Sunday, July 27, 2014

Factors That Might Determine A Batter's GIDP Rate

I looked at how some variables impact GIDP rate using regression analysis. All data is from Baseball Reference. The variables were

Lefty-A dummy variable for being a lefty (1). Righties got 0. Switch hitters got .67, assuming they face righties about 2/3 of the time

SO%-Strikeouts divided by PA - IBB. Players who strikeout alot probably won't hit into too many DPs

IsoCon-Isolated power on contact. So extra bases divided by AB - SO. I wondered if how hard you hit the ball mattered. Maybe the harder you hit it, the easier it is to turn a DP. But maybe hard hit balls go through more or a line drives or long flies.

Speed-Triples divided by 2B + 3B. This is an idea from Voros McCracken. Fast guys turn those balls hit into the gap into 3Bs, slow guys get 2Bs.

GB/FB ratio-not used in the first regression

I looked at all AL batters who had 800+ PAs combined in 2012-3. Here is the first regression equation:

(1) DP Rate = .185 - .029*Lefty - .178*SO% - .031*IsoCon - .168*Speed

 Here are the t-values for each variable

Lefty) -4.49
SO%) -2.45
IsoCon) -1.04
Speed) -3.03

The r-squared was .393 and the standard error was .029. The mean for all the players in the group (93) was .1122.

I wanted to see how much conventional stats could explain the rate first. Then I added in GB/FB ratio. Here is the second regression equation

(2) DP Rate = .075 - .025*Lefty - .15*SO% + .071*IsoCon - .191*Speed + .096*GB/FB

The r-squared jumped to .599 and the standard error of the regression fell to .024. Here are the t-values

Lefty) -4.61
SO%) -2.51
IsoCon) 1.37
Speed) -4.21
GB/FB) 6.69

So 4 of the variables look significant being greater than 1.96 in absolute value (at the 5% level). Interesting that IsoCon saw a change in sign. The GB/FB ratio sure increased the r-squared alot.

Being a lefty lowers your rate by .025. So if you were otherwise average (.1122), then you fall to .0872. If your strikeout rate goes from .1 to .2, your DP rate falls by .015.

The GB/FB ratio ranged from 1.85 (Jeter) down to .475 (Reddick). A .25 drop in the GB/FB ratio would lower your DP rate by .024. Jeter had the highest DP rate of .245. Kendrick was 2nd with .199. Granderson was lowest with .032

The Speed variable went from .224 (Austin Jackson) down to 0. An increase of .1 here means a drop of .019 in DP rate.

Most of this is not a surprise. It might be worthwhile to add in more years of data. It will also be interesting to see what happens to the sign on the IsoCon variable with more data.


Wednesday, July 23, 2014

Is Adrian Beltre Bending The Aging Curve?

The graph below shows his OPS+ at each age starting at 19 continuing through this year at age 35. Then a table of the actual numbers. 5 of his best 6 seasons have come (if we include this year) at ages 31-35. Except for his age 25 season (in 2004) none of his seasons before age 30 are very close to any of his age 31-35 seasons




19 73
20 102
21 114
22 91
23 97
24 88
25 163
26 93
27 105
28 112
29 109
30 83
31 141
32 131
33 139
34 137
35 143
 
If we did pairs of seasons and average them (just a simple average), the last two are the best


19-20 87.5
21-22 102.5
23-24 92.5
25-26 128
27-28 108.5
29-30 96
31-32 136
33-34 138

I found all the players who had 300+ PAs at every age from 20-35 like Beltre. Then I averaged their OPS+ at each age. Here is the graph with trend line using a polynomial order 2 equation.




Those players were

Sam Crawford       
Ty Cobb                 
Mel Ott                 
Eddie Mathews      
Hank Aaron           
Al Kaline               
Roberto Clemente 
Frank Robinson     
Vada Pinson          
Ted Simmons        
Buddy Bell            
Robin Yount          
Rickey Henderson 
Roberto Alomar     
Ivan Rodriguez      
Alex Rodriguez     

Beltre averaged a 105.18 OPS+ from age 20-30. From 31-35, it is 138.2. So 138.2/105.18 = 1.314. So he has done 31.4% better. Here are the same ratios for the other guys


Roberto Clemente   1.407
Rickey Henderson  1.109
Al Kaline   1.051
Frank Robinson  1.049
Sam Crawford          1.046
Robin Yount  1.039
Hank Aaron  1.026
Roberto Alomar  0.978
Buddy Bell   0.970
Alex Rodriguez  0.961
Mel Ott                    0.936
Ivan Rodriguez  0.923
Ty Cobb                    0.871
Vada Pinson  0.852
Eddie Mathews         0.794
Ted Simmons  0.781

I also used wRC+ from Fangraphs. I found all the guys who had 3000+ PAs through age 30 and 1000+ PAs from 31-35. Then I took their Old to Young ratio. Beltre is near the top, doing about 33% better at his old age. The average ratio for the group was .953, meaning they did 4.7% worse. There were 1286 players at the young age and 781 of them made it to the old age. Here is the top 15


Ken Caminiti 1.505
Bret Boone 1.379
Roberto Clemente 1.348
Augie Galan 1.327
Luis Gonzalez 1.327
Adrian Beltre 1.327
Sammy Sosa 1.325
Jerry Grote 1.320
Hal Chase 1.306
Frank White 1.299
Mark McGwire 1.284
Brady Anderson 1.267
Tony Phillips 1.265
Willie Stargell 1.263
Ozzie Smith 1.250

Monday, July 21, 2014

Top 29 Seasons in FIP ERA Since 1920

From Baseball Reference

Rank
Player (age that year)
FIP ERA
Year
1
Pedro Martinez (27)
1.395
1999
2
Dwight Gooden (19)
1.685
1984
3
Clayton Kershaw (26)
1.739
2014
4
Bob Gibson+ (32)
1.774
1968
5
Sandy Koufax+ (27)
1.852
1963
6
Sandy Koufax+ (29)
1.927
1965
7
Tom Seaver+ (26)
1.931
1971
8
Hal Newhouser+ (25)
1.966
1946
9
Matt Harvey (24)
2.005
2013
10
Steve Carlton+ (27)
2.009
1972
11
Luis Tiant (27)
2.041
1968
12
Felix Hernandez (28)
2.052
2014
13
Bob Moose (20)
2.063
1968
14
Sandy Koufax+ (30)
2.071
1966
15
Sandy Koufax+ (28)
2.076
1964
16
Randy Johnson (31)
2.081
1995
17
Sam McDowell (22)
2.081
1965
18
Don Sutton+ (23)
2.082
1968
19
Bob Veale (29)
2.107
1965
20
Bill Gullickson (22)
2.111
1981
21
Dave Righetti (22)
2.116
1981
22
Dwight Gooden (20)
2.127
1985
23
Randy Johnson (37)
2.128
2001
24
Sandy Koufax+ (26)
2.15
1962
25
Mike Scott (31)
2.157
1986
26
Don Sutton+ (26)
2.157
1971
27
Bob Feller+ (27)
2.157
1946
28
Pedro Martinez (28)
2.171
2000
29
Roger Clemens (25)
2.174
1988