Shriyans Raj

IPL Data Analysis 2015–2019: Python, Pandas & 70K+ Ball-by-Ball Records

IPL fans and commentators constantly debate whether winning the toss matters, which phase of the game is most important, and who the best performers are — but these opinions are rarely backed by data. Most analysis stops at aggregated match results or highlights reels. I wanted to go deeper and analyze actual ball-by-ball delivery data from IPL 2015–2019 to answer three specific questions: Do toss winners actually win more? Which phase — powerplay, middle, or death overs — correlates most with winning? Who are the top 5 batters and bowlers across these 5 seasons? Process Downloaded ball-by-ball IPL data covering 2008-2026 from Cricsheet, converted JSON to CSV,and filtered out 5 seasons - 2015-2019 and cleaned it using Python and Pandas. Removed null values, handled extras like wides and no-balls, and excluded run-outs from bowler wicket counts. Split the analysis into three focused questions. Initially assumed death overs would be the decisive phase, but when I calculated average runs per phase for winning vs losing teams, middle overs showed the biggest gap — 7.8 runs compared to just 2.5 in death overs. Built an interactive dashboard with Chart.js to visualize findings and cross-verified every number against the raw dataset to ensure accuracy. Results Toss winners win 54% of matches vs 46% for toss losers-a real advantage, but barely better than a coin flip. The bigger finding: middle overs (7–15) are the true match differentiator. Winning teams average 73.7 runs in middle overs vs 65.9 for losers — a gap of 7.8 runs. Death overs only show a 2.5-run gap. This busts the myth that death overs decide matches. Top performers: V Kohli leads all batters with 2,780 runs, closely followed by DA Warner at 2,743. B Kumar tops bowlers with 89 wickets, just ahead of YS Chahal at 88. The consistency at the top is remarkable-margins are incredibly tight. Reflection I'd add venue-specific and chasing vs batting-first splits to the phase analysis. Chasing teams in death overs likely behave very differently than teams batting first, and certain venues favor specific phases. I'd also move beyond descriptive statistics and build a predictive win probability model using these phase-wise run patterns, possibly adding strike rotation analysis like dot ball percentage and boundary frequency.