Who Delivers When It Matters
A pressure-adjusted rating of T20 batters and bowlers, built from 200,961 deliveries across 867 matches. Pressure is measured from match state, not from the clock — and every player is compared against the opposition they actually faced.
Top 10 under pressure
Ranked by pressure rating after adjustment for opposition, phase, season and venue. Only players who appeared in an IPL 2026 playing XI are included.
What counts as pressure
A 0–100 index computed for every delivery from four match-state signals. The mean rises through the innings, but the spread within a single over is almost as large as the spread between overs — so the index tracks the state of the game, not just the clock.
Pressure index by over
Mean with interquartile band, and the 70th-percentile threshold.
What drives it
The four components, each normalised to 0–1.
Where pressure concentrates
Share of deliveries classified as high-pressure, by phase and innings.
Pressure performance against ordinary performance
Each dot is an eligible player. Above the dashed line means better under pressure than in ordinary situations. The top-ranked players sit in the upper-right region — strong in both conditions, not specialists in one.
Batting
Runs above expected per ball. Higher is better on both axes.
Bowling
Runs conceded above expected per ball. Lower is better on both axes.
A finding that changed the model
Ranking on uplift — how much better a player is under pressure than in ordinary situations — is the intuitive design, and it fails. Out of sample it correlates with future pressure performance at −0.06 for batting, against +0.43 for pressure level alone. Uplift is a difference of two noisy estimates, so it rewards players with weak baselines. It is shown in the tables as a diagnostic and is not allowed to drive the ranking.
How the rating is built
Every choice below was tested against out-of-sample prediction rather than assumed.
1Pressure index
Four match-state signals combined into a 0–100 score per delivery.
2Win probability
A gradient-boosted model on 199,029 deliveries with a known result. Tension is 4·p·(1−p), weighted by innings progress so an early coin-flip is not equated with a late one.
3Adjusted plus-minus
Ridge regression with batter, bowler, phase, season and venue effects fitted together, so each player is measured against the opposition they actually faced.
4Empirical-Bayes blend
The pressure estimate is blended with the player's ordinary-situation record, weighted by measured split-half reliability.
Retention status: data not available
The IPL 2027 retention deadline is 17 November 2026 and the mini-auction is 16 December 2026. No franchise has published a retained or released list, so retention could not be used as a filter and was not guessed. These rankings answer the question the brief asks — who performs best under pressure — filtered for players demonstrably active in the IPL. They are not a list of who will be in the auction pool.
Availability screen
A performance ranking is not an auction shortlist. Three tiers were applied; only Tier C is submitted, and Tiers A and B are published in full so nothing is hidden.
| Tier | Player | Overall rank | Reason |
|---|