THE BOOK--Playing The Percentages In Baseball

That’s nifty, Thanks!  Do you think that’s a better way than normal multiplicative park factors?

It’d be interesting to see splits based on that that methodology by the handedness of the pitchers and hitters.  You’d probably need to use LWTS or BaseRuns splits instead of actual runs.

The best way would be odds ratio at the component level (say park effect of “Fenway on lefthanded FB hitters”, among a multitude of components).  Once you have all that, a combination of Odds Ratio and BaseRuns. 

Otherwise, I like differentials.  It means more to me than seeing “1.04”.

Sorry if this is stupid and wrong, but isn’t it “So a Padres pitcher’s ERA is about 0.50 LOWER than it should be”?  My understanding is that the SD park is a severe pitcher’s park, so any pitcher pitching in SD should have their ERA lower than at a neutral ballpark.  Thus 0.50 should be added to his ERA.

I’m the one who is silly.  I should have said “better” not “higher”.

And here’s one for BIP, similar to what I did for 1999-2002, but this time, it’s for 2000-2006.

Coors and Fenway still lead the way (harder to convert a BIP into an out in those ballparks).  Nats are way in the bottom.

Park hTeam BIP
DEN02 COL 0.039
BOS07 BOS 0.018
KAN06 KCA 0.013
PHO01 ARI 0.013
MIN03 MIN 0.009
CIN08 CIN 0.008
ARL02 TEX 0.007
PIT08 PIT 0.006
MON02 MON 0.005
TOR02 TOR 0.004
ATL02 ATL 0.004
ANA01 ANA 0.004
HOU03 HOU 0.003
PHI13 PHI 0.002
SFO03 SFN 0.001
DET05 DET -0.001
STP01 TBA -0.002
STL09 SLN -0.002
MIA01 FLO -0.002
CLE08 CLE -0.002
NYC17 NYN -0.004
CHI12 CHA -0.006
CHI11 CHN -0.006
CIN09 CIN -0.007
PHI12 PHI -0.007
NYC16 NYA -0.007
SJU01 MON -0.008
MIL06 MIL -0.009
OAK01 OAK -0.011
SAN01 SDN -0.012
STL10 SLN -0.012
BAL12 BAL -0.012
PIT07 PIT -0.015
LOS03 LAN -0.015
SEA03 SEA -0.016
SAN02 SDN -0.016
WAS10 WAS -0.020
MIL05 MIL -0.022

This one I call “visibility”, and it’s K minus non-intentional walks per PA.

In San Diego, it’s alot easier to get lots of K (and/or be low in the walks) than elsewhere.  Colorado is a the bottom. 

Park hTeam VISIBILITY
PIT07 PIT 0.023
SAN01 SDN 0.022
PHI12 PHI 0.021
WAS10 WAS 0.019
MIA01 FLO 0.015
MIN03 MIN 0.014
CIN09 CIN 0.014
MIL06 MIL 0.014
CHI11 CHN 0.011
SAN02 SDN 0.011
STP01 TBA 0.010
NYC17 NYN 0.010
SJU01 MON 0.009
LOS03 LAN 0.008
SEA03 SEA 0.006
HOU03 HOU 0.002
CLE08 CLE 0.001
NYC16 NYA 0.001
MON02 MON 0.000
STL09 SLN 0.000
ATL02 ATL -0.001
PHI13 PHI -0.002
TOR02 TOR -0.003
SFO03 SFN -0.003
PIT08 PIT -0.003
ANA01 ANA -0.004
OAK01 OAK -0.004
BOS07 BOS -0.006
CHI12 CHA -0.006
ARL02 TEX -0.007
STL10 SLN -0.008
MIL05 MIL -0.010
BAL12 BAL -0.011
CIN08 CIN -0.011
DET05 DET -0.011
KAN06 KCA -0.018
PHO01 ARI -0.020
DEN02 COL -0.023

I call this “CONTACT”.  A contact rate is 1 minus (K+BB)/PA.  The higher the contact rate, the more the ball is… contacted.  Think of the high contact rate to be a pitcher like Brad Radke, and a low contact rate to be a pitcher like RJ.

Coors makes hitters contact the ball more.  In Philly, lots of sitting and waiting.  Seeing that VISIBILITY is average in Philly, this means hitters are getting both more K and more walks in Philly.

Park hTeam CONTACT
KAN06 KCA 0.022
DEN02 COL 0.020
PHO01 ARI 0.017
SFO03 SFN 0.016
DET05 DET 0.013
PIT08 PIT 0.013
ANA01 ANA 0.012
BAL12 BAL 0.010
PHI13 PHI 0.009
BOS07 BOS 0.009
ARL02 TEX 0.008
OAK01 OAK 0.007
HOU03 HOU 0.006
MON02 MON 0.006
STL10 SLN 0.006
NYC16 NYA 0.003
MIL05 MIL 0.003
ATL02 ATL 0.002
CIN08 CIN 0.000
CHI12 CHA -0.001
SJU01 MON -0.001
NYC17 NYN -0.002
CHI11 CHN -0.002
TOR02 TOR -0.004
STL09 SLN -0.007
LOS03 LAN -0.007
WAS10 WAS -0.007
MIN03 MIN -0.008
CLE08 CLE -0.010
STP01 TBA -0.010
SAN02 SDN -0.014
SAN01 SDN -0.015
CIN09 CIN -0.016
PIT07 PIT -0.020
SEA03 SEA -0.022
MIL06 MIL -0.023
MIA01 FLO -0.031
PHI12 PHI -0.039

TRIPLES: rate of triples per triples plus doubles. 

Much harder to hit triples in both Cincy parks.  Kinda weird.  Easiest to do it in Detroit.  If a guy has 1 triple and 14 doubles on the road, he’ll have 2 and 13 in Detroit.

Like I said, I don’t think the additive approach is the correct one.  The Odds Ratio would be better.  So, if the typical 3B to 2B rate is 0.10, then it would be something like 0.17 in Detroit.  And if some player has a ratio of 0.30 3B to 2B, then it would be 0.50 in Detroit.  So, a guy with 6 triples and 20 doubles on the road would be 9 triples and 17 doubles in Detroit.

Park hTeam TRIPLES
DET05 DET 0.068
SAN02 SDN 0.044
MIA01 FLO 0.037
PHO01 ARI 0.033
DEN02 COL 0.031
SFO03 SFN 0.031
HOU03 HOU 0.029
WAS10 WAS 0.024
SAN01 SDN 0.019
STP01 TBA 0.018
MIN03 MIN 0.016
ARL02 TEX 0.015
PIT07 PIT 0.011
KAN06 KCA 0.004
MIL06 MIL 0.003
PHI13 PHI 0.003
PHI12 PHI 0.002
MIL05 MIL -0.003
TOR02 TOR -0.008
ATL02 ATL -0.008
CHI12 CHA -0.010
STL10 SLN -0.012
CHI11 CHN -0.012
SEA03 SEA -0.016
OAK01 OAK -0.016
STL09 SLN -0.017
MON02 MON -0.018
NYC16 NYA -0.018
BOS07 BOS -0.020
PIT08 PIT -0.021
ANA01 ANA -0.022
BAL12 BAL -0.022
NYC17 NYN -0.024
CLE08 CLE -0.034
SJU01 MON -0.038
LOS03 LAN -0.039
CIN08 CIN -0.049
CIN09 CIN -0.053

***

Hmmm… I think the VISIBILITY should have the sign reversed.  The more K and the fewer walks means that visibility is not as good for the hitter.

HOMERS: homeruns per “AB plus SF minus K”

Coors near the top.

Park hTeam HOMERS
CHI12 CHA 0.012
DEN02 COL 0.010
CIN09 CIN 0.008
PHI13 PHI 0.008
ARL02 TEX 0.006
SJU01 MON 0.006
HOU03 HOU 0.006
CIN08 CIN 0.005
MIL06 MIL 0.004
TOR02 TOR 0.004
PIT07 PIT 0.004
PHO01 ARI 0.004
NYC16 NYA 0.004
LOS03 LAN 0.003
MON02 MON 0.003
CHI11 CHN 0.002
BAL12 BAL 0.001
OAK01 OAK 0.000
ATL02 ATL -0.001
STL09 SLN -0.001
KAN06 KCA -0.002
CLE08 CLE -0.002
PHI12 PHI -0.002
STP01 TBA -0.002
ANA01 ANA -0.003
SEA03 SEA -0.003
MIN03 MIN -0.004
MIA01 FLO -0.004
NYC17 NYN -0.005
STL10 SLN -0.006
BOS07 BOS -0.006
SAN02 SDN -0.006
SAN01 SDN -0.006
PIT08 PIT -0.006
WAS10 WAS -0.006
SFO03 SFN -0.010
DET05 DET -0.010
MIL05 MIL -0.012

GAP: doubles plus triples per nonHR hit.

Fenway near the top. 

Park hTeam GAP
MON02 MON 0.044
PIT07 PIT 0.040
PHI12 PHI 0.034
BOS07 BOS 0.030
TOR02 TOR 0.026
PHO01 ARI 0.025
CIN08 CIN 0.017
MIL06 MIL 0.015
ARL02 TEX 0.014
MIN03 MIN 0.010
PIT08 PIT 0.005
STP01 TBA 0.005
KAN06 KCA 0.003
CLE08 CLE 0.003
DEN02 COL 0.003
OAK01 OAK 0.003
SFO03 SFN 0.002
HOU03 HOU 0.002
MIA01 FLO 0.001
STL09 SLN -0.001
WAS10 WAS -0.002
PHI13 PHI -0.003
CIN09 CIN -0.007
CHI11 CHN -0.008
CHI12 CHA -0.009
SAN02 SDN -0.011
MIL05 MIL -0.011
ATL02 ATL -0.012
NYC17 NYN -0.013
SEA03 SEA -0.015
ANA01 ANA -0.016
NYC16 NYA -0.017
SJU01 MON -0.020
DET05 DET -0.021
LOS03 LAN -0.027
STL10 SLN -0.029
SAN01 SDN -0.033
BAL12 BAL -0.036

STEALS: SB - 2*CS per Opp, where Opp = 80% of singles plus 60% of NIBB and HBP

Not much success stealing in San Diego.

Park hTeam STEALS
PIT07 PIT 0.021
CIN08 CIN 0.018
MIN03 MIN 0.015
KAN06 KCA 0.011
NYC17 NYN 0.011
LOS03 LAN 0.010
PHI12 PHI 0.010
SEA03 SEA 0.009
TOR02 TOR 0.008
MON02 MON 0.007
WAS10 WAS 0.007
NYC16 NYA 0.007
CIN09 CIN 0.006
HOU03 HOU 0.005
CHI12 CHA 0.005
MIL05 MIL 0.004
PHI13 PHI 0.003
SFO03 SFN 0.003
ANA01 ANA 0.001
DEN02 COL 0.000
PIT08 PIT -0.001
STP01 TBA -0.002
MIL06 MIL -0.002
PHO01 ARI -0.002
ATL02 ATL -0.003
STL10 SLN -0.004
OAK01 OAK -0.004
ARL02 TEX -0.005
STL09 SLN -0.006
CLE08 CLE -0.006
BAL12 BAL -0.007
DET05 DET -0.008
BOS07 BOS -0.012
CHI11 CHN -0.013
MIA01 FLO -0.018
SAN02 SDN -0.018
SJU01 MON -0.019
SAN01 SDN -0.027

***

I think this concludes all the basic data.

Here’s your Google Docs data all in one place.

Found a little bug.  Ignore previous, and use
Google Docs

what process are you using to determine the time period that teams played their road games, in terms of assigning to their current home park?

IE, the Cards first home game at new Busch wasn’t until 4/10/06 - is their first week of games from 06 being assigned to new or old Busch?

Thanks

It’s whatever Park ID is determined by Retrosheet:

http://www.retrosheet.org/boxesetc/S/PK_STL09.htm
http://www.retrosheet.org/boxesetc/S/PK_STL10.htm

From what you are saying, Retrosheet has it wrong?  If so, please provide a reference for your statement, and I’ll post it on the Retrolist group.

Oh, I think I misunderstood you.  The “time period” for road parks is by year.

I am coming up with slightly, slightly different numbers.

In Coors, for example…

At home:
(3395+3605)/(30066/54) = 12.572

On road:
(2303+2795)/(30120/54) = 9.140

Run Diff = 12.572 - 9.140 = 3.430

Am I going about this the wrong way?

hittingRuns/hittingOuts + pitchingRuns/pitchingOuts, home and road., rather than (hittingRuns+pitchingRuns)/(hittingOuts+pitchingOuts)

Even my way is a little off, because of runners left on base, if the game ends in the bottom of an inning.

got it - thanks!

David comes up with more component park factors:
http://www.hardballtimes.com/main/article/batted-balls-and-park-effects/

Seeing SD so hard to steal at prompted me to look at their catchers from 00-06, I found it interesting and thought I’d share:

I am not sure where to find opportunities for each though.

bumping

Bumping…

Tom, have you thought about trying to account for the unbalanced schedule in these factors? 

For example, because of the unbalanced schedule, Texas plays a disproportionate percentage of its road games in Seattle/Oakland/Anaheim, all of which are negative run environments.  This would tend to make their home park look even more run-friendly than it most likely is.

Also, I expect you got an answer already about GABP and triples, but the answer is clearly the closer outfield fences suppress triples… and the opposite of course for Comerica Park…

I’ve thought about it, sure.  When I do, I then think I should focus on the players on those teams (not presume that the A’s will find an average team at each park).  And then I pause and wonder if I should go there.

"not presume that the A’s will find an average team at each park”

As long as they face the same team at home and on the road in around the same number of games, what difference does it make?

Tango, when I go to your Google Docs page, there doesn’t seem to be a way to download and save that file.

I can’t access Google docs from teh office.  I’ll have to check tonight.

Remember a “team” is a collection of players.  Just because the A’s play a certain set of players at one park doesn’t mean they play the same set of players at another park.  I’m thinking, specifically, of pitchers, like Greinke or Felix or Peavy who are disproportionately represented relative to their teammates at various parks.

Royals+Greinke is over a .600 team
Greinke+other is under a .400 team

(Or something like that.)

So, it makes no sense to lump the Royals+Greinke with Royals without Greinke, when they really should count like a Yankee team or something.

That’s what I mean: we need to look at the actual players in that game.

Sure, I agree that you want to adjust for pools of players, which is not that hard to do, but even without that, you are still way better off adjusting for the imbalanced schedule than not adjusting for it.

If you adjust for the schedule, you should also adjust for the league, right?

As for “way better off"… well, I’m a pretty simple guy.  I like my Marcel, I like my LWTS, I like my (K-BB)/PA, I like my WOWY.  And when I throw big samples at it, well, there’s not that much imbalance left is there?  If I do a park factor 1998-2009 the basic way like I’m doing it here, won’t that pretty much get me where I want to be (other parks changing, notwithstanding)?

Try this MGL: run UZR for 2002-09… the full UZR.  Now, strip EVERYTHING out, except batted ball location.  The basic UZR.  All the other adjustments you make, remove them. Now, tell me what was Jeter’s UZR in both ways?

There is no real argument here.  We are both in agreement that adjusting for the schedule would be an improvement and not one with any real disadvantages, as some adjustments can have.

Certainly the more the spread in park factors and the more extreme the schedule is, the more it would become necessary to do the adjustments.

But in the end, we are left with “how much difference does it make?” That is a matter of personal opinion. I was being a tad hyperbolic when I said “way better” I guess.

I am not crticizing your work or your opinion on whether it is “worth it” to do schedule adjustments.  To each his own.  To me, when something can be made better rather easily and “cleanly,” there is little downside to it, and the gain is more than infinitesimal (again, even that is a judgment call), I usually try and do it.  If nothing else, you forestall the people who say, “What about the imbalanced schedule?”

It is not right to just tell them, “Well, it won’t make much difference, so don’t worry about it.” You have somewhat of a responsibility to show them that it won’t make much difference.

Do you guys have an accurate list of ballpark factors that you use (other than the commonly found 3 yr park averages)?

Not really.  I do my own stuff.  There is a Yahoo group dedicated to park factor issues run by Patriot. I am not sure if it is private or not.  If you Google Patriot and park factors you will probably come up with lots of stuff.  There are definitely more than the basic 3-year PF’s out there.  I am not very good at remembering or recording where all this stuff is located.  Tango is much better than I am with that stuff..

MGL, what do you use for your own park factors (referring to your own stuff)?

What do you mean?  You mean my actual numbers?  I don’t release those publicly.  Sorry.

I believe the park factor Yahoo group MGL mentions is run by KJOK.  But I shouldn’t have said anything because KJOK does great work.

Seems very difficult to find accurate park factors around here, patriot you have any advice as where to look?

What does “accurate park factors” mean?  What’s wrong with what I’ve published?

If you mean it’s outdated, that’s ok.  But, they are accurate.