap here, and I fell into it myself. Football's composite metrics look sophisticated when imported into cricket and are wrong. Cricket is a discrete chain of twenty-over events in which every ball is a separate, bounded episode with no continuous flow. In football "pressure" is a spatial concept, measurable in passes played within thirty metres. Cricket has no spatial equivalent.

So I did not carry football-derived composites across. I built three cricket-native measures — phase-adjusted strike rate, wicket equity, and win probability from ball one — and I define each in-line the first time it appears. That is not over-explanation. In thirty-three years I have noticed that the people who look most impressive are often the ones who most need the definition.

The 2026 empty stadium and an audit of the crowd story

The 2026 IPL was played in the United Arab Emirates without crowds. For me it was a rare natural experiment: if the presence of a crowd manufactures "pressure", removing the crowd should remove the effect.

I coded all sixty matches. Chasing sides won thirty-four. The superficial conclusion is that pressure vanished in empty stadiums and chasing became easier. That conclusion was wrong, because I split it by venue. At Dubai's larger boundaries, death-over scoring fell 8.4 percent against 2026. The effect belonged to dimensions, not to crowds.

I presented the study to my firm's editorial team and refused to let any writer cite "empty-stadium nerves" without a venue split. Causal-caution notes became a habit: no performance claim without a control for venue, toss, and innings phase. My hiatus coverage became a template for separating real signal from broadcast mood.

The same issue applies to death-over economy. Four overs in Mumbai's sea air against four overs on Sharjah's short square are not the same column of data; pooling them is the same error as comparing four teams without a venue split. The board does not see the difference. The board sees one number.

Uncapped Indian bowlers: the largest evidence vacuum

The story told about India's bench strength is a story about talent. I think it is a selection problem, and a selection problem is fundamentally an audit problem.

Franchise scouting now does what zonal selectors once did. It carries the same defect: outside the knockout weeks of the Ranji season, ball-by-ball domestic data is not clean enough for the decision the auction demands. A bowler who produced two good knockout games has a sample on the auction board; a bowler who was quietly consistent all season does not.

The consequence is that the room prices available information, and the size of available information is set by whom the media showed. I do not chase the crowd; I audit what the crowd left in the data. Death-overs specialism and left-arm wrist spin are the two categories in India where the record has been left out most.

The model beat an information gap, not the market

Now the part I like least, because it is where my own idiom turns against me.

The Rashid Khan case is often rewritten into the false sentence that the phase-adjusted model understood he would be great. I do not accept that. The model did not predict his excellence; it said only that information about him was thin and the market was discounting that thinness far too heavily. It was a valuation of an information gap, not of ability.

Why does the distinction matter so much? Because if the model had been capturing ability, it would still work today. It does not. Afghan bowlers are in every league, every scout has four seasons of footage. The absence on which the discount rested is now present. The same model no longer pays.

My sharpest and most necessary claim is this: an edge whose source is an information gap is always temporary, and those who mistake it for insight end up empty-handed within a season.

The direct consequence is that I rebuild the model every season, because last season's model has become part of the market. Phase-adjusted valuation is no longer a secret; every franchise holds some version of that sheet. What remains is not model advantage but sample quality — who has the cleanest domestic data.

This is where auction-room proximity cautions me most. Franchise access, leaked purses and retention chatter are the currency of this beat, and proximity feels like sourcing. The day I can no longer keep reported fact separate from model output in the copy, every bit of my credibility goes with it.

The condition I should have written first

I now attach this to every death-over claim, and it applies to this entire piece: the sample is small. If next season's picture breaks the pattern — if the most expensive death bowlers return least and the cheapest return most — that will be a failure of this argument, not of my method. I will write it down. I will not defend it.

The field: before the next bid

The first thing to watch at the next auction is the ratio at which death-bowling specialism is priced against powerplay batting. Across recent cycles that ratio has leaned persistently toward the powerplay, while the win-probability arithmetic says the opposite. When the ratio turns, you will not be told. It will turn.

The question is how many boards will realise, by then, that they are still reading last season's file.