IPL Data Analysis — Crunch 26 Submission

00 — Problem Statement

What we set out to answer

IPL analysis is dominated by commentary and intuition. Fans assume toss winners have an edge, that death overs decide games, and that strike rate is the only batter metric that matters. This project challenges those assumptions with evidence.

The questions that drove this

Three specific questions formed the brief: (1) Do teams that win the toss actually win more matches? (2) Which phase — powerplay, middle overs, or death overs — is most linked to winning? (3) Who are the top 5 batters and bowlers across 5 seasons?

But the more interesting meta-question was: what does conventional cricket wisdom get wrong? That became the thread running through the entire analysis.

Deliveries analyzed

84,588

Ball-by-ball granularity

Matches

353

2021–2025 IPL seasons

Wickets recorded

4,041

Excluding run-outs

Franchises

With 10+ matches

01 — My Process

How I approached this

Six steps from raw CSV to submission. The reasoning behind each choice matters as much as the output.

Data acquisition & format decision

Downloaded ball-by-ball CSV from Cricsheet.org. Each row is one delivery — batter, bowler, runs, wicket type, match outcome, toss info. 289,000 rows total before cleaning.

I chose this because… JSON from Cricsheet is richer but requires custom parsing. The CSV variant already has match-level fields (toss, winner) embedded per row — faster to join without a separate match file.

Data cleaning & scoping

Converted season to numeric (mixed dtype issue), filtered to 2021–2025 for recency, merged both RCB name variants (Royal Challengers Bangalore / Bengaluru) into one, classified wicket types excluding run-outs and retired hurt from bowler tallies.

I chose this because… Using all 16 seasons would dilute modern T20 patterns — the game has changed dramatically since 2009. 5 seasons gives enough sample size (353 matches) while reflecting current playstyle. RCB's naming inconsistency inflated their apparent record — merging was essential for honest team rankings.

Toss analysis — base rate first

Calculated toss-winner win rates overall, then split by toss decision (bat/field), then by season. Also checked if the trend was improving or degrading over time.

I chose this because… The common mistake is to report the headline number (50.4%) and stop. Slicing by decision revealed the real story: batting after winning the toss is actively harmful, which the headline alone would have hidden.

Phase segmentation — runs per match not total

Defined three phases (Powerplay 1–6, Middle 7–15, Death 16–20), calculated average runs per match per phase for winning vs losing teams, then computed the winner-loser gap as the key metric.

I chose this because… Total runs across all matches would favour phases with more overs. Using average-per-match normalises for match count and innings count — apples-to-apples comparison. The differential (not just the raw average) is what reveals which phase is the actual differentiator.

Performer ranking — beyond raw volume

Ranked batters by total runs, then added strike rate, fours, and sixes. For bowlers, ranked by wickets but foregrounded economy rate as the secondary metric.

I chose this because… Runs alone favour openers who face more balls. Strike rate alongside runs gives a fairer picture of impact. For bowlers, economy in T20 is as important as wickets — a bowler who takes 80 wickets at economy 9.5 is less valuable than one with 70 wickets at 7.9.

Visualisation — dark dashboard + written report

Built an interactive HTML dashboard with Chart.js (9 charts), a structured markdown report, and this combined case study. All three serve different audiences: hackathon judges, data peers, and general readers.

I chose this because… A single static PDF loses the interactive detail that makes chart exploration rewarding. An HTML dashboard works without any install. The report provides the narrative that charts alone can't convey.

❌ What didn't work — iterations

First attempt — used all 16 seasons (2009–2026): Phase analysis results were muddied by rule changes (powerplay overs changed in early IPL seasons). Scrapped and scoped to 2021–2025 only.

Second attempt — "average runs per over" as the phase metric: This seemed more normalised but actually introduced noise because partial overs (last ball of innings) skewed the per-over average. Switched to total runs per match per phase, then averaged — cleaner signal.

Third attempt — reported RCB as two separate franchises: The name change mid-dataset (Bangalore → Bengaluru) made them appear twice in team rankings with lower sample sizes, inflating uncertainty. Merged on string matching — honest reporting required it.

02 — Toss Analysis

Does winning the toss win the match?

Short answer: no. At 50.4%, toss winners are statistically indistinguishable from a coin flip. But the decision they make with the toss matters enormously.

Toss winner vs loser — overall win rate

353 matches · 2021–2025

Toss winnerToss loser

Win rate by toss decision — bat or field?

85 chose bat · 268 chose field

Chose to batChose to field

Toss win rate by season — no stable trend

Fluctuates between 43.7% (2024) and 58.1% (2025) — year-specific conditions matter more than the toss

03 — Phase Analysis

Which phase decides the match?

Middle overs (7–15) show a 7.4-run gap between winners and losers — larger than powerplay (5.6) and death overs (3.6). The game is won in the middle, not at the edges.

Average runs per phase — winners vs losers

Per-match average across 353 games

WinnersLosers

Run advantage by phase (Winners − Losers)

Middle overs dominate — 7.4 run gap

Middle overs run trend — winners per season (2021–2025)

+18% growth in 5 years — scoring is accelerating in the most critical phase

04 — Top Performers

Best batters & bowlers (2021–2025)

Raw volume tells part of the story. Strike rate and economy complete it.

Top 5 batters — total runs

2021–2025 combined · with strike rate

#	Batter	Runs	4s	6s	SR
1	Shubman Gill	2927	285	95	143.7
2	V Kohli	2783	268	90	137.6
3	KL Rahul	2582	219	104	136.4
4	F du Plessis	2471	235	101	142.5
5	JC Buttler	2406	241	108	149.3

Top 5 bowlers — wickets taken

Excluding run-outs · economy rate is the key differentiator

#	Bowler	Wickets	Economy	Overs
1	HV Patel	105	9.09	249.7
2	YS Chahal	100	8.42	276.8
3	Arshdeep Singh	85	9.12	250.8
4	Avesh Khan	83	9.16	239.8
5	Rashid Khan	83	7.95	286.5

Top 3 batters — season-by-season progression

Gill peaks in 2023 (890 runs). Kohli accelerates 2023–2025. Rahul consistent throughout.

Shubman GillV KohliKL Rahul

05 — Team Performance

Franchise win rates (2021–2025)

RCB's consistent improvement across 75 combined matches puts them at 57.3% — second only to Gujarat Titans. Mumbai Indians' 44.6% is their weakest 5-season stretch in franchise history.

Team win percentage — all franchises (10+ matches)

RCB merged: both "Bangalore" and "Bengaluru" name variants combined into single record

06 — Dismissal Patterns

How do batters get out?

72% of all wickets come from catches — making aerial fielding placement and the ability to create top edges arguably the most critical bowling skill in IPL cricket.

Dismissal type distribution

4,041 wickets · 2021–2025

Wicket count by dismissal type

Absolute counts

07 — Key Insights

What the data actually says

INSIGHT 01

The toss is noise, not signal. 50.4% is statistically indistinguishable from 50%. Captains, team selectors, and pundits who analyse toss trends are analysing randomness.

INSIGHT 02

Games are won in the middle, not the margins. A 7.4-run middle-overs advantage outweighs powerplay (5.6) and death overs (3.6). Teams that build in overs 7–15 win more — not teams with the best finishers.

INSIGHT 03

Rashid Khan is an outlier in the mathematical sense. 7.95 economy with 83 wickets. No other top-5 bowler breaks below 8.4. He operates in a different statistical universe.

INSIGHT 04

Scoring is in structural acceleration. Winner middle-over runs grew 18% (71.4→84.2) in 5 seasons. This is not random variance — it's a systemic shift in how T20 batsmanship has evolved.

⚡ The one genuinely surprising finding

Toss winners who chose to bat first won only 47.1% of matches — lower than the 49.6% base rate for toss losers. Batting first after winning the toss is literally worse than not winning the toss at all. Yet 75.9% of captains who won the toss chose to field — meaning the professionals already know this intuitively. The data confirms what elite T20 captains learned through experience: the "bat first, set a target" playbook is extinct. Chasing is a structural advantage in modern IPL.

08 — Results & Outcomes

What this analysis delivered

Outputs produced

9 interactive charts covering toss impact, phase differentials, performer rankings, team win rates, and dismissal patterns.

3 submission artefacts: Interactive HTML dashboard (this file), structured markdown report, and combined case study document — each targeting a different reader: judge, peer, and general audience.

1 headline finding that directly contradicts popular cricket commentary: batting first after winning the toss is counterproductive.

09 — What I'd Do Differently

Next iteration improvements

🎯

Add venue-level analysis

Toss-field advantage likely varies by ground — Wankhede (small boundaries) vs Chepauk (spin-friendly) should show different patterns. Venue-stratified toss win rates would be more actionable for captains.

📊

Model win probability per ball

Instead of phase averages, build a delivery-level win probability model. Which over specifically turns a game? Which wickets are "high leverage"? This would make the phase finding much more precise.

🔁

Head-to-head matchup data

Batter vs specific bowler performance would reveal whether Rashid's economy holds against top-ranked batters or whether it's inflated by weak opposition in certain matches.

🧮

Statistical significance tests

The toss finding (50.4%) is directionally clear but deserves a chi-square test for rigor. With 353 matches the power is modest — reporting confidence intervals would make the "coin flip" claim defensible to a statistics audience.

📱

Mobile-optimised dashboard

Chart.js on mobile requires careful viewport handling. Next version would use a responsive grid that stacks to single column and reduces chart heights for phone screens.

🏃

Run-rate pressure modelling

Required run rate when a wicket falls is a more nuanced "pressure" metric than phase runs. Adding pressure context to the phase analysis would explain why middle overs matter — teams under pressure there collapse.

Skills demonstrated

Python pandas numpy Data Cleaning EDA Feature Engineering Chart.js Data Visualisation Dashboard Design Statistical Analysis Hypothesis Testing Cricket Analytics T20 Strategy Sports Data

Cricket bythe Numbers