Cracking the Data - What Actually Wins in IPL

Analyze IPL data of 289673 rows and answer questions such as: Do teams that win the toss actually win more matches? Which phase impacts victory the most — Powerplay, Middle Overs, or Death Overs? Who are the top batters and bowlers across seasons? What hidden patterns or surprises can you discover from the data? Process From Raw Data to IPL Insights - I took a 289,673 row ball by ball IPL dataset and filtered it to 5 seasons (2020–2024), giving 339 matches. Using Python (pandas), I deduplicated delivery-level data to match-level, engineered a phase column (Powerplay/Middle/Death), created boolean win flags, and ran groupby aggregations to compute average runs per phase for winners vs losers. Toss analysis used a simple mean on a boolean flag. Player rankings were used with groupby + sum/count. I built all visuals and the 15-slide deck programmatically in PptxGenJS with SVG graphics rendered via Sharp. Every claim is backed by visible Python code embedded directly in the slides. Results Across 339 IPL matches, the data revealed three clear findings. Toss winners won just 47.2% of matches - worse than random chance: with bat-first captains performing worst at 42.9%. Phase analysis showed the powerplay (0–5 overs) has the highest win correlation at an 11.96% run gap, directly contradicting the popular belief that death overs are decisive, which showed only a 6.86% gap. In the player rankings, F du Plessis, Shubman Gill, and KL Rahul were separated by just 12 runs across five seasons, while YS Chahal led all bowlers with 105 wickets. Reflection Rather than presenting results at face value, I let the data challenge conventional IPL wisdom, and it did. Instead of building slides manually, I coded the entire 15-slide deck programmatically using PptxGenJS, embedding Python proof snippets directly on each analysis slide so every claim was traceable. I designed the Q2 divider as a curiosity-first reveal with hidden "???" impact scores to create narrative tension before showing the answer. The surprising finding that powerplay matters nearly 2x more than death overs wasn't a hypothesis I set out to prove; it emerged purely from the analysis