PCA Request Line

[Matthew C.]

I am saying that Koufax "over-excelled" mightily at Dodger stadium. I know he got penalized by WAR for pitching half his games at Dodger stadium.

Here is a link to how much he over-excelled.

http://www.beyondtheboxscore.com/2006/5/5/12349/01432

The article basically states that adjusting for park factors does not always tell the whole story. Some pitchers benefit disproportionately from their parks, and that was exactly what Koufax did.

As Brett stated, Koufax had a road era of 2.70 his lat 5 years. I just don't see how that would have translated to seasons of 10+ WAR had he pitched for a different team, at a home park other than Dodger stadium.

Koufax would have completed a lot fewer games. From 62-66, he completed 66% of his games at home, and only 48% on the road. He averaged around 8.4 innings per start at home, but only about 7.2 innings per start on the road. The lower innings would have made a dent in his WAR total.

His road numbers, a 2.70 era, and 7.2 innings per start just don't add up to a 10+ WAR season because that would be a good baseline to start at should Koufax never have had the luxury of Dodger stadium. That is why I just don't see him putting up a 10+ WAR with his new team had he been traded after his 1966 season.

How do you know that the reason that he didn't "overperform" at Dodger Stadium (compared to what was expected) is because he had a skill unique to that park or took advantage of the park?

If Koufax's splits were more extreme than his mates, then one of two things - 1. he had something in his repertoire that helped him excel in that park more than his mates or 2. he was lucky.Lets break down what it was about his pitching style, etc. that helped him "over-excel." What did he do that caused his splits to be more dramatic than his teammates? Until we can answer that question, I don't think we can assume anything.

[curveball]

What makes you think this?

I think it is more likely that good pitchers 1. have the ability to excell regardless of their circumstances more than other pitchers and 2. have the skill to take advantage of park effects, those similar to what brett was refering too.

I think Matt agrees that top players are hurt less by LQ and park issues than margainal/ poor players. But he can confirm that.

I agree with you. When I said that "I don't have any evidence, but I think that good pitchers are helped more by pitcher friendly parks than average to poor ones", I am saying good pitchers can take more advantage of a pitcher friendly park than average to poor ones, which is why it helps those good pitchers more.

[curveball]

How do you know that the reason that he didn't "overperform" at Dodger Stadium (compared to what was expected) is because he had a skill unique to that park or took advantage of the park?

If Koufax's splits were more extreme than his mates, then one of two things - 1. he had something in his repertoire that helped him excel in that park more than his mates or 2. he was lucky.Lets break down what it eas about his pitching style, etc. that helped him "over-excel." What did he do that caused his splits to be more dramatic than his teammates? Until we can answer that question, I don't think we can assume anything.

That is the million dollar question. Was it all Sandy' skill, or was it because Dodger stadium was even more friendly to left handed pitchers, or was it something else? I don't think any of us know for sure. But what we do know for sure is that he did over perform at his home park. That over performance led to even more WAR value.

If Koufax gets traded to another team, do you think he would also over perform to the same extent in his new stadium as he did at his old one? I don't think so, and that is why I don't take his peak at face value. I believe there was something about Dodger stadium that he would never be able to replicate anywhere else.

[Matthew C.]

That is the million dollar question. Was it all Sandy' skill, or was it because Dodger stadium was even more friendly to left handed pitchers, or was it something else? I don't think any of us know for sure. But what we do know for sure is that he did over perform at his home park. That over performance led to even more WAR value.

If Koufax gets traded to another team, do you think he would also over perform to the same extent in his new stadium as he did at his old one? I don't think so, and that is why I don't take his peak at face value. I believe there was something about Dodger stadium that he would never be able to replicate anywhere else.

Surely someone would have some insite regarding Dodgers Stadium and lefties...that shouldn't be too hard to rule on.

[BigRon]

But the parks are nuetralized in WAR - that is my point. If you think Koufax couldn't have produced as high of a peak not in Dodger Stadium...WAR agreed...which is why it adjusted his WAR to 10.8. IF you do not think 10.8 was possible even after adjustment that you do disagree with his WAR total.

Been following this a bit. It seems to me that there is some misunderstanding by at least one poster about the application of War or Win Shares or OPS+ or ERA+ in trying to iron out park effects. They attempt to do just that. The results try to emulate basically neutral park performance. Are they perfect? No. Are they basically "right"? I think so.

[Matthew C.]

Been following this a bit. It seems to me that there is some misunderstanding by at least one poster about the application of War or Win Shares or OPS+ or ERA+ in trying to iron out park effects. They attempt to do just that. The results try to emulate basically neutral park performance. Are they perfect? No. Are they basically "right"? I think so.

We are on the same page now - he is sayong that Koufax outperformed even what a normal Home/Away split Dodger Stadium split would suggest.

We now have to decide if the park favored Koufax, if Koufax had a skillfull hand in the difference himself, or if it was plain luck. Given the sample size "luck" seems not likely.

So was Dodger stadium constructed in such a way to allow Koufax particularly to outperform his expected splits or did Koufax have some skill/attribute that allowed him to outperform his expected splits?

Not sure there is a lot of difference between the two options.

[BigRon]

So was Dodger stadium constructed in such a way to allow Koufax particularly to outperform his expected splits or did Koufax have some skill/attribute that allowed him to outperform his expected splits?

Not sure there is a lot of difference between the two options.

Well, Dodger Stadium is symmetrical, so there's no obvious left-right thing there. I never heard anything about unusual wind currents there- not saying it doesn't exist, but I never heard anything. No unusual temperature issues that I'm aware of- usually pretty benign there.

I would think that pitchers with very high K rates relative to average are somewhat less influenced by park issues than others. True, Koufax struck out batters at a lower rate on the road, but still very high against league norms.

Interesting that Koufax completed "only" 48% of his road starts during the referenced time. Still WAY higher than average.

[curveball]

Well, Dodger Stadium is symmetrical, so there's no obvious left-right thing there. I never heard anything about unusual wind currents there- not saying it doesn't exist, but I never heard anything. No unusual temperature issues that I'm aware of- usually pretty benign there.

I would think that pitchers with very high K rates relative to average are somewhat less influenced by park issues than others. True, Koufax struck out batters at a lower rate on the road, but still very high against league norms.

Interesting that Koufax completed "only" 48% of his road starts during the referenced time. Still WAY higher than average.

From 1962-1966, Juan Marichal completed 58.51% of his games at home, and 59.09% of his road games. These weren't exactly Bob Gibson's best years, but he completed 41.46% of his home starts, and 59.09% of his road starts.

[Matthew C.]

I would think that pitchers with very high K rates relative to average are somewhat less influenced by park issues than others. .

This is correct and one notch in Koufax's favor, perhaps.

[brett]

I am saying that Koufax "over-excelled" mightily at Dodger stadium. I know he got penalized by WAR for pitching half his games at Dodger stadium.

Here is a link to how much he over-excelled.

http://www.beyondtheboxscore.com/2006/5/5/12349/01432

The article basically states that adjusting for park factors does not always tell the whole story. Some pitchers benefit disproportionately from their parks, and that was exactly what Koufax did.

As Brett stated, Koufax had a road era of 2.70 his lat 5 years. I just don't see how that would have translated to seasons of 10+ WAR had he pitched for a different team, at a home park other than Dodger stadium.

Koufax would have completed a lot fewer games. From 62-66, he completed 66% of his games at home, and only 48% on the road. He averaged around 8.4 innings per start at home, but only about 7.2 innings per start on the road. The lower innings would have made a dent in his WAR total.

His road numbers, a 2.70 era, and 7.2 innings per start just don't add up to a 10+ WAR season because that would be a good baseline to start at should Koufax never have had the luxury of Dodger stadium. That is why I just don't see him putting up a 10+ WAR with his new team had he been traded after his 1966 season.

So I decided to play around with the park factors and other expectations as well. For someone with a 2.70 road ERA we would expect about a 2.45 to 2.50 home ERA due to normal home/road splits given a neutral home ballpark. Koufax' "park factor" for those 5 years was about 91-92, but that is the total effect on runs caused by simply playing half time in your home park. It means that Dodgers' stadium actually gave up only about 82-84% as much as average, (plus his road parks would have been 101-102%, so we might expect that home ERA of 2.45-2.50 to drop down into the range of about 1.95 to 2.10. In other words a 2.00 home ERA would have been right in line with a 2.70 road ERA during that period, and in at least one of the seasons, his home ERA is higher that projected from his road ERA. Remember his home ERA was about 1.40 during that stretch.

If that 1.40 is significantly different than the expected 2.00, I would postulate that it may be simply because there is non-linearity in run production corresponding to very low allowed OB%, and slugging percentages. In other words, take a pitcher with a 150 ERA+ in a 4.00 run environment, and put him into a 3.00 run environment and he now becomes a 180 ERA+ pitcher because offense drops off even faster at low rates allowed in a very low run setting.

(On second thought though, Koufax' OPS+ against was not out of line with his ERA+'s during that period.)

Another factor I remember was that Koufax was one of the first pitchers to get highly righty platooned against. Teams would put up right handed hitters who were not in the unsual lineup against him, and I think I remember that they did this MORE in LA than in their home parks, perhaps because starters were more reluctant to sit on the bench in a home game.

[cbenson5]

Hey Matt,

I am curious how you handle League Quality in PCA. More broadly, can you give any advice on how one should handle LQ when comparing recent stars to those from the early part of the 20'th century? I am asking, because I am beginning to think that I have been overrating the early stars from baseball. For example, I broke down my personal top 20 by year of debut into three segments pre 1933, 1933-1965, 1966-1999. This is what I came up with:
Pre- 1933
8 players (Wagner, Cobb, Ruth, Speaker, Hornsby, Collins, Gehrig, Ott)
1933-1965
8 players (Williams, Musial, Mantle, Morgan, F. Robinson, Mathews, Mays, Aaron)
1966-1999
4 players (Bonds, Henderson, Schmidt, Brett)

With the exception of George Brett and, maybe, Mathews, I don't think I have an unusual top 20. However, logic tells me that more top players should come from recent years. I am just hoping for some advice from an expert. Thank you for your time.

[SABR Matt]

Hey Matt,

I am curious how you handle League Quality in PCA. More broadly, can you give any advice on how one should handle LQ when comparing recent stars to those from the early part of the 20'th century? I am asking, because I am beginning to think that I have been overrating the early stars from baseball. For example, I broke down my personal top 20 by year of debut into three segments pre 1933, 1933-1965, 1966-1999. This is what I came up with:
Pre- 1933
8 players (Wagner, Cobb, Ruth, Speaker, Hornsby, Collins, Gehrig, Ott)
1933-1965
8 players (Williams, Musial, Mantle, Morgan, F. Robinson, Mathews, Mays, Aaron)
1966-1999
4 players (Bonds, Henderson, Schmidt, Brett)

With the exception of George Brett and, maybe, Mathews, I don't think I have an unusual top 20. However, logic tells me that more top players should come from recent years. I am just hoping for some advice from an expert. Thank you for your time.

I wish I'd had more time to pursue a deeper league quality analysis. I used the skewness and kurtosis of each league's RS/G/Side distribution to make an estimate of league quality (the more skewed the distribution, while at the same time possessing low kurtosis (a measure of the sharpness of a curve)...the more tail-heavy the run scoring was, thus implying that there were teams beating up on other teams...implying a deep imbalance in league parity and, likely, poor league quality. That was just a rough guess.

PCA itself does not use that as I'm not confident in its' assessment...what it does instead is force the player win creation rate distribution in each league to fit a standard distribution as represented by the all-time distiribution at least position for defense, and for each pitcher and batter...I call that normalization...it's not perfect, but it at least creates a final ranking that is a little less ancient-heavy than the original PCA.

Mty top 20 is somewhat different than yours, though obviously many of the same names are present...I'll post it when I have access to my data and can incorporate all moder data not included i the old database (as well as make tweaks to cccount for what I believe are weaknesses in the original PCA method).

[brett]

Matt, it was recently brought to my attention that George Brett had apparently significantly better home than road hitting stats in his early years, '74-'82 with the exception of the 1980 season. Then starting in '83 through the end of his career he had almost the exact opposite trend, almost completely balancing out his hitting value by the end of his career. Is there any explanation for this? I remember you wrote once that Royals stadium benefitted line drive hitters, and perhaps those who ran well. I noticed too that Brett had over 2/3 of his triples at home. Does it seem plausible that Brett did that much better at home from '74-'82 when he ran much better and tried to hit line drives, and then when he lost some speed and maybe tried to hit more flyballs that he was actually hurt (relatively) by his park trends?
Is there any easy way to get his combined home-road splits for '74-'82 excluding 1980 and for '83-'90. I think he batted 70 points higher at home from '74-'82.

[SABR Matt]

So the question would be...why exclude 1980? What was different that year?

I am usually inclined to believe that even several-year trends in a split stat are dubious unless a solid reason is provided that they might exist...but it is possible that when Brett was faster and more of a line drive hitter, he had more of an advantage at home. Yes...their home park did strongly benefit speedy players (hit it on the ground hard and the artificial turf was like superball material)...are Brett's home splits strongly favoring triples and singles at home during that stretch? And then strongly favoring home runs on the road later in his career?

[brett]

So the question would be...why exclude 1980? What was different that year?

I am usually inclined to believe that even several-year trends in a split stat are dubious unless a solid reason is provided that they might exist...but it is possible that when Brett was faster and more of a line drive hitter, he had more of an advantage at home. Yes...their home park did strongly benefit speedy players (hit it on the ground hard and the artificial turf was like superball material)...are Brett's home splits strongly favoring triples and singles at home during that stretch? And then strongly favoring home runs on the road later in his career?

I am going to start to piece together the splits. Actually, '80 was balanced, '81 was nearly balanced, and '82 started a trend in which he performed better on the road consistently every year. In '80 he started to save his legs, (walked more, hit more home runs, became willing to hit the ball in the air more which he really tried to avoid through '79).

So let's start with his splits from '74-'79.

At home he went:

603 for 1628 for a .370 batting average!
132 doubles
48 triples
32 home runs
.569 Slugging %

On the road he went:
474 for 1827 for a .259 batting average
77 doubles
25 triples
42 home runs
.398 slugging percentage!

He tripled in 1.4% of his at bats on the road and 2.9% at home.

Those look like some huge splits over a span of 6 years.
The reason that Brett was so interesting in his splits was that his road rates might not have given him a single all star type season going into 1980, while in reality he was an established star with a batting title, and runner-up MVP in '76, and had started 3 straight all star games, and was also runner up MVP in '79. His '77 and '79 road slugging percentages were solid at .495 abd .492.

Then in '80 he hits .388 and sluggs .640 on the road, .391 and .685 at home. He only hit .300 on the road once at .301 with a .398 slugging in '76,

From '82 to '93 his numbers look a lot more balanced, but he seemed to do better on the road, despite most hitters not doing quite as well at home.

At home he went:
889 for 3005 and a .296 average
188 doubles
36 triples
89 home runs
.471 slugging

On the road
899 for 3053 and a .294 average
206 doubles
just 12 triples
.496 slugging percentage.
124 home runs (That's about a 420 home run pace on the road prorated through his career at bats and even better with a normal home boost)


his .294 average and .496 slugging on the road over that 11 year period are definitely hall of fame level production for that era. His overall productivity almost balances out over the length of his career, but it seems to show me that a players approach DOES impact their performance in a given ballpark. It seems to support Brett's own statements that he was taught to hit the ball down in his early years, to keep it on the ground or hit line drives, but that fly balls were bad, and then again that he became willing to try to pull the ball more and hit it in the air more.

Overall Brett still went .290/.356/.469 on the road with over 1500 hits and 181 home runs in a league that as a whole went about .265/.323/.380 composite on the road over his career. .320/.383/.506 at home with 136 home runs.

[brett]

Duplicate post

[cbenson5]

I wish I'd had more time to pursue a deeper league quality analysis. I used the skewness and kurtosis of each league's RS/G/Side distribution to make an estimate of league quality (the more skewed the distribution, while at the same time possessing low kurtosis (a measure of the sharpness of a curve)...the more tail-heavy the run scoring was, thus implying that there were teams beating up on other teams...implying a deep imbalance in league parity and, likely, poor league quality. That was just a rough guess.

PCA itself does not use that as I'm not confident in its' assessment...what it does instead is force the player win creation rate distribution in each league to fit a standard distribution as represented by the all-time distiribution at least position for defense, and for each pitcher and batter...I call that normalization...it's not perfect, but it at least creates a final ranking that is a little less ancient-heavy than the original PCA.

Mty top 20 is somewhat different than yours, though obviously many of the same names are present...I'll post it when I have access to my data and can incorporate all moder data not included i the old database (as well as make tweaks to cccount for what I believe are weaknesses in the original PCA method).

Thank you Matt. I imagine your analysis would go a little over my head, but it sounds like we are going through the same issue. That is trying to find a way to get more modern players in the higher echelons of our ratings. I look forward to seeing your top 20. Hopefully, it is not too far off from mine. Then I will feel a little better that I am on the right track.

[Sultan_1895-1948]

Matt, whats your new system called and do you have a thread for it?

[Honus Wagner Rules]

I hope Matt returns. He hasn't posted since August 13, 2011.

[Sultan_1895-1948]

Oh really? Didn't know that. Maybe he's hoarding info, gathering his arsenal, and will come back in with a crashing wallop that will knock us off our feet :P

[Matthew C.]

Oh really? Didn't know that. Maybe he's hoarding info, gathering his arsenal, and will come back in with a crashing wallop that will knock us off our feet :P

Actually, he just got tired of what he perceived as snarky people always snarkly whining about how snarky sabermetricians were.

[Sultan_1895-1948]

Actually, he just got tired of what he perceived as snarky people always snarkly whining about how snarky sabermetricians were.

Always thought Matt balanced his interaction with pure stat crowd and the pure traditionalist crowd pretty well.

It isn't smart to only use one or the other; we should allow them to compliment each other.

Anyway, yeah hope he comes back.

[Matthew C.]

Always thought Matt balanced his interaction with pure stat crowd and the pure traditionalist crowd pretty well.

It isn't smart to only use one or the other; we should allow them to compliment each other.

Anyway, yeah hope he comes back.

There were a handful of members that constantly gave him fits - and he got tired of it. I conversed with Matt a lot of times about stats off-site. I miss him around here too.

[Sultan_1895-1948]

There were a handful of members that constantly gave him fits - and he got tired of it. I conversed with Matt a lot of times about stats off-site. I miss him around here too.

Yeah I also took a long break, just choosing not to deal with the nonsense and ignorance. Just gets old. Even now, I read some things that just ooze with stupidity. Matt's passion was probably attacked on a personal level one too many times. Well, if you have a way to tell him, let him know his work is appreciated.

[Honus Wagner Rules]

I want Matt to come back so he can explain to me the inner workings of PCA.