MLB Pitching Data

MLB Pitchers and their statistics 1871-2015

Problem

Though may not be considered a problem, there is a lot that can be learned through Major League Baseball pitching statistics.It is of the best interest of the pitcher to lower their ERA and win more games, as well as lower other major stats.

Solution

The dataset has 30 different categories of statistics and identifiers of individual players (by season played). Through looking at many advanced statistics we can help to determine what is most important to focus on to provide a team with the best chance to win their game and let up the least number of runs possible (the ultimate goal of baseball). **The goal of this project is to identify and use a relationship between Wins and Earned Runs Average and other major pitching stats to teach pitchers how best to succeed.

Impact

If good correlations are found, then the best statistics for predicting pitching success will be shown and focused on for pitchers at all levels to imporve

import pandas as pd

#reads in data set as a whole
df_pitching = pd.read_csv("Pitching.csv")

#visualization of the complete dataframe
df_pitching

	playerID	yearID	stint	teamID	lgID	W	L	G	GS	CG	...	IBB	WP	HBP	BK	BFP	GF	R	SH	SF	GIDP
0	bechtge01	1871	1	PH1	NaN	1	2	3	3	2	...	NaN	NaN	NaN	0	NaN	NaN	42	NaN	NaN	NaN
1	brainas01	1871	1	WS3	NaN	12	15	30	30	30	...	NaN	NaN	NaN	0	NaN	NaN	292	NaN	NaN	NaN
2	fergubo01	1871	1	NY2	NaN	0	0	1	0	0	...	NaN	NaN	NaN	0	NaN	NaN	9	NaN	NaN	NaN
3	fishech01	1871	1	RC1	NaN	4	16	24	24	22	...	NaN	NaN	NaN	0	NaN	NaN	257	NaN	NaN	NaN
4	fleetfr01	1871	1	NY2	NaN	0	1	1	1	1	...	NaN	NaN	NaN	0	NaN	NaN	21	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
44134	youngch03	2015	1	KCA	AL	11	6	34	18	0	...	0.0	5.0	0.0	0	500.0	3.0	44	4.0	2.0	NaN
44135	zieglbr01	2015	1	ARI	NL	0	3	66	0	0	...	3.0	2.0	1.0	0	263.0	46.0	17	1.0	0.0	NaN
44136	zimmejo02	2015	1	WAS	NL	13	10	33	33	0	...	3.0	2.0	8.0	1	831.0	0.0	89	8.0	2.0	NaN
44137	zitoba01	2015	1	OAK	AL	0	0	3	2	0	...	0.0	0.0	0.0	0	37.0	1.0	8	0.0	0.0	NaN
44138	zychto01	2015	1	SEA	AL	0	0	13	1	0	...	0.0	1.0	2.0	0	76.0	4.0	6	0.0	0.0	NaN

44139 rows × 30 columns

Data

Kaggle Dataset of Pitching Statistics to observe the features as displayed in the dictionary created below.

#I have decided to drop a large number of columns to only focus on statistics and identification that can be useful to 
#draw conclusions that can help pitchers improve
df_pitching.drop(["stint", "GS", "H", "HR", "IBB", "WP", "HBP", "BK", "BFP", "GF", "R", "SH", "SF", "GIDP"], axis = 1, inplace = True)

#creates a dictionary that describes the data that is in the Dataframe, telling what each column would represent
pitching_dict = {}
pitching_dict["playerID"] = "The first five letters of the pitchers last name,then the first two letters of their first name, followed by a number starting from 01 and increasing for each repeated name."
pitching_dict["yearID"] = "The year represented that the player pitched in."
pitching_dict["teamID"] = "Team that the player is on represented by a 3 letter symbol."
pitching_dict["IgID"] = "Conference that the team is a part of."
pitching_dict["W"] = "Number of wins."
pitching_dict["L"] = "Number of losses."
pitching_dict["G"] = "Number of Games."
pitching_dict["CG"] = "Number of complete games pitched."
pitching_dict["SHO"] = "Number of complete games that were a shutout."
pitching_dict["SV"] = "Number of saves."
pitching_dict["IPouts"] = "Number of innins pitched with outs as a decimal."
pitching_dict["ER"] = "Number of Earned Runs."
pitching_dict["BB"] = "Number of walks."
pitching_dict["SO"] = "Number of strikeouts."
pitching_dict["BAOpp"] = "Batting average of the opponent."
pitching_dict["ERA"] = "Earned Run Average."

#displays the dictionary
pitching_dict

{'playerID': 'The first five letters of the pitchers last name,then the first two letters of their first name, followed by a number starting from 01 and increasing for each repeated name.',
 'yearID': 'The year represented that the player pitched in.',
 'teamID': 'Team that the player is on represented by a 3 letter symbol.',
 'IgID': 'Conference that the team is a part of.',
 'W': 'Number of wins.',
 'L': 'Number of losses.',
 'G': 'Number of Games.',
 'CG': 'Number of complete games pitched.',
 'SHO': 'Number of complete games that were a shutout.',
 'SV': 'Number of saves.',
 'IPouts': 'Number of innins pitched with outs as a decimal.',
 'ER': 'Number of Earned Runs.',
 'BB': 'Number of walks.',
 'SO': 'Number of strikeouts.',
 'BAOpp': 'Batting average of the opponent.',
 'ERA': 'Earned Run Average.'}

#displays the updated dataframe with the correct number of columns that were detailed above
df_pitching

	playerID	yearID	teamID	lgID	W	L	G	CG	SHO	SV	IPouts	ER	BB	SO	BAOpp	ERA
0	bechtge01	1871	PH1	NaN	1	2	3	2	0	0	78.0	23	11	1	NaN	7.96
1	brainas01	1871	WS3	NaN	12	15	30	30	0	0	792.0	132	37	13	NaN	4.50
2	fergubo01	1871	NY2	NaN	0	0	1	0	0	0	3.0	3	0	0	NaN	27.00
3	fishech01	1871	RC1	NaN	4	16	24	22	1	0	639.0	103	31	15	NaN	4.35
4	fleetfr01	1871	NY2	NaN	0	1	1	1	0	0	27.0	10	3	0	NaN	10.00
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
44134	youngch03	2015	KCA	AL	11	6	34	0	0	0	370.0	42	43	83	0.202	3.06
44135	zieglbr01	2015	ARI	NL	0	3	66	0	0	30	204.0	14	17	36	0.197	1.85
44136	zimmejo02	2015	WAS	NL	13	10	33	0	0	0	605.0	82	39	164	0.264	3.66
44137	zitoba01	2015	OAK	AL	0	0	3	0	0	0	21.0	8	6	2	0.387	10.29
44138	zychto01	2015	SEA	AL	0	0	13	0	0	0	55.0	5	3	24	0.239	2.45

44139 rows × 16 columns

Potential Problems

I feel that some of the data could be problematic and be too easily predictable. Using a statistic such as ERA is literally telling the number of runs that a pitcher allows. It is the hope of this to surface a statistic that is more telling and can direct a pitcher to a trait or attribute that should be improved

Method

We will determine groups of levels of pitching and the statistics that usually contribute to these areas. From this we will have numerous regressions and look for the best regressions for each category. For example a level could be pitchers that win 70% of their games and a large number of SO (strikeouts) could have the largest correlation with it.