WAR-Based Elo Power Rankings: MI's 1653 Rating Is 131 Points Clear of the Field
WAR-Based Elo Power Rankings: MI's 1653 Is 131 Points Clear of the Field
The old Elo model ranked teams by franchise history. RCB entered 2026 at #1 because they won the 2025 title. PBKS sat at #2 because Shreyas Iyer carried them to a strong season. CSK languished at #10 because their aging core underperformed.
After the mega auction, those rankings are meaningless. CSK 2025 and CSK 2026 share a name and a jersey. They do not share a squad.
The new WAR-based Elo model initializes team ratings directly from projected squad composition — the same approach FanGraphs uses in baseball. Each team's projected WAR/M is converted to a win rate, then mapped to the Elo scale. The result: a pre-season rating that reflects the actual players on each roster, not the ghosts of last year's results.
The Full Table
| Rank | Team | Elo | Projected WAR/M | Win Rate |
|---|---|---|---|---|
| 1 | MI | 1653 | 4.51 | .707 |
| 2 | SRH | 1522 | 3.80 | .531 |
| 3 | GT | 1519 | 3.79 | .527 |
| 4 | DC | 1498 | 3.67 | .498 |
| 5 | KKR | 1493 | 3.64 | .490 |
| 6 | RCB | 1489 | 3.62 | .485 |
| 7 | LSG | 1486 | 3.59 | .479 |
| 8 | RR | 1472 | 3.52 | .460 |
| 9 | CSK | 1449 | 3.39 | .427 |
| 10 | PBKS | 1426 | 3.26 | .395 |
MI: The Outlier at 1653
MI's projected WAR of 4.51 is 0.71 above the next team. That is not a narrow edge. It is the equivalent of adding a league-average player to a lineup that already has 11. No team in the WAR-based era has entered a season with a larger projected advantage.
The source: Bumrah (0.93), SKY (0.68), Boult (0.43), Tilak (0.31), Rohit (0.31), Rickelton (0.37), Jacks (0.34). Seven above-replacement batters from 1 through 7, plus the IPL's best death bowler. The vulnerability — middle-overs bowling via Thakur and Bosch — exists, but the depth elsewhere absorbs it.
MI vs PBKS produces a 79% pre-match win probability. MI vs any team in the 4-7 band (DC, KKR, RCB, LSG) sits around 70%. Only SRH and GT project competitive matchups against MI, and even those favor Mumbai at 68/32.
The Pack: 6 Teams Within 37 Points
From SRH at 1522 to RR at 1472, six teams sit within a 50-point Elo band. In practical terms, any match between two teams in this group is a 53/47 proposition at most. The WAR model sees these six teams as essentially interchangeable in squad strength.
SRH and GT separate slightly from the pack, driven by specific structural advantages:
- SRH: Abhishek Sharma (0.74) and Travis Head (0.65) form the IPL's best opening pair by projected WAR. Combined: 1.39 WAR/M from two players.
- GT: Buttler-Gill-Sudharsan form the highest-projected top three of any squad, and Washington Sundar anchors the best middle-overs bowling unit.
The middle four (DC, KKR, RCB, LSG) have projected WAR/M ranging from 3.59 to 3.67. A 0.08 WAR/M difference is smaller than the projection uncertainty on any single player. These teams will be separated by matchday form, venue conditions, and in-season player development — not pre-season squad composition.
The Defending Champions at #6
RCB won the 2025 title. The WAR model ranks them 6th. That disconnect is the entire point of WAR-based initialization. The 2025 RCB squad — Salt, Kohli, du Plessis, Harshal, Siraj — is not the 2026 RCB squad. The mega auction reshuffled every roster.
RCB's 3.62 projected WAR sits just 0.03 above the league mean of 3.58. Their structural profile is balanced but not dominant: Salt provides an elite opener, Krunal anchors the bowling, but Romario Shepherd's -0.12 negative projected WAR at No. 7 is the only negative projection in any team's starting XII.
The Bottom: PBKS at 1426
PBKS's 3.26 projected WAR is the lowest of any team by 0.13. That gap translates to a pre-season Elo of 1426, producing a projected win rate of just .395. Against the top three (MI, SRH, GT), PBKS face 30-37% win probabilities.
The roster has batting depth through seven positions, but no single player projects at the elite tier. Iyer leads at 0.45 projected WAR/M — strong but not at the Bumrah/Abhishek/Buttler level. The bowling relies on Arshdeep, Chahal, and Brar, none of whom project above 0.25 WAR/M.
CSK at 1449 face a similar structural challenge: spin-heavy bowling but death-overs vulnerability through Khaleel Ahmed's 12 death economy.
How WAR-Based Elo Works
The model converts squad projected WAR to team Elo through three steps:
1. Sum projected WAR/M for each team's optimal XII
2. Convert to win rate: `winRate = 0.500 + 0.25 × (teamWAR - meanWAR)`
3. Map to Elo: `elo = 1500 - 400 × log10((1/winRate) - 1)`
Once the season starts, match results update Elo via the standard K=32 formula, same as before. A win against a stronger team earns more points; a loss to a weaker team costs more. By mid-season, actual results will dominate the WAR-based starting point.
The scale factor of 0.25 is calibrated so that the Elo spread (~227 points) aligns with historical IPL win-rate variation. Historical seasons (2020-2025) remain unchanged in the model — the WAR-based initialization applies only to 2026+, where the mega auction makes franchise-history Elo unreliable.
What Changes When Matches Start
The WAR-based ratings are pre-season projections. They will move, fast, once real matches produce real results. With K=32, a single upset can shift team ratings by 20+ points. By match 7, teams with 5 wins will have climbed 60-80 points regardless of where they started. By the midpoint of the season, the WAR initialization will be a small fraction of each team's total Elo signal.
The value of WAR-based initialization is not permanence. It is accuracy at the starting point — giving the model a better prior than "everyone starts at 1500" when every roster has changed.