Predicting and Preventing Injury in Major League Baseball

An analytical study of baseball analysis trends

Transcript

1. Baseball Injuries Renuka Kulkarni-Kelapure
2. What are Injury and Risk Factors? Injury Event that causes changes in the environment of cell leading to its malfunction External factors ex.-environment, country of origin, overuse etc. Internal Factors ex- genetic, BMI, age, etc.
3. Is this really about Baseball though? — Data Gathering — Cleaning — Metric analysis — Findings — Recommendations
4. Data Source Players Injuries Player Statistics Scraping for player country data
5. Data Cleaning and Formatting Literally needed these many tools and 4-5 iterations to clean this data!!! No more Mess Plz!
6. How did we do it? — rbind- to make the dataframes from list — Join- to join the dataframes in each list to get one data from — Formatted each columns in three dataframes Packages used- lubridate, plyr, devtools, stringr, scrapeR Other functions - filtering, ymd, %in%. Ifelse, gsub, grep, grepl, as.factor, as. Integer,
7. Data is no fun without problems! — Data variables- 54- too many variables — Mixed variables- Categorical and Numerical — R couldn’t recognize foreign letters — Web scraping challenges-^A — Choosing proper analysis method — Corplot applies for continuous variables only — Missing data
8. Final Data — Decided to use only variables which have less covariance among each other — No use of performance variables — Created a matrix of Continuous variables — Corplot — Mixedcorplot- tried for all variables- Failed — Use Excel to fixed web scraping issues — Decided to do separate analysis for categorical variable
9. Corrplot variables • Hitters games played • Pitches faced • BMI • Age • Salary height weight Fielders games played Years of experience
10. CorPlot
11. Corrplot Analysis — Dependent Variable- Injury Status- Binomial 0- No injury 1- Injury Only 3 variables are show weak correlation with Injury Status (BMI, Years of experience, Games played by players) All the correlations looks Weak
12. Logistic Regression Analysis Model#1- Is injury correlated with BMI Years of experience Age BMI salary Games played
13. Deviance Residuals: Min 1Q Median 3Q Max -2.174 -1.195 0.697 0.947 1.579 Coefficients: e stimate Std. Error z value Pr(>|z|) (Intercept) 3.899e+00 1.600e+00 -2.436 * PLITpstat_2$years_of_experience 1.327e-01 5.739e-02 2.313 PLITpstat_2$humanized_salary -1.549e-08 2.648e-08 -0.585 0.558484 PLITpstat_2$age -4.931e-03 4.501e-02 - 0.110 0.912766 PLITpstat_2$BMI 1.547e-01 4.397e-02 3.517 PLITpstat_2$pitches_thrown 3.871e-04 1.429e-04 2.709 PLITpstat_2$pitcher_games_played -1.315e-02 1.070e-02 - 1.229 0.219109 PLITpstat_2$fielder_games_played -9.353e-03 1.190e-02 -0.786 0.431792 PLITpstat_2$hitter_games_played 8.238e-03 7.464e-03 1.104 0.269689 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for binomial family taken to be 1) Null deviance: 635.00 on 480 degrees of freedom Residual deviance: 591.84 on 472 degrees of freedom (1775 observations deleted due to missingness)
14. Models Model 2 Model 3 Experience, age, BMI BMI, Experience Z score=0.01789 * Z score= 2e-16 *** Z Score= not significant Z score=0.00110 ** Z score= 0.00542 ** AIC: 733.97 AIC: 2628.1 Missing-1681 observations Missing- 164 observations R2- R2-
15. Scatter Plots- Injury Vs BMI
16. Injury vs Experience
17. Categorical Analysis — Null Hypothesis- Injury is not associated to batting hand, education, country of birth — Test- Chi – Square test of independence
18. Chi-square test Bat Highschool Undergrad Country of birth Position X-squared = 30.829 X-squared = 2.0658e-05 X-squared = 13.1621 X-squared = 2.0339, X-squared = 236.75, P- 9.234e-07 p-value = 0.9964 p-value = 0.0002857 p-value = 0.1538 p-value < 2.2e-16 Df-3 Df-1 Df-1 Df- 1 Df- 4
19. Upper Extremiti e Lower Extremiti es Abdomen Back Neck Chest General Head face Noninjury 17.37% 15.91% 1.28% 5.27 2.08% 1.19% 16.17%
20. Injuries with respect to Position in game F H P R — Abdomen 8 1 17 3 — Chest Neck Back 54 2 41 22 — General 12 1 12 22 — Head and Face 13 1 6 7 — Lower extremity 179 15 84 81 — Noninjury 193 31 11 130 — Upper extremity 119 4 2 13 56
24. Descriptive Analysis — #No of obs- 2256- all players16- 1669 — # No of Injured players Unique-756 — #Most Injuries - Gerald Posey- 17 records — #Team with Most injuries-Dodgers -100 records. — #Unique Healthy players- 913 — Incidence of injuries- 45% — No of players healthy during season – 55%
25. BMI for Injured — BMI-
26. BMI for healthy
29. Recommendations Conclusions • BMI and year of experience are weakly correlated to injuries. • Position of player in the game is also associated with injuries. • Need more research with education level, BMI, Position, handedness.
30. References used — https://www.beyondtheboxscore.com/2014/11/16/722584 7/looking-at-who-played-in-all-162-games — https://www.ncbi.nlm.nih.gov/pubmed/19826310 — http://m.mlb.com/glossary/standard-stats