Source http www baseball reference com Project member 1 Tae Kyun Kim SAS data output analyst 2 Hyun Jung Kim project editing 3 Hwan Suk Lee project design mentor Project Design Based on the record of 2008 Major league season we assumed that good batters are batters who has AVG batting average higher than league average Also we assumed that good pitchers are pitchers who has ERA earned runs average lower than league average From the record we set AVG and ERA as our two variables From these two variables we compare each team and counts hitters who are above the entire season of AVG and pitchers who are below the entire season of ERA Then we find out the relationship between wins and both AVG and ERA We only considered batters who attended more than 100 times at bat Also we only considered pitchers who attended equal or more than 40 innings for same reason Batters and hitters who have not been satisfied above these conditions are excluded due to the possibility to be outliers We analyze 2 correlations one is batting AVG and WINS and the other is between ERA and WINS Figure out each correlation to know which factor ERA AVG affect more on the numbers of WINS i Present scatter plot ii Present correlation coefficient Prediction In the baseball game at least the team is going to win at the game when the pitcher played well in the game However no matter how well hitters hit the ball there are still possibilities that the team might lose the game For those reasons some people are saying that pitcher has more influence on the number of wins and teams reputation than the hitters In order to prove this hypothesis we did our project following this prediction Higher correlation coefficient is more effective to number of WINS According to our research we got too low correlations that between number of wins and either our factors AVG or ERA have too weak relationships So we infer there must be some other lurking variables that lower our results To get more clear correlation we added one more variable to each pitcher and hitter variable and find out another correlation coefficient by putting in these two variables in our study WHIP Walks and Hits per Innings Pitched that is related to pitchers ability and OPS On Base Percentage Slugging Percentage that is related to hitters ability Pitcher who has lower WHIP tends not to load bases so make fielders less pressure of defense Thus we assumed that the pitcher who has lower WHIP contribute team s win as same as the pitcher who has lower ERA On the other hand OPS is generally used for defining power hitter Of course power hitter tends to make more extra base hits than just contact hitter and extra base hit has more possibility that runner makes score than just single base hit Hence power hitters have high slugging percentage as well Also for preventing extra base hit and power hitters usually hit so hard and lose their contact pitchers threw more balls to outside of strike zone against power hitter than contact hitter power hitters usually have more four ball than contact hitters Thus we assumed that OPS which is the sum of hitter s On Base Percentage and Slugging Percentage contributes for team s win We only included WHIP that is above the season average and OPS that is below the season average First we found out correlation between Average at bat A VG and number of Wins and also about Earned Runs Average ERA and Wins SAS output indicate that two variables AVG and ERA do not show strong relationship with number of Wins The SAS System 08 54 Monday April 27 2009 Plot of wins batter wins 100 99 98 97 96 95 94 93 92 91 90 89 88 87 86 85 84 83 82 81 80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 1 Symbol used is 60 59 3 4 5 6 7 8 9 10 11 12 13 batter Plot of wins pitcher wins 100 99 98 97 96 95 94 93 92 91 90 89 88 87 86 85 84 83 82 81 80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 Symbol used is 2 3 4 5 6 7 8 9 10 11 12 pitcher Batters and wins point estimate r is 0 18041 Pitchers and wins point estimate r is 0 33924 Patterns of scatter plots do not show the linear pattern which means these two variables do not show consistency Through the result pitchers absolute value of r is higher than that of batters However correlations are too low to evaluate precise effect of between two variables Therefore we can assume that there are other lurking variables that are affect the number of wins So we are including other variables such as WHIP and OPS to see the relationship and correlation with the number of wins So after considering these additional values we get the results Correlation BOPS batter OPS and wins point estimate r is 0 25455 r square is 0 0638 95 Confidence interval is 54 54347 to 87 6493 PWHIP pitcher WHIP and wins point estimate r is 0 50532 r square is 0 3106 95 Confidence interval is 48 69236 to 76 25275 It is clear that the correlation of BOPS and wins is higher than that of batters and wins Also the correlation of PWHIP and wins is higher than that of pitchers and wins What if we contain more variables in our experiments above Are we going to get better results Yes we will get more trusted results Therefore we concluded that It is a fact that good pitchers and batters affect the number of wins However we cannot judge the number of wins by these two or more variables For example teams concentration players patience or endurance the weather of the day etc are influencing variables which can give us better results However we cannot express all variables to the number If we can put all of the variables that can be expressed as number we would get higher correlations SAS code DATA project input name batter pitcher OPS WHIP BOPS PWHIP wins datalines CHC NYM PHI STL FLA ATL MIL COL PIT ARI HOU CIN LAD WSN SFG SDP TEX BOS MIN DET CHW CLE NYY BAL TBR LAA TOR KCR SEA OAK run 9 10 9 8 19 18 97 9 8 8 7 17 15 89 6 9 8 8 14 17 92 11 8 9 6 20 14 86 7 7 8 7 15 14 84 10 10 10 8 20 18 72 6 8 8 6 14 14 90 9 6 10 5 19 11 74 8 5 6 5 14 10 67 4 5 8 9 12 14 82 9 7 5 9 14 16 86 6 6 8 5 14 11 74 9 9 9 7 18 16 84 7 6 8 4 …