UZR - and Park effects

[boomboom]

I understand that park effects change each season. What doesn't make any sense is why they change. Unless the park dimensions actually change, the park effects shouldn't So... why do they use a short sample size of 1 year when in fact they should use something like the last 10-20 years or since the park opened effect.

Most parks haven't changed dimensions, if a new park opens or a old park changes the fences then sure, redo the park effects and discount anything before that time.

The variance of a park going from a hitters park to a pitchers park in a matter of one season I think comes down to one thing, luck. They should use more than one season of data when they create park effects.

[HoustonGM]

Baseball-Reference uses three-year park factors for their park-adjusted stats, and every park-adjusted stat I've seen that explains their park factors does so as well. I would assume that UZR does as well.

From the update article (http://www.fangraphs.com/blogs/index.php/uzr-updates/):

Of course, park factors in general are updated every year, as we get more data in each park, and as new parks come into existence and old parks make material (to fielding) changes.

This sounds to me as if the park factors are UPDATED every year to account for the new information, not that they use solely single-year factors each year.



Most parks haven't changed dimensions, if a new park opens or a old park changes the fences then sure, redo the park effects and discount anything before that time.

The variance of a park going from a hitters park to a pitchers park in a matter of one season I think comes down to one thing, luck. They should use more than one season of data when they create park effects.
Most parks don't have large swings from year to year. I don't recall any park that, without actual changes, has swung from a good hitter's park to a good pitcher's park and back in the matter of one season. As for what can cause changes outside of actual physical changes, it's not solely "luck", which should really just be called statistical variation, but there's also things like changing weather patterns that can influence it.

[200tang]

Well, if we're to assume they never change, then using 1 year of data should make no difference, correct?

Second, we're trying to figure out what happened last year, weather is a huge factor. Take Wrigley field for example where one night it can play like a pitchers park and the next night wind will be blowing like crazy out to LF. Also, parks change minor things a lot more than one might think, even something like changing the batters eye can affect things like strikeouts.

[kenny1234]

Well, if we're to assume they never change, then using 1 year of data should make no difference, correct?

Second, we're trying to figure out what happened last year, weather is a huge factor. Take Wrigley field for example where one night it can play like a pitchers park and the next night wind will be blowing like crazy out to LF. Also, parks change minor things a lot more than one might think, even something like changing the batters eye can affect things like strikeouts.
If one year of fielding data is not sufficient to judge the quality of a player, then it also isn't sufficient to determine a park factor. For fielding data, I would suggest that even three years isn't enough, given the error inherent to the measurement.

And if weather can vary a lot from day to day, the only solution would be to create day-specific park factors - which is impossible. It might be true that some parks need day factors vs. night factors, though I'm guessing that someone has looked into that.

[HoustonGM]

Regarding weather, the relevant thing is that it can affect yearly park factors. If there happens to be more days during the year where the wind blows out, the park will be more hitter-friendly, but the next year could have more days with the wind blowing in, and thus play more pitcher-friendly. I think that weather is one of the more significant factors that contribute to park factors varying from year to year, besides just simple variation.

[200tang]

If one year of fielding data is not sufficient to judge the quality of a player, then it also isn't sufficient to determine a park factor. For fielding data, I would suggest that even three years isn't enough, given the error inherent to the measurement.

Well, they use three years of data, I was just saying that if boomboom doesn't think they change than it should be enough. The case is pointless though since they don't do yearly data.


And if weather can vary a lot from day to day, the only solution would be to create day-specific park factors - which is impossible. It might be true that some parks need day factors vs. night factors, though I'm guessing that someone has looked into that.

What HGM said. Wrigley moves from a hitters park one year to more pitchers next. Day by day would be impossible.

[HoustonGM]

What HGM said. Wrigley moves from a hitters park one year to more pitchers next.
It actually doesn't. Checking the ESPN park factors back to 2001, it was a pitcher's park in 2001 and 2002 but has been a strong hitter's park every year since.

I was just saying that such weather patterns do play into the yearly changes that occur in park factors.

[200tang]

It actually doesn't. Checking the ESPN park factors back to 2001, it was a pitcher's park in 2001 and 2002 but has been a strong hitter's park every year since.

I was just saying that such weather patterns do play into the yearly changes that occur in park factors.

Ok, I guess I should say, *can*.

[Dyzalot]

I used to play a tabletop baseball sim called Pursuit the Pennant. It had weather effects for each region of the country, by month coded into the game. Wrigley would be a pitcher's park in April, May and October and a hitter's park the rest of the time.

[dps]

Also, the physical factors of parks change a lot more than people realize. It's not just changes in the actual dimensions of the playing field that can change the park factors. A new playing surface may have a big impact, even if it's just replacing the old surface and not a change in the type of surface--even the groundskeepers cutting the grass a little bit short or not as short as the year before. Or a new scoreboard or pressbox could change the wind patterns. Or a slight change in the color of centerfield walls can change the hitting backdrop and cause a big difference. Or--well, you get the idea.

[SirKodiak]

One thing that could change park factors is how tall the grass is, and that can be changed basically whenever a team wants it to. Quality of the playing surface is another one. The batter's eye is another (Twin's new park is going to be having a changing batter's eye for a while). Mound quality. Rain plays a part, not just the wind. The number of day games would matter too, as teams tend to field lesser teams on day games.