Filter posts by...
Parks
Sunday, December 18, 2011
Too much is made of matchup stats (i.e., how batter X does against pitcher Y after 20-40 plate appearances, while ignoring each player’s career 2000 plate appearances).
At the same time, is too little made of individual player-park factors? If you have 3000 plate appearances of Barry Bonds at 3Com and away from 3Com (is that what it was called then? I can’t keep up with corporate names), and he hits as many HR at home as he does on the road, do we really care that the average LHH hits two-thirds as many HR at 3Com than away? I say: NO! Barry Bonds has very little in common with those hitters, other than handedness, but (traditional) park factors DEMAND that we treat Bonds as being tightly coupled with that group.
This is Mat Latos’s career Petco / non-Petco slash line (BA, OBP, SLG):
.229 .287 .348
.224 .286 .351
I won’t even bother to tell you which is Petco and which is not, since it’s the same thing! The question is: did he simply have made luck at Petco and good luck away from Petco? Or, is his talent level such that it’s an important parameter when looking at the park factors of the rest of the players? That is, with Bonds, he hits HR so far, that it doesn’t matter that 3Com suppresses HR.
Similarly, is Petco such that Latos can’t benefit from its pitcher-friendliness? I DON’T KNOW. But, these are the kinds of valid questions that are out there.
Latos’ strikeout rate per PA is 23.7% at home and 23.4% on the road. His walk rate is 6.9% at home, 7.9% on the road. His HR rate is the same 2.2% at home and on the road. His BABIP is .283 at home and .276 on the road.
Therefore, we need to do just like we do with handedness splits: personalize them. In The Book, we showed that you take the observed handedness splits and regress them toward the league mean (by adding 1000 PA for LHH and 2200 for RHH, to the number of their PA against LHP). We need to come up with the same thing for parks.
The problem is that you may have to come up with a different regression amount for each park AND for each quality of player. It makes it a tougher job.
So, we can’t presume that Latos was either:
a. completely unaffected by Petco (as evidenced by his splits)
b. extremely unlucky at Petco (as evidenced by all MLB players at Petco)
The truth is somewhere in the middle.
Tuesday, November 29, 2011
Can we really simply just do this, for Barmes?
Was Adrian Gonzalez disproportionately suited toward Fenway and against Petco? Clearly the opposing pitcher is going to try to mitigate whatever advantage a player has. If you just take Gonzalez extra base hits at Petco, and pro-rate to 351 PA, that gets you 27 extra base hits. If you look at all the other parks he’s played in, he pro-rates to 42 extra base hits in 351 PA. So, bring him to Fenway, and we’re thinking that he’ll do great right? He hit 31 extra base hits at Fenway.
Now, I’m not saying that Barmes at PNC will or won’t be affected. But there seems to be this presumption of overlaying performances, to see how someone will be affected. And, yes, I’m sure this is true… to some degree.
What a saberist’s job to figure out is the level of degree… how far can you make that claim.
Wednesday, October 19, 2011
Poz reels off some interesting info.
It probably helps to understand the home/away context. A team will score about 0.4 more runs per 27 outs at home than on the road. (This is why for example, they have a .540 win% at home: 0.4 runs = 0.04 wins, and 0.04 plus .500 is .540. Isn’t math fun?)
Another interesting thing to note is that a pitcher’s wOBA is around .160, while a DH is around .350 (more or less, and depending on the year). Pitchers come to bat 2.5 times, and then you have PH the rest of the way (PH are around .300 wOBA, and we’ll give them 1.5 PA). Remembering that to convert wOBA to runs you divide by 1.2, and we get: DH over P = (.350-.160)*2.5/1.2 = 0.39 runs. DH over PH = (.350-.300)*1.5/1.2 = .06 runs. So, the DH advantage is about 0.45 runs per game. Basically, the DH advantage is about as large as the home field advantage.
Ok, so he notes that:
On the road, with the pitcher hitting for himself, AL teams have averaged about 3.77 runs a game. With the DH, they average about a half run more per game. I think that fits in with what you would expect.
We have three effects: the home/away, DH/P effect, and the unfamiliarity of playing “out of position” on the road (AL team using a P, NL team using a DH).
So, being at home, AL would be +.4 more runs than on the road. Being at home, DH generates +.45 more runs than P/PH. Being at home, the familiarity of playing by your rules, you’d get some positive effect, say something like +.15 runs (just pulled it out of my a$$). The total of these two is +1.00 runs, but Poz is reporting only +.50 runs scoring at home for AL teams than on the road. There’s another 0.50 runs unaccounted for.
Anyway, we go on to the NL:
The National League, though, offers the shocker. At home, with pitchers hitting, they average 4.15 runs per game, which is pretty close to what American League teams score at home. But on the road, using the DH, National League teams have scored only 3.4 runs per game, meaning they score seven-tenths of a run LESS per game with the DH than they do with pitcher’s hitting.
So, at home, we’d expect them to score +.4 more runs. The DH impact would mean scoring -.45 runs at home (i.e., fewer runs at home than on the road). And then, some impact for familiarity at playing by home rules of +.15 runs. In total, we’d expect NL teams to score +.10 more runs at home than on the road. Poz is reporting that they score +.75 more runs at home than on the road. That’s an extra 0.65 more runs unaccounted for.
Is it possible therefore that there’s a huge familiarity factor of playing by your home league rules, way above the +.15 I was originally contemplating?
I mean just look at the data right here:
American League home record: 44-21 (.677 winning percentage)
National League home record: 37-29 (.561 winning percentage)
So, the average playoff team at home wins at close to a .620 clip, far higher than the .540 that you’d see in the regular season. And getting a +.08 win advantage, that’s a +.80 run advantage.
And that pretty much closes the gap in my numbers above.
That’s +.08 win advantage at home (over and above what you’d see in the regular season) is almost two standard deviations from the mean. It’s possible that this is just small sample size. But, seeing that we have a huge parameter here (playing by home-league rules), a .540 prior would seem to be too low to test against.
I’d love to see more data breakdowns and analysis on the AL/NL breakdown, including home-field advantage for intra-league playoff games as another comparison point.
Thanks Joe, learned something new today!
Friday, October 14, 2011
Tom,
I added the rest of the regular season data for 2011 and created the attached graphics to show the correlation between the sun’s position and hitting success. The big thing that jumped out at me was that hitting drops off when the sun is behind home plate and at an elevation of 20-29 degrees. According to the sketch-up model for Busch Stadium, shadows are cast between the mound and home plate when the sun is at a solar elevation of 21-29 degrees.
My biggest concern is that as you get to the extremities of the chart, fewer and fewer stadiums are contributing data points. I’m wondering if there are other park specific issues that could be impacting the results.
Let me know what you think.
Chris
Read More
Tuesday, September 20, 2011
How this was allowed to happen in the first place is incredible. But at least it won’t happen next year.
Glove-slap: Dan.
Sunday, September 18, 2011
All data from 1993-2010.
I look at each player’s game, and captured whether they were in the starting lineup or not, and which park they played in. And of course, their wOBA. I matched on those categories (starting or not, park), to make sure that the player pool was equally represented in day and night. (That is, if Sosa had 1000 PA at Wrigley in the day and 2000 PA at Wrigley at night, then I counted Sosa with 1000 PA in each.)
For hitters in the starting lineup, their wOBA was .3429 in day and .3423 at night. Basically, a match. However, I then broke it down by “times through the order”. And that’s where things became interesting.
Their 1st time through the order, it was 0.331 day, 0.335 night. That difference, on 219239 matching PA, is 3.9 standard deviations from the mean. The 3rd time (and later) through the order, it REVERSED, with 0.347 day, 0.344 night, or 4.0 standard deviations (going the other way). If you think of day time as games playing from 13:00 to 16:00 and night as playing from 19:00 to 22:00, then it would seem that the optimal point to maximize offense is somewhere between 16:00 to 19:00.
Basically, the above data would be consistent with this idea. If you peak at around 5 o’clock, then comparing your 1st time through the order performance (at 1 o’clock) in the day to 1st time through (at 7 o’clock) at night, we see an advantage for the night game. However, your performance the 3rd and later time in the day (at 3 or 4 o’clock) to the same at night (at 9 or 10 o’clock) would give the advantage to the day.
Interestingly, for substitute players, we don’t see that! The first time they come up, their day and night performances are a match (.305 in day and night). This part makes sense, because they’d be playing at say 2 or 3 o’clock in the day and 8 or 9 o’clock at night. Equidistant let’s say from 5 o’clock. But, the 2nd and later time they come up, their performance in the day is .313 while it .329 at night, which is quite the reversal. (Based on 4417 PA, which is 2.1 standard deviations from the mean, but in the reversed direction of the starting group.) Maybe these guys are subs for some particular reason, maybe they partied too hard, and didn’t recover for day games? I dunno.
For pitchers, their performances are more consistent with the 5 o’clock theory. Relievers, who would be pitching between 3 and 4 o’clock, perform with a wOBA of .328 in the day. But at night, say after 9 o’clock, the wOBA drops to .324. With 270287 PA, this is 3.7 standard deviations from the mean.
As a general rule, if you need to visualize it, your wOBA is going to peak at around 5 o’clock (maybe 6? someone else can figure it out), and the wOBA is going to drop by a point or two for each hour from that peak time.
This has some serious implication for the “times through the order” effect, and may provide partial explanation as for why starting pitchers perform as well as they do the 4th time through the order.
Saturday, August 13, 2011
Interesting charts here:
This is non-Colorado pitchers, and we see they’ve decided to throw fewer sinkers at Coors:
But this is Colorado pitchers, and they’ve decided to throw more sinkers at Coors (though it could be that it’s not the same pitchers in the same proportion in the two groups).
And here’s the piece de resistance, comparing how pitches move at Coors compared to the league overall. The “eye” is where a pitch would be observed to be thrown if you were playing catch. So, a MLB fastball looks like it “rises” in a regular park in comparison, but at Coors, it moves “straighter”.
Wednesday, August 10, 2011
ESPN reports:
Now, by themselves, the above splits aren’t conclusive, so to measure the effect of Rogers Centre more precisely, The Mag consulted with Wyers. He has developed a method that generates park factors by comparing a player’s performance in any given park with his performance in all other parks, not just in road games for that player. This reduces statistical noise and offers a better estimate of how a park actually plays in a given season. Wyers found that for every ball that batters made contact with in 2010, Rogers added .011 home runs, up from a rate of just .002 from 2005 to 2009. That puts Rogers Centre in 2010 among the top 3 percent of home run ballparks since 1950.
But only the Blue Jays, and not their opponents, got a home run boost in Toronto. When the Jays were on the road in 2010, they hit home runs in 4 percent of plate appearances in which they made contact, compared with an AL average of 3.6 percent. At Rogers, their home run on contact rate soared to 5.4 percent, which is a home-field advantage seven times the magnitude teams typically enjoy.
Opposing batters, however, actually homered on contact at a below-average rate in Toronto. As a result, the power differential between home and visiting hitters at Rogers in 2010 was the third largest of any park in any season over the past 60 years (see chart).
I’ve never looked, but let me take a really quick look here. And sign stealing would go beyond just batted balls. Walks, strikeouts, the whole thing comes into play. So, let’s see what we have:
In 2011, so far, Toronto batters’ OPS is 69 points higher at home. Their opponents are 47 points higher at Rogers. Since the typical home field advantage is about 30 or 40 points, seeing a 23 point advantage for Toronto batters over their opponents actually works against the theory, but in reality, it’s well within random chance.
How about 2010? Toronto batters had a 64 points advantage at home, while their opposing batters are 3 points under, giving a whopping 67 point differential (compared to the standard 30-40). Is that a big deal though? That’s about a 15 point wOBA advantage on 3000 PA. One standard deviation would be 9 points, so we’re talking under two SD. By itself, maybe there’s something to it. Maybe. But couple it with 2011, and there’s nothing there.
2009? The Toronto hitters were 8 points UNDER, while opposing hitters were 63 points UNDER. Quite a reversal of fortune for all concerned. Anyway, that’s a 57 point advantage for Toronto hitters. Putting the three years together, the 57, 67 and 23 points of advantage averages out to 52 OPS points, compared to the league average of 30 to 40 points. That’s about 7 wOBA points. Given 8000 PA, one SD is 5.7 wOBA points.
I don’t see it. And if you were to do all 30 teams, I’m sure you’ll find a couple of other teams with a bigger advantage at home than the Jays hitters.
If anything, the outlier is 2008, before all this apparently started. The Toronto hitters had a 28 point advantage at Rogers. But opposing hitters were 50 points UNDER, for a 78 point difference, far ahead of all subsequent seasons. In 2007, it was 37 point advantage for Toronto hitters, while opposing hitters were 47 points under, for an even greater 84 point differential. In 2006, Toronto hitters had a 100 point advantage, while opposing hitters were 48 points UNDER, for a super duper differential of 148 points! You want to talk about something weird, then go back to 2006. In 2005, it was 70 point advantage for Toronto hitters, while opposing hitters were 4 point UNDER, for a 74 point differential.
So, 2005-2008 is where Rogers was the huge advnantage for the hitters, averaging 96 points of OPS advantage (compared to presumably a 30-40 league average, though at this point, I’m too lazy to look that up). Anyway, that’s about a 25 wOBA difference on 12,000 PA. With one SD being 4.7 wOBA points, that’s 5 SD from the mean. That’s about as big an outlier as you’ll ever find.
If you want to investigate, go back to 2005-2008, and find out what happened at Rogers. 2009-2011? Nothing in comparison.
Thanks to B-R.com for the ease of data access.
Thursday, July 21, 2011
Good stuff.
This analysis confirms a significant inverse correlation (more AF, fewer SO; less AF, more SO) between AF and SO, for the 3 time periods reported. It reaffirms the previously-reported relationship identified in the AL 1964-68, as presented at SABR 40 . Analysis of the effect of each square foot of foul territory on SO is currently in progress. It is hypothesized that the Cleveland Indians pitching staff would have SO more hitters in 1964-68 had they not pitched 81 games yearly in Cleveland Municipal Stadium.
Basically, the true relationship is between foul area and foul-strikes. If you have no foul territory, then those foul balls are out of play foul balls (and the at bat is extended). If you have a huge foul territory, then those foul balls are in-play foul balls. And if they are in-play, then they have a chance of ending the at bat “prematurely” (disproportionately early).
Strikeouts is what you use to infer this relationship.
Monday, June 27, 2011
On ESPN.
Thursday, June 02, 2011
Another honest mess. New park parameter should be used, but has not been handled.
There were 20 teams that played in 1968 that also played in 1969 (i.e., in 1969, it was the 20 teams of 1968, plus the 4 expansion teams). If we ONLY look at the matching teams (the two that played in both), we see that their per-game attendance increased from 14,220 to 14,760, for an increase of 4% in attendance. However, their won-loss record went from .500 to .527 (the wonders of expansion), and that kind of jump would demand an 8% increase in attendance. We can say therefore that attendance was DOWN 4% (4% minus 8%) relative to expectations of win percentage.
I did this for all seasons. The biggest year over year drop was 1994 to 1995: a drop of 20% in attendance per game.
Anyway, I also “chained” the results. What this provides is a timeline adjustment.
For example, in 1948, the timeline adjustment is 0.98. That’s the same timeline adjustment as 1997-98. Therefore, we can say that even though there were about 17,000 per game in 1948, that’s equivalent to the 29,000 per game in 1997-98.
We see that MLB was most popular in 1993-94, though 2007-08 comes close.
Again, I’m providing an honest framework, where the mess of not considering new ballparks should be handled (by someone else).
Generally speaking, the eras can be set as:
1946-1951 (88% attendance of 2009)
1952-1960 (66%)
1961-1976 (58%)
1977-1986 (77%)
1987-2009 (99%)
Data (quasi-formatted… copy to Excel to see it better formatted):
Read More
Wednesday, June 01, 2011
There have been several great articles that track attendance by city over time, and controls for such factors as expansion, new ballpark, past won-loss record. The article I am writing below is not going to be one. That’s because I don’t control for any of those three key parameters. There was an excellent article written a few years back in the Phil-edited By The Numbers. If you want something good, read that one.
What I’m going to do here will establish the framework, and then some aspiring saberist can improve upon it.
Let me tell you what I did, and why I did it. The why will always be the same answer: because it was easy. I choose the honest mess because I don’t want to spend more than a few minutes doing what I’m doing, and by leaving it honest, you can clean up the mess if you want to. Or not. Indeed, I will spend more time writing this article than doing the actual research.
There are three eras for baseball, in terms of attendance. There is 1946-1976, where the per-game attendance was around 14,000 fans, give or take a few thousand each year. There is 1987-present, where the per-game attendance was around 28,000 fans, give or take a few thousand each year. And there was the transition period of 1977-1986, as baseball’s fandom rose from one plateau to the next. Why this happened, I will leave to the historians. I will note that I became a baseball fan right around 1977, and a regression equation will therefore conclude that I am completely responsible. Which is why I hate regression equations.
Free agency and money was probably the real reason. People like to see valuable things, and if you pay alot for someone, then people will want to see that player. It’s like McNall said when he and Gretzky bought the Honus Wagner card: you buy the most important card, and you pay top dollar, and that drives its value. This is true even for small things, like books. When we had to price our book, we thought about pricing it for 9.99$, but there was a theory that if you priced your book too low, it couldn’t be valuable! Indeed, the lower we wanted to price the book, the less units it would sell apparently. A comedian had a joke about his father buying a VCR (old joke), and it was priced at 99$, and the father told the salesman he wanted to pay 49$ for it. The salesman was so flustered with the back-and-forth, he told the father to take it for 19$. And the father said “19$? What’s wrong with it?”.
Anyway, historians and comedians and economists have a better handle on this than I do. Listen to them, and ignore me on why circa 1977 was pivotal.
Ok, what I did:
Read More
Thursday, May 26, 2011
I enjoy the site and the wide ranging topics that are discussed in the Blog. I ran a search for an idea that interests me about MLB stadiums and their construction. I came across 106 articles that mention ‘stadiums’, but do not address my query. My question is this, considering park factors with a 3 year data regression correlates if a stadium is a pitchers or hitters park, is it possible given a site, a team can have an engineering team working in conjunction with the construction contractor and quite possibly a physicist to determine the dimensions within regulation required to construct a park with a neutral, hitter positive or pitcher positive park or would they need to build within the allotted area taking in consideration all of the non game related factors, open the park, check the data and readjust later to achieve the desired result, preferably a neutral park? It would seem to be an advantage to have a stadium with a neutral park factor and it does not hinder you r roster construction. I apologize for the rambling nature of this question.
How well can we call a new park? I know that MGL will come up with an initial forecast for a new park, so maybe he can shed some light from his experience.
Has anyone tracked this? It does seem that the mew Yankee Stadium and Mets park took people by surprise. Did anyone guess Coors?
Anyway, what conditions make it the hardest to establish the neutrality of a park? Being by the lake? Being in a windy city?
Thursday, May 05, 2011
Sean has expanded his Play-Index to use the weather parameters as noted in Retrosheet.
http://www.bbref.com/about/coverage.shtml
Weather Data
The weather data is based on conditions at the start of the game. Below we show the percentage of each data set (temp, wind speed & dir, etc) which are not null (or unknown). This data is included in the RetroSheet data files and is provided as is and most certainly contains some errors. There is no weather data pre-1950.
For example, these are HR hits in 2010-2011 where the temperature (using your father’s scale) of at least 95 degrees (35 in the worldwide scale).
Wednesday, May 04, 2011
Study:
Earned runs allowed by home pitchers were lowest on clear days at 3.93, climbing to 4.26 on cloudy days. For visiting pitchers the ERA was 4.50 in the clear and 4.68 under the clouds.
...
The analysis is based on statistics from 10,758 major league day games obtained from STATS LLC and weather data collected by the National Climatic Data Center, showing the conditions at the nearest National Weather Service office to each stadium at game time. The findings are published in the current issue of the journal Weather, Climate and Society.
Kent said he had expected to see better hitting in cloudy conditions but was surprised by how strong the effect was on strikeouts. Home pitchers averaged 6.65 strikeouts on clear days, but in cloudy conditions that fell to 6.22. For visiting pitchers, the drop from clear to cloudy was from 6.14 to 5.67.
...
On clear days, home teams won 56 percent of their games...When it was cloudy, that fell to 52 percent home wins…
...
Home teams had 0.98 home runs per game for clear days and 0.96 when it was cloudy. For visitors, the change from clear to cloudy was from 0.95 to 1.01. Home pitchers gave up 3.37 walks on clear days and 3.43 when it was cloudy. Visiting hurlers averaged 3.56 walks for clear days and 3.50 under clouds.
Anyone who’s seen an outfielder lose a ball in the sun won’t be surprised to hear there are more errors on clear days than cloudy ones.
The difference is largest for visiting teams who would not be accustomed to the glare and light angles in someone else’s stadium. Visiting teams averaged 0.80 errors on clear days and 0.73 on cloudy days. For home teams the decline was from 0.77 on clear days to 0.75 as it got cloudier.
If someone finds the paper, please post below.
Tuesday, March 22, 2011
Kevin’s been providing this service for several years via downloadable files, which has been great. Now, he’s put it up online as well.
Thursday, March 17, 2011
Wonderful stuff:
Digging a little further revealed a turning point for this NL homer rate in ’67:
April 2.56%
May 2.64%
June 2.99%
July 2.60%
August 1.82%
Sept/Oct 1.89%
Wow. That looks to me like something happened to dramatically reduce homers either at the beginning of August that season or somewhere in the back end of July.
In 1968, that low rate resumed:
1st half: 1.94%
2nd half: 1.95%
April 2.37%
May 1.69%
June 2.04%
July 1.72%
August 2.03%
Sept/Oct 2.00%
And the possible cause:
“Rawlings had ... six manufacturing plants--four in Missouri and two in Puerto Rico--when it was sold in 1967 to Automatic Sprinkler Corp. of America. This conglomerate made the company a division under its prior Rawlings Sporting Goods name.”
It seems a remarkable coincidence that a change in home run rates should come in the same year that the ball manufacturer changes ownership.
Tuesday, March 08, 2011
Looks interesting. I’d like to see this split between home and road, and v LHH and RHH. I just don’t necessarily trust anyone park adjustments.
Glove-slap: Lee
Friday, February 25, 2011
Great stuff from Max.
Saturday, January 29, 2011
J-Doug, among several others, want to talk to you. Phil too.
I love it that we have this sabermetric army these days ready, willing and able to knock everyone back to reality.
from Amazon
Recent comments
Older comments
Page 1 of 320 pages 1 2 3 > Last »Complete Archive – By Category
Complete Archive – By Date