Using statistics and optimization to achieve 67% win rate
I love cricket and played fantasy cricket casually, but my win rate was terrible—around 30%. I'd pick my favorite players (Kohli, Dhoni) regardless of form or cost efficiency. After losing ₹500 in entry fees, I realized: what if I treated this like a math problem instead of a fan exercise? Could statistics actually beat intuition?
Built a data-driven model using Expected Value, consistency metrics, form weighting, and budget optimization (knapsack algorithm). Tracked every player's performance for 6 months: points per game, standard deviation, last 5 matches form. Created ROI rankings (points per credit cost). Stopped picking favorites, started picking high-EV players. Result: 67% win rate, climbed from rank 342 to rank 3 out of 500 players, won ₹4,200 in prizes.
from last place to league champion: IPL 2023 - joined Dream11 league with school friends. finished DEAD LAST in first 3 matches. everyone mocked me. decided to treat it like a math problem. started tracking: player form (last 5 matches), venue stats (batting avg at each stadium), match-ups (Bumrah vs specific batsmen), value-for-money (points per credit). built Excel model with weighted scoring. tested on past season data - 78% accuracy! applied model for remaining matches. climbed from 12th place to 1st. won league. prize: ₹5,000 + bragging rights forever. friends asked for my "algorithm". taught them probability, expected value, optimization under constraints. turned a game into a statistics masterclass. math = superpowers. 🏏📊
Statistical framework to predict player point potential
Standard deviation tracking to minimize variance risk
Weighted moving average of last 5 matches for current form
Knapsack algorithm to maximize team EV within 100 credits
6 months of player statistics from Cricbuzz and ESPNcricinfo: runs, wickets, strike rates, economy rates, match conditions (home/away, pitch type). Total: 1,200+ player-match records in Excel database.
Calculated: Mean points, Std Dev (consistency), Rolling 5-match average (form), Playing XI probability (injury/selection tracking), EV = Mean × PlayProb. Validated model predictions against actual points weekly.
Knapsack problem: 100 credit budget, 11 players (1 WK, 3-5 batters, 1-3 all-rounders, 3-5 bowlers). Greedy algorithm: sort by ROI, pick top players meeting constraints. Verified no better solution exists via brute force on smaller samples.
After each match: updated database, recalculated EV, adjusted form weights. Discovered bowlers undervalued (average 6.2 ROI vs batters 5.1 ROI). Shifted strategy mid-season—key to climbing leaderboard.
Every great project starts with a personal problem. What's yours?
Get Started