## Pitcher Win Values Explained: Part Six

We took a day off from the pitcher win value explanations yesterday so I could help a friend move (when you need to move on a weekday, call the baseball writing friend with the flexible schedule – he’s always available), but we’ll tackle park factors this afternoon and wrap up the series on Monday and Tuesday of next week.

As mentioned earlier, the win values are based on a park adjusted FIP. However, we never covered how we handled the park factor. There are lots of different park factors floating around out there, so I figured it would be useful for us to spend a bit of time talking about them. For those that aren’t aware, a park factor is basically the run environment of a particular ballpark expressed as a decimal, where 1.00 is average. A ballpark with a park factor of 0.90 would depress run scoring by 10%, so that if the league average runs per game is 5.00, then the runs per game in that park would be 4.5. On the flip side, a ballpark with a park factor of 1.10 would have an average of 5.5 runs per game.

Park factors are determined by the relative offensive level between each park and the league average. One of the common misperceptions about park factors is that they will be overly influenced by the home team. However, because the home team plays equal amounts of games per season in their home park and on the road, and the visiting team’s also play 81 games per year in that park, we get a decent sized sample with which to understand how parks affect run scoring.

‘That doesn’t mean that there isn’t noise in a single year’s park factor, however. Let’s take Turner Field in Atlanta as an example, for instance. Here are the single year park factors for that park since 2002:

2002: .88
2003: 1.04
2004: .94
2005: 1.01
2006: 1.02
2007: .95
2008: 1.01

That’s a six year average of .98, which makes it just barely below average in term of runs per game, but it obviously hasn’t been very consistent from year to year. The 2002 to 2003 change, especially, would suggest that the park went from being something like Petco Park to being more like Fenway Park. Most parks don’t have swings that large, but single year park factors can still be a bit unreliable. So, to calculate the win values, we’ve used a five year regressed park factor. For 2008, here are the park factors we used for all thirty teams:

```Season	FullName	         PF
2008	Arizona Diamondbacks	 1.05
2008	Atlanta Braves	         1.00
2008	Baltimore Orioles	 1.01
2008	Boston Red Sox	         1.03
2008	Chicago Cubs	         1.04
2008	Chicago White Sox	 1.04
2008	Cincinnati Reds	         1.02
2008	Cleveland Indians	 0.99
2008	Detroit Tigers           1.00
2008	Florida Marlins	         0.97
2008	Houston Astros	         0.99
2008	Kansas City Royals	 1.00
2008	Los Angeles Angels       0.99
2008	Los Angeles Dodgers	 0.98
2008	Milwaukee Brewers	 1.00
2008	Minnesota Twins	         0.98
2008	New York Mets	         0.97
2008	New York Yankees	 1.00
2008	Oakland Athletics	 0.98
2008	Pittsburgh Pirates	 0.98
2008	San Francisco Giants	 1.01
2008	Seattle Mariners	 0.96
2008	St. Louis Cardinals	 0.98
2008	Tampa Bay Rays	         0.98
2008	Texas Rangers	         1.04
2008	Toronto Blue Jays	 1.01
2008	Washington Nationals	 1.01 ```

Dave is a co-founder of USSMariner.com and contributes to the Wall Street Journal.

### 25 Responses to “Pitcher Win Values Explained: Part Six”

You can follow any responses to this entry through the RSS 2.0 feed.
1. drew says:

You weren’t helping Keith Hernandez move by any chance?

Anywho… I’m suprised the by the Giants and Nationals. I thought they would grade out closer to the Mets/Marlins range.

• The Nationals will change a bit in the following years. It’s just 1 year of data in their new park so it’s regressed more heavily.

• drew says:

That makes sense. I guess I figured that you’d used the data from RFK to figure the value of pitchers in the past years.

Unrelated to pitching but I’m hoping someone can help me with this. If you click teams at the top tab and then select values you can see the batting values of all the teams in 2008. But I don’t understand the postional column. Obviously the AL teams have a negative positional rating that is greater than their NL counterparts (because you’re factoring in the DH) but what makes up the difference between the teams in their leaugues? Thanks in advance.

• For RFK we’re using .96 and it’s 3 years of data and it’s near the bottom, but 2008 is just that 1 year.

That positional column might be going away. I’m not sure it makes a whole lot of sense to have it there as Dave pointed out to me a few weeks ago.

2. Guillermo says:

Chavez Ravine just 2% below the league average?

That doesn’t seem right. All we ever hear is how difficult it is to hit there… But hey, 5 years should be a good sample size.

• Nick says:

Since they cut down on foul territory a few years ago, it plays a bit more neutral overall. Actually, in the summer I think it’s a bit hitter friendly, which is more or less evened out in the early spring and late fall. Don’t have the numbers for this, however. It’s just a guess.

3. Evan says:

Blanket park factors? Really?

I would have thought you’d break them down into their components. HRs, LDs – that sort of thing.

4. TangoTiger says:

You can compare to 2000-06 data:

5. Thor says:

Count me in the “misunderstanding” group, but aren’t park factors still not taking into account synergies between individual players and parks? Suppose (ignoring the matchup problems) that the Mariners stacked their lineup with nothing but left-handed pull hitters. Wouldn’t Safeco’s park factor look bigger than it actually is?
The implication of this in general would be that teams that acquired players that fit their parks offensively would tend to have too-high park factors.

• Colin Wyers says:

The home team is only half of the park factor – the visiting team’s performance counts as well.

6. TangoTiger says:

Thor: you are comparing how those guys do to how they did away from Safeco. It is possible that you get guys like Boggs and Lynn who take particular advantage of Fenway, and therefore your sample of players is not representative of the population in whole. That’s a limitation.

7. Well, simply put, you can make park factors as complicated and as granular as you want them to be. There’s a lot of work still being done on various park adjustments, including park size, weather, etc…

We’re not going to delve into implementing component park factors on FIP or wRAA right now, since I’d guess that the gains in accuracy won’t be earth shattering. If someone wants to run the numbers and show that it’s worth my while to run those calculations, I’m all ears.

• TangoTiger says:

I agree. No matter what you do, the reader will automatically include his own uncertainty range of (at least) +/- 0.25 wins anyway. I really don’t see the point of trying to figure out if Youkilis is +2.5 wins or +2.7 wins, if the reader sees the +2.5 wins and automatically thinks 2.25-2.75, anyway.

If the impact is at least 0.50 wins, then ok, sure. Otherwise, my suggestion is to accept the uncertainty level, so that people can apply the “human factor” that they would anyway.

8. Samg says:

• Nope, that’s already in there. These are the 1+PF/2 versions.

• Samg says:

Thats what I was asking, and thanks!

9. Samg says:

By the way any chance of a speed score?

• I used to calculate Speed Score when the site first launched, but it was taken off for one reason or another. I think there’s a high chance it makes a return this season, maybe in its new 6 component form, though I don’t know if there’s much of a difference between the 6 component and 4 component forms.

10. Samg says:

I only know of the six component form. I think it is a great stat, but maybe some tweaks on the sixth (fielding) factor?

11. OldDogNewTricks says:

I understand park factors and why they are a good idea. But what is meant by “5 year regressed park factor”? How is the regressed part done?

12. Bryan says:

Does anyone try to do park factors that depend on handedness? For example, Safeco doesn’t hurt LHB like Ibanez but it kills RHB like Beltre. LHP do better than RHP for the same reason. Seems like this is a pretty straightforward thing that would dramatically improve their accuracy.

13. Samg says:

Why not adjust every facet of a hitter’s/pitcher’s game separately, because they way they derive value may or may not benefit from his park factor?

14. Samg says:

Any chance of combining the hitting and pitching values for pitchers?

15. VegasPaleHoseFan says:

Are the Park Factors used for 2009 available?