football_fit_v1

Transcript

1. Monica Pangilinan Football Fit
2. Problem: The passing, rushing, and defense of a team are important factors in winning games. But how valuable is each factor in making the playoffs? •Is the rushing (RB), passing (QB) or defense making my team win? •Are they still performing at a high level? •Do they contribute to my bottom line of making it to the playoffs? M. Pangilinan 2
3. DATA/ALGORITHM • Training set: 2002 - 2010 play by plays from compiled game logs to extract variables based on main game components: passing, rushing, and defense • Random forest trained to classify using the variables of the team and their opponent • Validation using 2010 - 2012 dataset M. Pangilinan 3
4. Passing, Rushing, and Defense information for team and their future opponents Random forest predicts win/loss of remaining games Product Data/Algorithm Flow: M. Pangilinan 4
5. VALIDATION Training Set (2002 - 2010) Testing Set (2010 - 2012) 85% accuracy 82% precision 68% accuracy 67% precision M. Pangilinan 5
6. TOP FEATURES: 2002 - 2010 M. Pangilinan 6
7. TOP FEATURES OVER TIME 2002 - 2005 2005 - 2010 • Passing importance increased over time M. Pangilinan 7
8. POSITION VALUE OVER TIME •increase in total yards from passing rather than from the rushing M. Pangilinan 8
9. Monica Pangilinan
10. Back up slides M. Pangilinan 10
11. WEBSITE (INPUT) M. Pangilinan 11
12. M. Pangilinan 12 WEBSITE (OUTPUT)
13. DATA/ALGORITHM • Data from csv files compiled here and appears to be from ESPN game logs: http://archive.advancedfootballanalytics.com/2010/04/play-by-play-data.html • Random Forest: 50 estimators with 20 minimum events required in each node with 54 features used • Training set: ~2000 events M. Pangilinan 13
14. Feature ranking (Importance): 1. DEFOppAveTotRushYds (0.087999) 2. TEAMAveScore (0.050015) 3. TEAMOppAveScore (0.045839) 4. DEFOppAveTotPassYds (0.040182) 5. DEFAveTotRushYds (0.036854) 6. TEAMOppAveWin (0.036126) 7. QBAveQBRating (0.035942) 8. QBAveTotYds (0.034374) 9. QBOppAveQBRating (0.033573) 10. DEFOppAveSackYds (0.031315) 11. TEAMAveWin (0.030825) 12. DEFOppAveTotSacks (0.026769) 13. DEFAveTotPassYds (0.026061) 14. DEFAveSackYds (0.023985) 15. QBAveTotTD (0.023011) 16. QBOppAveTotYds (0.022486) 17. DEFOppAveTotInterception (0.021540) 18. QBAveSackYds (0.021090) 19. QBOppAveSackYds (0.020247) 20. DEFAveTotInterception (0.018266) 21. TEAMAveOppScore (0.018138) 22. TEAMOppAveOppScore (0.017857) 23. QBOppAveCompletions (0.017364) 24. QBOppAveTotTD (0.016749) 25. QBAveAttempts (0.015207) 26. QBOppAveTotSacks (0.015057) 27. DEFAveTotSacks (0.014287) 28. QBOppAveAttempts (0.013892) 29. RBOppAveTotYds (0.013256) 30. RBAveTotYds (0.012546) 31. QBAveCompletions (0.012189) 32. QBOppAveRushYds (0.012161) 33. QBAveTotSacks (0.011961) 34. RBAvePassAttempts (0.011956) 35. RBAveAttempts (0.011492) 36. DEFOppAveFumbles (0.010831) 37. DEFAveFumbles (0.010802) 38. RBOppAveAttempts (0.010164) 39. QBAveRushYds (0.009261) 40. QBAveTotRush (0.008844) 41. QBOppAveTotInterception (0.008620) 42. RBOppAveTotTD (0.008496) 43. QBAveTotInterception (0.008446) 44. QBOppAveFumbles (0.007896) 45. RBOppAvePassAttempts (0.007505) 46. QBAveFumbles (0.007382) 47. RBOppAveFumbles (0.006138) 48. QBOppAveTotRush (0.006002) 49. RBAveTotTD (0.004612) 50. RBAveFumbles (0.004391) 51. RBAvePassTotYds (0.000000) 52. RBAveTotInterception (0.000000) 53. RBOppAvePassTotYds (0.000000) 54. RBOppAveTotInterception (0.000000) TOP FEATURES: 2002 - 2010 M. Pangilinan 14
15. POSITION VALUE OVER TIME M. Pangilinan 15