10 — Personal Analytics / Scouting · SupeRR Selector Season 2

Who delivers under pressure?

A Pressure Index on every ball of IPL and T20I cricket from 2022 to 2026, built on a win-probability model. Each player is judged against what the exact situation usually produces and against their own normal level, then the best pressure performers still available are turned into a Rajasthan Royals shortlist for the IPL 2027 auction.

DataIPL + T20I, 2022–26
Deliveries200,961
ModelWin prob · GBM
Players rated100 + 100
01 / Problem

Some players rise when the game is on the line. Can data tell which?

Rajasthan Royals’ SupeRR Selector challenge asked for the ten most reliable batters and bowlers under pressure, measured against their usual level, using IPL and T20I data from 2022 onwards. Pressure was left for each entrant to define.

Two traps make this hard. A raw “death-overs strike rate” rewards finishers for the situation, not the skill. And pressure samples are small, so most “clutch” numbers are luck. The model below is built to avoid both.

02 / Defining pressure

A 0–100 score on every ball.

▸ Pressure Index ◂
PI = 0.40 · stakes  // expected win-probability swing of the next ball
    + 0.20 · closeness  // near 50–50, weighted up late in the match
    + 0.20 · scoring_demand  // required rate, or runs below par
    + 0.20 · wicket_stress  // more wickets down than usual
    × 1.10 knockout · × 0.3 when the result is already decided

Stakes come from two gradient-boosted win-probability models trained on the data (chase AUC 0.86, first innings 0.69). For every ball, the six usual outcomes are played out and the average swing in win probability is the leverage. The top 25% of balls count as pressure.

The index separates real situations: from the lowest band to the highest, league strike rate falls from 146 to 128, dot balls rise from 35% to 41%, and each ball moves the result about 8 times as much.

03 / Rating players

Judged against the exact situation.

▸ Per ball ◂
runs_above_expected = (runs − expected_runs) − wicket_value · (out − expected_out)
// expected values by format × innings × phase × wickets down × pressure band × pace/spin
▸ Pressure Reliability Rating ◂
rating = 0.35 · pressure_value + 0.25 · win_prob_added + 0.20 · rise_vs_own_baseline + 0.20 · consistency

T20Is against associate nations count half, A-team games 0.7, and each season back counts 15% less. Eligibility: 150+ pressure balls, 8+ pressure games, active in 2025–26. A bootstrap over 200 resamples of matches gives every player a rank range.

04 / The key finding

Raw clutch numbers are mostly noise.

▸ Reliability check

A player’s pressure stats in odd matches barely predict their even matches.

0.06
Split-half reliability (r) of pressure performance at about 150 pressure balls per half. Rank on raw pressure stats and you mostly rank luck.

So each player’s pressure score is shrunk toward their own overall level by an amount set from that measured reliability. The ranking still rewards players who lift under pressure, but 150 lucky balls can’t top the list.

05 / Output

The auction-available top 10s.

Filtered to players not retained on 18 Nov 2026 and clearly available to the IPL, then ranked within each player’s best phase using the SR / runs per innings / balls per dismissal / balls per boundary method from my SRH scouting work, blended 60/40 with pressure value.

Batters
01Kamindu MendisAnchor
02Harry BrookNo. 3
03Shahrukh KhanFinisher
04Reeza HendricksTop order
05Devon ConwayAnchor
06Kusal MendisOpener, wk
07Kusal PereraMiddle order
08Pathum NissankaOpener
09Jacob BethellDeath hitter
10Lorcan TuckerMiddle, wk
Bowlers
01Mitchell StarcDeath pace
02Spencer JohnsonDeath pace
03Wanindu HasarangaSpin, mid + death
04Matheesha PathiranaDeath pace
05Liam DawsonSpin, middle
06Sean AbbottDeath pace
07Trent BoultNew ball
08Allah GhazanfarSpin, PP
09Lockie FergusonPace, middle
10Mujeeb Ur RahmanSpin, middle
06 / Building RR’s XI

Strong powerplay bat, weak death bowling.

In IPL 2025–26 RR scored at 181 in the powerplay against a league 152, but conceded 11.41 an over at the death against 10.67. Sandeep Sharma and Ravi Bishnoi rank 113th and 122nd of 137 death bowlers under pressure.

Plan A: Jaiswal, Sooryavanshi, Jurel, Parag, Shahrukh Khan, Ferreira, Jadeja, Hasaranga, Archer, Starc or Spencer Johnson, Shami. Two of RR’s three overseas slots go on bowlers who replace its two weakest pressure bowlers, and an Indian finisher means no overseas slot is spent on batting.

07 / Limitations

Where it is less certain.

01
Small pressure samples. 150–700 pressure balls per player; ranks 6–10 move across resamples.
02
First-innings stakes are weaker. The first-innings win model has AUC 0.69, so leverage is blended with observable signals rather than used alone.
03
Missing context. No venue scoring norms, pitch, dew or batting-position intent.
04
No domestic data. Uncapped Indian players can’t be rated from IPL and T20I alone; Syed Mushtaq Ali Trophy data is the next addition.
05
Availability changes. Bangladesh players and Harry Brook carry eligibility questions for 2027; prices are not modelled.
Explore the ratings

See who lifts under pressure.

Filter 100 batters and 100 bowlers by auction status, compare normal and pressure numbers, and see RR’s phase gaps.