📚 Research & Methodology

Our predicted XI algorithm combines multiple data sources to generate optimal team compositions:

Scoring Components

• Base Classification (0-30 pts) - Role-based scoring for position fit
• Performance Tags - Phase-specific tags (PP_ELITE, DEATH_SPECIALIST, etc.)
• Derived Metrics - boundary%, consistency_index, death_dot_pct
• Price Tier Bonus (5-15%) - Investment level consideration
• Variety Optimization - LHB/RHB balance, spin/pace mix

Hard Constraints

• C1: Captain cannot be Impact Player
• C2: Maximum 4 overseas players
• C3: Minimum 20 overs bowling coverage
• C4: At least 1 wicketkeeper
• C5: At least 1 spinner

Pro Football Focus revolutionized NFL analytics by grading every play. We adapt their methodology:

Key Concepts

• Ball-by-Ball Grading - Evaluate each delivery on a -2 to +2 scale
• Process Over Outcome - Good decisions that fail still get positive grades
• Context Adjustments - Situation, opposition, and conditions matter
• WAR (Wins Above Replacement) - Aggregate value metric

Cricket Applications

• P0 Batter grading vs bowling type and phase
• P0 Bowler grading by delivery outcome quality
• P1 Fielding impact assessment
• P2 Captain decision grading

📄 PFF Research Document (Internal)

Ken Pomeroy's basketball ratings remove pace from the equation. We apply similar concepts:

Key Concepts

• Adjusted Efficiency - Opponent-normalized performance
• Tempo-Free Stats - Per-possession (or per-ball) metrics
• Four Factors - Decompose performance into components
• Strength of Schedule - Quality of opposition faced

Cricket Applications

• P0 Four Factors: Boundary%, Dot Ball%, Extras, Bowling Changes
• P0 Venue Park Factors: Adjust for pitch and ground size
• P1 Opposition Strength Index: Weight by opponent quality
• P1 Adjusted Strike Rate: SR normalized by context

📄 KenPom Research Document (Internal)

We use K-means clustering to identify natural player archetypes:

Batter Archetypes

• EXPLOSIVE_OPENER - High SR, PP aggression, boundary-heavy
• PLAYMAKER - Consistent scoring, adaptable approach
• ANCHOR - Low dot%, innings builder, lower SR
• MIDDLE_ORDER - Middle overs specialist, rotation focus
• FINISHER - Death overs specialist, high SR at end

Bowler Archetypes

• WORKHORSE - Consistent economy, regular overs
• NEW_BALL_SPECIALIST - PP wickets, swing/seam
• DEATH_SPECIALIST - Low death economy, yorker execution
• WICKET_TAKER - High wickets, aggressive approach

📄 Creative Archetype Descriptions (Internal)

CricPom adapts college basketball's KenPom methodology to T20 cricket, producing opponent-adjusted, venue-neutral player ratings that account for tournament quality and conditions similarity.

Core Adjusted Metrics

• AdjBRR (Adjusted Batting Run Rate) — Batting run rate adjusted for bowling quality faced, venue park factor, and match context. Formula: raw_RR × (league_avg_bowling / opponent_bowling_quality) × venue_factor
• AdjBE (Adjusted Bowling Economy) — Economy rate adjusted for batting quality faced and venue. Lower is better. raw_econ × (league_avg_batting / opponent_batting_quality) × venue_factor
• CEM (Composite Efficiency Metric) — All-rounder evaluation combining AdjBRR and AdjBE into a single efficiency score. CEM = w₁·AdjBRR_percentile + w₂·(1 - AdjBE_percentile)
• OSI (Opponent Strength Index) — Weighted average quality of opponents faced, used as the adjustment denominator in all CricPom ratings

5-Factor Tournament Quality Engine

CricPom consumes the Tournament Quality Weighting system (see below) to weight data from 426 T20 tournaments. Each tournament's data receives a composite weight computed as:

W = ∏(fᵢ^wᵢ)^(1/Σwᵢ) — geometric mean of 5 factors

• PQI (25%) — Player Quality Index: average career quality of participants
• Competitiveness (20%) — Match balance and outcome distributions
• Recency (20%) — Exponential decay favoring recent data
• Conditions Similarity (15%) — How closely conditions match IPL 2023-2025
• Sample Confidence (20%) — Statistical reliability from match volume

How It Differs from Raw Stats

Aspect	Raw Stats	CricPom Adjusted
Opposition quality	Ignored	OSI-adjusted
Venue effects	Ignored	Park factor adjusted
Tournament relevance	All equal	5-factor weighted
Recency	All equal	Exponential decay
All-rounder evaluation	Separate batting/bowling	Unified CEM score

Status: Tournament weights computed (TKT-187). CricPom foundation metrics (AdjBRR, AdjBE, CEM) implemented (TKT-190). Groundwork research complete.

📄 KenPom Research Foundation (Internal)

Not all T20 data is equal. Our weighting system quantifies tournament quality across 426 tournaments and 9,357 matches:

5-Factor Composite Weight

• Player Quality Index (PQI) - Average career quality of tournament participants
• Competitiveness Index (CI) - Match outcome balance, margin distributions
• Recency Decay - Exponential decay weighting recent tournaments higher
• Conditions Similarity - How closely tournament conditions match IPL 2023-2025 (Founder Decision #6)
• Sample Size Confidence - Statistical reliability based on matches played

IPL 2023+ Baseline (Founder-Locked)

All conditions comparisons use IPL 2023-2025 as the baseline, not all-time IPL averages. The data shows the 2021-22 to 2023-25 transition produced the largest single jump in IPL history: Run Rate +1.00, Boundary% +3.2, Six% +1.65. Using the all-time average (RR 7.86) would dilute comparisons against a fundamentally different era than the modern IPL (RR 8.98).

Tournament Tiers (Provisional)

• Tier 1A IPL - Baseline (1.0x weight)
• Tier 1B PSL, SA20, The Hundred, MLC, BBL, CPL - Major franchise leagues (0.70-0.85)
• Tier 1C ILT20, LPL, Super Smash, Vitality Blast - Established leagues, lower overlap (0.50-0.70)
• Tier 2 T20 World Cup, Asia Cup - High-quality international (0.60-0.80)
• Tier 3 SMAT - Domestic Indian T20 (0.40-0.50)

Status: Plan approved by Founder. IPL 2023+ baseline locked. Implementation via TKT-183 (8-12 days).

📄 Tournament Weighting Plan (Internal)

Every analytical view now exists in two scopes to balance historical context with current-form accuracy:

Architecture

• _alltime views - Full IPL history (2008-2025, ~1,169 matches). Used for career records, historical comparisons
• _since2023 views - Current analytical window (2023-2025, ~219 matches). Used for all predictive outputs, tags, archetypes
• 80 dual-scope views - 40 pairs covering batting, bowling, phase, venue, matchup, and Film Room tactical analysis

Why 2023+? The Data Evidence (Founder-Approved)

DuckDB analysis of 1,169 IPL matches reveals a structural break at 2023 — the largest single-era shift in IPL history:

Era	Matches	Run Rate	Boundary%	Six%	Dot%
2008-2012	322	7.60	15.2	4.04	36.1
2013-2017	314	7.90	16.1	4.63	34.8
2018-2020	180	8.17	16.9	5.55	33.8
2021-2022	134	7.98	16.5	5.41	35.2
2023-2025	219	8.98	19.7	7.06	31.6

2023+ vs 2008-2022 deltas: Run Rate +14.2%, Boundary% +23.1%, Six-hitting +49.6%, Dot Ball% -10.0%

What Caused the Break

• Impact Player rule (2023) — 12 effective players per side, inflating batting depth and scoring rates
• 2022 mega auction reset — team compositions fundamentally reshuffled, pre-2023 team context obsolete
• Evolved batting intent — six-hitting up 49.6%, batters attacking from ball one in the modern IPL
• 219 matches provides sufficient sample for statistical reliability across all analytical views

📄 TKT-181 Review Document (Internal)

Every insight published in the magazine needs a confidence assessment. This framework scores each analytical claim on a 0-100 scale with letter grades:

Scoring Breakdown

• Sample Size (40%) — Capped at 300 balls/innings. HIGH (≥300), MEDIUM (≥100), LOW (<100)
• Consistency (25%) — Metric stability across sub-samples (first half vs second half of career)
• Recency (20%) — How recent the underlying data is (1.0 = all 2025, 0.5 = mix)
• Cross-Validation (15%) — Does the insight hold across conditions (home/away, bat/field)?

Grade Boundaries

• A (≥85) Publish with confidence
• B (≥70) Publish with minor caveats
• C (≥55) Add sample size caveat and limitations
• D (<55) Do not publish as standalone insight

Status: Framework implemented. Bridges with existing Confidence Intervals (TKT-145). Used editorially for all stat pack claims.

Data Sources

• Cricsheet Ball-by-Ball - 219 IPL matches (2023-2025)
• IPL 2026 Auction Data - Squad compositions and prices
• Historical Records - Team vs team, venue performance

Derived Metrics (115+ Views)

• 80 dual-scope views (_alltime + _since2023 pairs)
• Batter consistency index, boundary percentage, dot ball percentage
• Partnership synergy scores, pressure sequences
• Phase-specific performance (PP, middle, death)
• Bowling type matchups, handedness analysis
• 13 Film Room tactical views (entry points, wicket clusters, bowling changes)

Validation

• Every insight reviewed by cricket domain expert (Andy Flower)
• 8-step quality assurance process
• Founder sign-off on all key outputs