Team Sports Analysis: Data Analysis

In my previous blog I looked at the impact that various team batting statistics have on the ability of a team to score runs. Now what I want to do is to determine how individual players produce runs using the regression results from the team runs created model. Since it looks like I did not do this for the 2012 MLB season, I will report the top 10 players for both the 2012 and 2013 seasons using this runs created methodology. (If you are interested, I have blogged previously on how to calculate MLB Player Production using MLB Team Runs Created).

Taking the data from Sean Lahman's database, here are the top 10 batters for 2012:

	playerID	yearID	teamID	Runs Scored
1	cabremi01	2012	DET	111.51
2	braunry02	2012	MIL	110.95
3	troutmi01	2012	LAA	109.20
4	mccutan01	2012	PIT	104.01
5	fieldpr01	2012	DET	102.52
6	encared01	2012	TOR	98.66
7	beltrad01	2012	TEX	98.31
8	canoro01	2012	NYA	97.68
9	poseybu01	2012	SFN	96.04
10	headlch01	2012	SDN	95.94

Again, taking the data from the same source, here are the top 10 batters for 2013:

	playerID	yearID	teamID	Runs Scored
1	troutmi01	2013	LAA	122.32
2	cabremi01	2013	DET	118.15
3	davisch02	2013	BAL	111.98
4	vottojo01	2013	CIN	105.13
5	goldspa01	2013	ARI	103.18
6	choosh01	2013	CIN	100.50
7	carpema01	2013	SLN	98.38
8	mccutan01	2013	PIT	96.15
9	donaljo02	2013	OAK	91.85
10	canoro01	2013	NYA	91.51

As you can tell Cabrera, Troutman and McCutchen both show up in the top 10 in each year. So the next blog will look at how consistent are MLB players with 100 AB's since 2000.

Monday, March 3, 2014

2013 MLB Team Runs Created Productivity

In my Sports Economics course we look at how to estimate the productivity of MLB players. In order to do that, we first look at estimating the productivity of MLB teams using a model created by Asher Blass that was published in 1992. Almost every year I update his model using MLB data from Sean Lahman's database. So I did the MLB Team Runs Created analysis (step-by-step procedure to estimate MLB Team Runs Created ) for the 2013 season, and here are the results. In step 8, I talk about the statistical analysis called a linear regression adjusting for heteroskedasticity, which is what I am reporting below.

Here are the estimated results for these 14 seasons of data (2000-2013).

Variable	Coefficient	Std. Error	t-Statistic	Prob.
SINGLE	0.523	0.020	26.460	0.000
DOUBLE	0.723	0.041	17.710	0.000
TRIPLE	1.116	0.127	8.785	0.000
HR	1.423	0.039	36.291	0.000
HBP	0.425	0.080	5.303	0.000
SB	0.112	0.038	2.931	0.004
SF	0.621	0.161	3.863	0.000
NBB	0.330	0.019	17.252	0.000
GIDPCS	-0.167	0.063	-2.671	0.008
OUTS2	-0.144	0.025	-5.857	0.000
Constant	152.521	107.078	1.424	0.155

A few observations: first, that other than the constant term, each of the variables is statistically significant at the 99% confidence level and of the correct sign. Only GIDPCS and OUTS are negative, which both are using up the finite and scarce resource in baseball: outs. Second,the coefficient on HR's is greater than the coefficient on Singles, which means that a HR will on average generate more runs than a single. All the other coefficients seem to make sense as well. Finally, note that the coefficient on grounded into a double play plus caught stealing (GIDPCS) is greater in absolute value than a stolen base (SB), which means that if a player made two stolen base attempts, and was successful on one attempt but not on the other, that player on average would have cost his team more than he benefited his team.

Up next is the best hitters for the 2012 and 2013 seasons.