Formulas and parameters · Data as of 18 Nov 2026
Method
The pipeline has four steps. A win-probability model scores every ball by how much its outcome can swing the match. Each player's deliveries are valued, weighted by that pressure, and shrunk toward the pool average so that a hot streak over 80 balls cannot outrank a career over 800. A consistency term counts how often a player beats his own baseline. Finally we validate the whole thing on matches the ratings never saw, and report the ugly parts as well as the good ones.
Data and corrections
The pack holds 867 matches and 200,999 deliveries across five seasons (2022 to 2026, IPL and T20I; 141 competition rows). A draft pass at this data reported 883 matches; re-counting the matches table gives 867, and 867 is what the report uses everywhere. Chases and the model's quality score come from the pipeline on the private ball-by-ball file, which is not part of the submission package, so we label them reported rather than independently recomputed: 845 chases, of which 842 were decided, and a cross-validated AUC of 0.889. Reported
Two scoring rules in the raw files were fixed. A delivery that goes for byes or leg-byes is a legal ball: the over advances, the batter did not score off the bat, and the bowler is not charged. Wides and no-balls are the opposite: not legal, and charged to the bowler. Getting this backwards inflates over counts by roughly one ball in fifty and distorts death-over economies most, which is exactly where the auction decisions live.
Pressure and win probability
WP is the chasing side's win probability from a match state, from a gradient-boosted tree classifier constrained to be monotonic: needing fewer runs raises it, more balls remaining raises it, more wickets lost lowers it. Trained on the 842 decided chases; grouped 5-fold cross-validation (whole matches held out, never a single match split across train and test) gives AUC 0.889. Reported Each over's mean PS is normalised to the overall mean (0.141) to remove first-over (0.159) and last-over (0.174) inflation against 0.126 in the middle overs.
How a player is scored
Ball value
Credited wickets are caught, bowled, lbw, stumped and hit wicket; run-outs are excluded because they are mostly the fielder's doing. The 12 is the run value of a wicket in a chase; results barely move anywhere in the 10 to 14 range.
Pressure weighting
Each ball's value is averaged with the ball's PS as its weight: what the player does when the game can swing counts more than what he does when it cannot.
Shrinkage toward the pool average
Vs is pulled toward the pool average in proportion to how little pressure the player has faced. Es is the headline number: value above an average player under pressure. A hot fortnight over 80 balls cannot outrank three seasons over 900.
Consistency
The Laplace-smoothed share of chase appearances where the player beat his own low-pressure baseline. The +1 and +2 stop a player who did it twice from claiming a perfect record.
The scores
Each part of PIS is standardised before combining. The 70/15/15 split is a disclosed judgement: pressure value does the work, consistency and familiarity with Indian conditions refine it. PIS v2, on 0 to 100, mixes the rating with things the auction actually prices: 40% overall value per ball (uniform weights), 20% recent form, 20% fit to RR's specific gaps, 15% clutch (the Es above), 5% durability. Where the two scales disagree, the signings page says which one is driving the pick.
Parameters
| Parameter | Value | Note |
|---|---|---|
| Wicket run value (chase) | 12 | Results barely move anywhere in the 10 to 14 range |
| Shrinkage K, batters | 160 | V_s = (S_w·V_p + K·pool) / (S_w + K) |
| Shrinkage K, bowlers | 60 | same formula |
| High-pressure bar | PS ≥ 0.30 | About 6.6% of chase balls clear it |
| Low-pressure bar | PS < 0.12 | |
| WP model AUC | 0.889 | Grouped 5-fold CV on 842 decided chases; pipeline-reported |
| Over-mean normalisation | 0.141 | First over 0.159 and last 0.174 normalised to the overall mean |
| Half-splits | 30 | Rate on one half, rank-check the other |
| Bootstrap resamples | 200 | Rebuild each top-10 per resample |
| PIS weights | 0.70 / 0.15 / 0.15 | z(E_s) / z(c) / z(Indian-venue share) |
| PIS v2 weights | 0.40 / 0.20 / 0.20 / 0.15 / 0.05 | value per ball / recent form / RR gap fit / clutch / durability |
| Indian relaxed threshold | 30+ balls, 2+ apps | Indian domestic data absent from the pack |
Definitions and conventions
"Confirmed" pool status means released by a franchise on 17 Nov 2026 or reported registered; "Likely" means on no retention list, with registration closing 23 Nov; the registered pool reaches franchises on 3 December 2026. "Reported" marks numbers computed on the private ball-by-ball file that is not in the package. Prices marked as bands or point estimates are estimates, not official figures; real prices depend on the 16 December 2026 auction in Goa.
The top-10 lists use PIS v2 empirically shrunk to 0–100; the signings section quotes PIS v2 on its own scale, as printed in the report. The pressure explorer on the Pressure page runs on a disclosed illustrative formula precomputed to JSON at build time (r_exp = 8.6 + 3.0·(1 − balls/120)·√(1 − w/10) − 2.0·(w/10); μ = (balls/6)·rexp; τ = 1.6·√balls + 0.8; WP = Φ((μ − need)/τ)); it carries the same monotone properties but is not the model. Illustration
Which file feeds which page
| Source file (data/csv/) | Used by |
|---|---|
| rr_retained_players.csv, rr_released_players.csv | Squad page |
| candidate_pool_with_status.csv, impact_scores.csv | Players page |
| final_signings.csv, final_signings_backups.csv, role_match_table.csv | Signings page, player drawers |
| top10_batters.csv, top10_bowlers.csv | Top-10 page |
| bootstrap_stability.csv, bootstrap_stability_bowlers.csv | Top-10 and Validation pages |
| validation_halfsplits.csv, validation_halfsplits_v2.csv | Validation page |
| indian_auction_watchlist.csv, exclusion_log.csv | Method page appendices |
| RR_Auction_Targets_IPL2027.xlsx | Signings prices, Squad gap mapping |
| sources_sheet.csv | Sources table below |
Sources
Every web fact behind the report, with source, URL, access date and confidence. Access date 18 Nov 2026 for all rows.
| Fact | Source | URL | Accessed | Confidence |
|---|---|---|---|---|
| RR retained 17 players for IPL 2027 (19-player squad incl. Shami + M Siddharth trades) | Rajasthan Royals official site | https://www.rajasthanroyals.com/latest-news/ipl-2027-rr-players-retention-list-retained-released | 2026-11-18 | high |
| RR released list (9): Hetmyer Deshpande Milne Maphaka Kuldeep Sen Yudhvir Singh Vignesh Puthur Ravi Singh Emanjot Chahal | Rajasthan Royals official site + Indian Express + Olympics.com | https://www.rajasthanroyals.com/latest-news/ipl-2027-rr-players-retention-list-retained-released | 2026-11-18 | high |
| RR squad slots: max 25 - 19 current = 6 open; overseas 8 - 5 = 3; purse INR 27.05 cr | Rajasthan Royals official site + Yahoo/Cricbuzz | https://sports.yahoo.com/articles/ipl-2027-rajasthan-royals-retention-200200696.html | 2026-11-18 | high |
| Trades: Shami + M Siddharth to RR; Sam Curran to LSG (INR 2.4 cr) | Rajasthan Royals official + Rediff | https://www.rediff.com/cricket/report/ipl-trade-shake-up-shami-curran-to-lsg/20261117.htm | 2026-11-18 | high |
| IPL 2027 auction: mini-auction 16 December 2026 Goa | ESPNcricinfo | https://www.cricinfo.com/story/ipl-2027-auction-to-be-held-on-december-16-1560085 | 2026-11-18 | high |
| Retention deadline extended to 17 Nov 2026; registered pool to franchises 3 Dec 2026 | ESPNcricinfo | https://www.cricinfo.com/story/ipl-2027-retention-deadline-moved-to-november-17-1561149 | 2026-11-18 | high |
| Purse INR 130 cr; squad min 18 max 25; overseas cap 8 squad / 4 XI | Yahoo Sports | https://sports.yahoo.com/articles/ipl-retained-players-2027-list-023001537.html | 2026-11-18 | medium-high |
| RR 2026 home split: 4 Jaipur (Sawai Mansingh) + 3 Guwahati (ACA Stadium) | Rajasthan Royals official schedule | https://www.rajasthanroyals.com/latest-news/rajasthan-royals-ipl-2026-schedule-match-time | 2026-11-18 | high |
| All-team IPL 2027 retained/released lists (MI CSK RCB KKR DC PBKS GT SRH LSG RR) | Crictracker (cross-checked vs ESPNcricinfo live blog) | https://www.crictracker.com/cricket-appeal/ipl-2027-retentions-full-list-of-players-retained-released-trades-for-all-10-teams-ahead-of-auction-1456 | 2026-11-18 | medium-high |
| IPL champions: 2023 CSK/GT 2024 KKR/SRH 2025 RCB/PBKS 2026 RCB/GT | Adda247 + 2026 final reports | https://currentaffairs.adda247.com/ipl-winners-list | 2026-11-18 | medium |
| MS Dhoni retired from IPL 17 Nov 2026; named CSK mentor | The Hindu | https://www.thehindu.com/sport/cricket/chennai-super-kings-ms-dhoni-team-mentor-retire-nov-17-2026/article71711142.ece | 2026-11-18 | high |
| Jacob Bethell long-term injury misses IPL 2027; Travis Head and Mitchell Starc withdrew (workload) | ESPNcricinfo | https://www.cricinfo.com/story/starc-head-bethell-released-by-dc-srh-rcb-ahead-of-the-2027-player-auction-1562958 | 2026-11-18 | high |
| 184 players retained (55 overseas) 89 released IPL 2027 | IPLT20.com official | https://www.iplt20.com/news/article/tata-ipl-2027-player-retentions-announced | 2026-11-18 | high |
| Sufiyan Muqeem PSL 2026 leading wicket-taker 22 at 14.40; CPL 2026 final 4/14 POTM | ESPNcricinfo profile + BBC Sport | https://www.cricinfo.com/cricketers/sufyan-moqim-1329697 | 2026-11-18 | medium |
| Abrar Ahmed PSL 2026 joint-top 33 wickets; former No.1 T20I bowler | ESPNcricinfo | https://www.cricinfo.com/story/abrar-ahmed-s-white-ball-rise-is-no-longer-a-mystery-1528962 | 2026-11-18 | medium |
| Shoriful Islam BPL 2026 Player of Tournament 26 wickets; record 5/9 | Google-indexed BPL reports | https://www.google.com | 2026-11-18 | medium |
| Adam Milne hamstring tear Jan 2026 (SA20); ruled out of T20 WC 2026 | ESPNcricinfo | https://www.cricinfo.com/story/adam-milne-ruled-out-of-t20-world-cup-kyle-jamieson-called-up-1520622 | 2026-11-18 | high |
| Mohsin Khan retained by LSG (NOT auction pool); SMAT 2026-27 return 3/28 | IPLT20.com + ESPNcricinfo | https://www.iplt20.com/news/article/tata-ipl-2027-player-retentions-announced | 2026-11-18 | high |
| Hazlewood retained RCB (15 wkts IPL 2026 incl. final); Rashid retained GT; Santner retained MI; Salt Kohli retained RCB; SKY retained MI | IPLT20.com official | https://www.iplt20.com/news/article/tata-ipl-2027-player-retentions-announced | 2026-11-18 | high |
| Shubman Gill IPL 2026: 732 runs at 45.75 SR 163.03 | Google-indexed Orange Cap lists | https://www.google.com | 2026-11-18 | medium |
| Vaibhav Sooryavanshi IPL 2026 Orange Cap 776 runs SR 237.30 (retained by RR) | Google-indexed Orange Cap lists + RR profile | https://www.rajasthanroyals.com | 2026-11-18 | medium-high |
| Hetmyer IPL 2026: 78 runs in 7 inns; CPL 2026: two centuries incl. 111* off 43 | Facebook/Cricinfo + CPLT20 | https://www.facebook.com/Cricinfo/ | 2026-11-18 | medium |
| Tushar Deshpande IPL 2026 RR death bowler; defended 11 in final over vs GT | ESPN.com | https://www.espn.com/cricket/story/_/id/48717140/ | 2026-11-18 | medium |
| SMAT 2026-27 began 14 Nov 2026; early leaders Mayank Yadav 9 wkts at 2.88 | ESPNcricinfo series page | https://www.cricinfo.com/series/syed-mushtaq-ali-trophy-2026-27-1546522 | 2026-11-18 | medium |
| Pakistan players de facto ineligible for IPL (none signed since 2008) | General cricket-history fact | https://www.iplt20.com | 2026-11-18 | high |
| All pressure/rating numbers (E_s PIS PRS stability validation) | Contest data pack (private) - own computation | local computation on data pack | 2026-11-18 | high |