Starting Pitcher stats

[CentralCABucsFan]

A year or two ago I went through the entire national league and determined just what makes a #1 starter, a #2, a #3, a #4, and a #5 starter.  I think I basically went through the actual stats for each team, and grouped the top 16 starters and called them #1's, the second 16 and called them #2's, etc.

My point at the time was that when we start looking at #4 and #5 starters, the stats aren't so hot across the league.  When I hear statements like "we have no pitcher better than a #5 starter", I disagree. 

I was curious if there is a collection of stats out there for the starting pitchers that could help make the distinction and give us an average stat for each level of starter.  I used ERA and Whip a few years ago, but I believe there are better overall stats now.

I'd be curious where Duke, Maholm, Ohlendorf, etc, really fall and would like to use it as a way of evaluating free agent starters this offseason.  Not by guessing, but by using actual stats.  Any thoughts on a starting point for something like this?

[Dale Berras Stash]

FIP (Fielding Independent Pitching) and xFIP are the hot stats for ranking stats over at Fangraphs.

[MattB]

Lookout Landing broke down rotation spots by TRA+ in 2008 (http://www.lookoutlanding.com/2008/9/23/618821/rotation-slots-in-reality).  I assume those numbers are still relevant.  TRA+ is available at http://statcorner.com.

1 -- 2 BARRIER: 118 tRA+
2 -- 3 BARRIER: 106 tRA+
3 -- 4 BARRIER: 95 tRA+
4 -- 5 BARRIER: 86 tRA+

[epoc]

Lookout Landing broke down rotation spots by TRA+ in 2008 (http://www.lookoutlanding.com/2008/9/23/618821/rotation-slots-in-reality).  I assume those numbers are still relevant.  TRA+ is available at http://statcorner.com.

1 -- 2 BARRIER: 118 tRA+
2 -- 3 BARRIER: 106 tRA+
3 -- 4 BARRIER: 95 tRA+
4 -- 5 BARRIER: 86 tRA+

One thing that is very important to note about these figures is that to be useful you need to account for a presumed number of innings pitched. For example, the way that the author of this article figured out these tRA+ figures is by ranking all starters by quality, and then using inning thresholds to determine the cutoff. For #1 starters, he presumed 200 IP, so the top 6000 IP of performance was used to determine the quality of a #1 starter. But since he (and anyone else who does a study like this, the seminal one being Sackmann's, which the author of this article references) presumes innings totals, you have to list the innings totals as part of what makes a #1 starter. If a starter can maintain a 106 tRA+ but you can't count on him for 192 IP, he is not truly a #2 starter.

So to flesh out the data that Matt provided, the cutoffs for each rotation spot are:
#1 = 118 tRA+ 200 IP
#2 = 106 tRA+ 192 IP
#3 = 95 tRA+ 183 IP
#4 = 86 tRA+ 175 IP
#5 = anything worse

When you look at it that way, it's not so easy to be a #3 starter! What ends up happening is that teams try to find guys who are much better than average, knowing that they're probably not going to go a whole season without injuries to their pitchers. But by the same token, this is a big part of the reason that guys like Duke and Maholm are valuable - especially Maholm. You can count on them for 180+ IP, which even with mediocre performances makes them very valuable. Over the past four years, Maholm has thrown 739.1 innings with tRA+s of 99, 107, 107, and 82. If you guess that he'll throw 25 more innings or so this year, that's an average of about 188 IP and a tRA of 95-100. That makes him a #3 starter easily. Duke is neither as good nor as reliable; he's pitched 642.2 innings with tRA+s of 77, 98, 104, and 74. If you figure 20 more IP for him this year, that's about 165 IP and a 85-95 tRA+. That's around the border of a 4/5 starter - good enough quality to be a #4, but not quite the innings.

Compare that to someone like Jorge de la Rosa, whose name has been tossed out there by a few people as someone who would be an upgrade for the rotation. He's thrown 533.1 innings with tRA+s of 82, 114, 117, and 106. Obviously, that's much better quality than Duke or Maholm, but far less durability. Figuring another 25 IP this season, and it's about 140 IP with a tRA+ of 105-115. If you wanted to be generous, you could bump the innings total up to 150 to account for the fact that he pitched out of the 'pen a bit in 2007-08. But anyway, that's #2 starter quality with #5 starter durability. One way to compare that to what Maholm does is to add in another 40 IP of replacement level starting pitching to even things out, since when de la Rosa can't start, his team has to use it's Karstenses and McCutchenses in his place. Replacement level tRA+ is about 75, so with de la Rosa plus a replacement you'd get about 190 IP with a 95-100 tRA+, or basically exactly what you'd get from Maholm.

Um, so I guess I just wanted to point out that it's very important to factor durability into your assessments of starting pitchers, and then I had fun playing around with the numbers. Hope some of you found this interesting.

[epoc]

Oh, by the way, the much, much easier way to judge this is by looking at WAR projections, since WAR factors in both quality and quantity. So, if you look at the updated CHONE projections - http://baseballprojection.com/2010/august2010p.htm - you can see that de la Rosa projects as a 2.5 WAR pitcher - a neutral ERA of 4.05 but only 139 IP. Maholm projects to have a worse neutral ERA (4.31) but still be a roughly equivalent pitcher (2.8 WAR) because of greater durability/dependability (186 IP). Just by doing a quick scan-test, I'd guess the chart would be something like:

#1 starter = 4+ WAR
#2 = 3-4 WAR
#3 = 2.5-3 WAR
#4 = 2-2.5 WAR
#5 = 1.5-2 WAR

That's just a quick estimate, though. Maybe tomorrow I'll sort through and figure it out for real. By that chart, the Pirates have one #3 (Maholm), one 4/5 (Duke), and a bunch of garbage, though CHONE is still projecting McDonald as a reliever instead of a starter.

edit: Those projections for Maholm and de la Rosa, by the way, are not projections of what they're final stat-lines for the season will be. They are projections of true talent for a full ML season.

[GoBucs21]

Lookout Landing broke down rotation spots by TRA+ in 2008 (http://www.lookoutlanding.com/2008/9/23/618821/rotation-slots-in-reality).  I assume those numbers are still relevant.  TRA+ is available at http://statcorner.com.

1 -- 2 BARRIER: 118 tRA+
2 -- 3 BARRIER: 106 tRA+
3 -- 4 BARRIER: 95 tRA+
4 -- 5 BARRIER: 86 tRA+

One thing that is very important to note about these figures is that to be useful you need to account for a presumed number of innings pitched. For example, the way that the author of this article figured out these tRA+ figures is by ranking all starters by quality, and then using inning thresholds to determine the cutoff. For #1 starters, he presumed 200 IP, so the top 6000 IP of performance was used to determine the quality of a #1 starter. But since he (and anyone else who does a study like this, the seminal one being Sackmann's, which the author of this article references) presumes innings totals, you have to list the innings totals as part of what makes a #1 starter. If a starter can maintain a 106 tRA+ but you can't count on him for 192 IP, he is not truly a #2 starter.

So to flesh out the data that Matt provided, the cutoffs for each rotation spot are:
#1 = 118 tRA+ 200 IP
#2 = 106 tRA+ 192 IP
#3 = 95 tRA+ 183 IP
#4 = 86 tRA+ 175 IP
#5 = anything worse

When you look at it that way, it's not so easy to be a #3 starter! What ends up happening is that teams try to find guys who are much better than average, knowing that they're probably not going to go a whole season without injuries to their pitchers. But by the same token, this is a big part of the reason that guys like Duke and Maholm are valuable - especially Maholm. You can count on them for 180+ IP, which even with mediocre performances makes them very valuable. Over the past four years, Maholm has thrown 739.1 innings with tRA+s of 99, 107, 107, and 82. If you guess that he'll throw 25 more innings or so this year, that's an average of about 188 IP and a tRA of 95-100. That makes him a #3 starter easily. Duke is neither as good nor as reliable; he's pitched 642.2 innings with tRA+s of 77, 98, 104, and 74. If you figure 20 more IP for him this year, that's about 165 IP and a 85-95 tRA+. That's around the border of a 4/5 starter - good enough quality to be a #4, but not quite the innings.

Compare that to someone like Jorge de la Rosa, whose name has been tossed out there by a few people as someone who would be an upgrade for the rotation. He's thrown 533.1 innings with tRA+s of 82, 114, 117, and 106. Obviously, that's much better quality than Duke or Maholm, but far less durability. Figuring another 25 IP this season, and it's about 140 IP with a tRA+ of 105-115. If you wanted to be generous, you could bump the innings total up to 150 to account for the fact that he pitched out of the 'pen a bit in 2007-08. But anyway, that's #2 starter quality with #5 starter durability. One way to compare that to what Maholm does is to add in another 40 IP of replacement level starting pitching to even things out, since when de la Rosa can't start, his team has to use it's Karstenses and McCutchenses in his place. Replacement level tRA+ is about 75, so with de la Rosa plus a replacement you'd get about 190 IP with a 95-100 tRA+, or basically exactly what you'd get from Maholm.

Um, so I guess I just wanted to point out that it's very important to factor durability into your assessments of starting pitchers, and then I had fun playing around with the numbers. Hope some of you found this interesting.

I would have to throw this evaluation method out from the start.  First it doesn't account for context.  We can only count on 180+ innings from Duke and Maholm because their are almost no other options.  When the entire staff is bad, someone still has to pitch.  That's how Duke and Maholm get their innings.  Second, any evaluation method that shows Duke and Maholm as threes or having value is clearly a highly flawed system.  As we saw in another thread, Duke and by extension his twins, Maholm are not good.

While I have my issues with WAR, it does seem to capture the quality of this year's staff.  Using the WAR standard, this year's rotation is nothing but 5th starters and worse.

http://www.fangraphs.com/winss.aspx?team=Pirates&pos=all&stats=pit&qual=0&type=6&season=2010&month=0

[bobster]

I know people love to play with the sabermetrics but a lot of teams just don't have the pitchers with the suitable numbers to plug in to determine their rotation according to any set MLB-wide formula. So even if we find the perfect formula to rank pitchers, a team is still limited to the pitchers it has or can obtain, regardless of whether their stats fit that formula or not. So it becomes very subjective. Somebody has to be the Pirates' #1, 2, 3, 4 and 5 even if they don't have the stats to fit the profile. You don't need a Harvard degree in math to see that Maholm and Duke are not high-quality pitchers and would be low or not even in the rotation of a decent staff. Ditto for the rest of the Pirates starters. A #1 starter on one team is a #5 or worse on another. No overall formula is going to change that.

[GoBucs21]

I know people love to play with the sabermetrics but a lot of teams just don't have the pitchers with the suitable numbers to plug in to determine their rotation according to any set MLB-wide formula. So even if we find the perfect formula to rank pitchers, a team is still limited to the pitchers it has or can obtain, regardless of whether their stats fit that formula or not. So it becomes very subjective. Somebody has to be the Pirates' #1, 2, 3, 4 and 5 even if they don't have the stats to fit the profile. You don't need a Harvard degree in math to see that Maholm and Duke are not high-quality pitchers and would be low or not even in the rotation of a decent staff. Ditto for the rest of the Pirates starters. A #1 starter on one team is a #5 or worse on another. No overall formula is going to change that.

There is probably no chance that Duke and Maholm would get as many innings as they do for Pirates, on any other team.  Morton probably wouldn't even be on the roster.

The dearth of quality starting pitching on this team is amazing.  There aren't fans of many teams, who could look at a guy who came out of nowhere and see him as the best they have after a half dozen starts. 

The Pirates just don't have any "proven" quality starters.  If Maholm were on the same kind of contract as Duke, I would non-tender both.

[gorillagogo]

I would have to throw this evaluation method out from the start.  First it doesn't account for context.  We can only count on 180+ innings from Duke and Maholm because their are almost no other options.  When the entire staff is bad, someone still has to pitch.  That's how Duke and Maholm get their innings.  Second, any evaluation method that shows Duke and Maholm as threes or having value is clearly a highly flawed system.  As we saw in another thread, Duke and by extension his twins, Maholm are not good.

While I have my issues with WAR, it does seem to capture the quality of this year's staff.  Using the WAR standard, this year's rotation is nothing but 5th starters and worse.

http://www.fangraphs.com/winss.aspx?team=Pirates&pos=all&stats=pit&qual=0&type=6&season=2010&month=0

I think the original point epoc was making was that Duke and Maholm are both durable enough to start as many innings as they have. A lot of other pitchers around the league put up better numbers when they're healthy enough to pitch but are physically unable to pitch as many innings as Duke or Maholm.

[CentralCABucsFan]

I think epoc's evaluation is quite good.  I'm not willing to throw it out the window just because the results don't fit GoBucs expected results.  Calling Duke and Maholm twins is about as bad as lumping Taillon and Allie together.  Both Duke and Maholm have had bad seasons this year (Maholm really has just been bad lately), dropping their status.  Maholm probably went from being a #2-#3 starter to a #3-#4 starter, and Duke probably went from being a solid #4 to a weak #5.  There is a big difference between the two.  Duke is at that point where many teams would not start him, but Maholm is not at that point.

Sure the Pirates lack pitching depth, but so do most teams in the majors.  What separates the Pirates from other teams is that their backup plan starters have failed miserably for several years.  They never seem to find that unexpected gem coming up from the minors like most teams do.  Signing a solid #3-#4 starter would not require the Pirates to call on their backup plan pitchers for that spot. 

We will probably drop Duke, and start the season with McDonald, Ohlendorf, Maholm, and a newly signed mediocre starter.  The number 5 spot will be a battle between the usual suspects (Morton, Lincoln, Karstens, McCutchen, etc.), and you can usually count on one starter dropping to injury, so probably there will be two spots to fill.

The following pitchers will be overpaid to the point that we should stay away:

Lee
Vazquez
Lilly
Kuroda
DelaRosa
Pavano

These pitchers might fit our range:

Padilla
Penny
Bush
Millwood
Westbrook
Heilman (I think he could be converted to a starter)

[GoBucs21]

I would have to throw this evaluation method out from the start.  First it doesn't account for context.  We can only count on 180+ innings from Duke and Maholm because their are almost no other options.  When the entire staff is bad, someone still has to pitch.  That's how Duke and Maholm get their innings.  Second, any evaluation method that shows Duke and Maholm as threes or having value is clearly a highly flawed system.  As we saw in another thread, Duke and by extension his twins, Maholm are not good.

While I have my issues with WAR, it does seem to capture the quality of this year's staff.  Using the WAR standard, this year's rotation is nothing but 5th starters and worse.

http://www.fangraphs.com/winss.aspx?team=Pirates&pos=all&stats=pit&qual=0&type=6&season=2010&month=0

I think the original point epoc was making was that Duke and Maholm are both durable enough to start as many innings as they have. A lot of other pitchers around the league put up better numbers when they're healthy enough to pitch but are physically unable to pitch as many innings as Duke or Maholm.

Maholm is but Duke has missed time this year and in the past.

There may be many pitchers on other teams, who are bad but could pitch a lot of innings, but won't get the chance because they are bad.  They Pirates don't need 2 innings eater type pitchers, who are supposed to be at the top of their rotation.  They don't need to pay upwards of $12M (Maholm is signed for $5.75M next, Duke will likely get the same, if not more) for the two of them.

[gorillagogo]

I would have to throw this evaluation method out from the start.  First it doesn't account for context.  We can only count on 180+ innings from Duke and Maholm because their are almost no other options.  When the entire staff is bad, someone still has to pitch.  That's how Duke and Maholm get their innings.  Second, any evaluation method that shows Duke and Maholm as threes or having value is clearly a highly flawed system.  As we saw in another thread, Duke and by extension his twins, Maholm are not good.

While I have my issues with WAR, it does seem to capture the quality of this year's staff.  Using the WAR standard, this year's rotation is nothing but 5th starters and worse.

http://www.fangraphs.com/winss.aspx?team=Pirates&pos=all&stats=pit&qual=0&type=6&season=2010&month=0

I think the original point epoc was making was that Duke and Maholm are both durable enough to start as many innings as they have. A lot of other pitchers around the league put up better numbers when they're healthy enough to pitch but are physically unable to pitch as many innings as Duke or Maholm.

Maholm is but Duke has missed time this year and in the past.

There may be many pitchers on other teams, who are bad but could pitch a lot of innings, but won't get the chance because they are bad.  They Pirates don't need 2 innings eater type pitchers, who are supposed to be at the top of their rotation.  They don't need to pay upwards of $12M (Maholm is signed for $5.75M next, Duke will likely get the same, if not more) for the two of them.

I'm not arguing that the Pirates need these guys or should pay $12M to keep them. I was responding to your assertion that epoc's evaluation method was somehow flawed because it lacked "context". When you state that Duke and Maholm only pitch as many innings as they do because the Pirates don't have anyone better, you're completely ignoring his point that many pitchers who put up better numbers can't pitch as many innings. He's pointing out that we tend to evaluate pitchers on quality while ignoring quantity, and noting that quantity has value too.

[epoc]

To follow up on my last post, I ran the numbers, and here are the actual WAR ranges for the rotation spots (I think you'll all be pretty surprised):

#1 starter = 4.6+ WAR
#2 = 3.2-4.5
#3 = 2.0-3.1
#4 = 1.1-1.9
#5 = 0.0-1.0

That's quite a bit more stratified than I predicted. I compared it to the CHONE projections to see how many true-talent #1 and #2 starters there are (according to CHONE). There were 10 #1 starters and 30 #2 starters. At first I was disappointed that the empirical data doesn't match the projections, but you have to consider that the empirical data will always be more stratified than the projected data, because the empirical includes random variation (meaning that the pitchers who perform like #1s in any given year will be a mix of guys with #1 talent and guys with #2 talent who are having good years). On the low end, it's the same thing in reverse. There are a lot of guys who are better than replacement-level in talent, but some of them will perform even worse than replacement because of random variation (like Morton, for instance), which drags down the empirical data.

When we talk about whether a guy is a #3 or a #4 or whatever, I presume we're talking about true-talent (i.e., what we should expect from that pitcher if we ignore random variation), so it probably makes more sense to base a chart like the one above on projections rather than on empirical data. If we do that, using CHONE again, we get:

#1 = 3.5+ WAR
#2 = 2.7-3.4
#3 = 2.3-2.6
#4 = 1.9-2.4
#5 = 1.6-1.8

But if you're building a roster rather than just arguing about how good certain players are, you'd probably want to use the empirical data and you probably wouldn't bother to call pitchers #1s or 2s or whatever. You'd just say, "Maholm can give us 180 IP at a bit worse than league average. Probably 2.5 wins or so," or "I think McDonald can post a league average ERA, but I'm not sure he can pitch a full workload. More like 140 innings, 2 wins or so."

[GoBucs21]

I would have to throw this evaluation method out from the start.  First it doesn't account for context.  We can only count on 180+ innings from Duke and Maholm because their are almost no other options.  When the entire staff is bad, someone still has to pitch.  That's how Duke and Maholm get their innings.  Second, any evaluation method that shows Duke and Maholm as threes or having value is clearly a highly flawed system.  As we saw in another thread, Duke and by extension his twins, Maholm are not good.

While I have my issues with WAR, it does seem to capture the quality of this year's staff.  Using the WAR standard, this year's rotation is nothing but 5th starters and worse.

http://www.fangraphs.com/winss.aspx?team=Pirates&pos=all&stats=pit&qual=0&type=6&season=2010&month=0

I think the original point epoc was making was that Duke and Maholm are both durable enough to start as many innings as they have. A lot of other pitchers around the league put up better numbers when they're healthy enough to pitch but are physically unable to pitch as many innings as Duke or Maholm.

Maholm is but Duke has missed time this year and in the past.

There may be many pitchers on other teams, who are bad but could pitch a lot of innings, but won't get the chance because they are bad.  They Pirates don't need 2 innings eater type pitchers, who are supposed to be at the top of their rotation.  They don't need to pay upwards of $12M (Maholm is signed for $5.75M next, Duke will likely get the same, if not more) for the two of them.

I'm not arguing that the Pirates need these guys or should pay $12M to keep them. I was responding to your assertion that epoc's evaluation method was somehow flawed because it lacked "context". When you state that Duke and Maholm only pitch as many innings as they do because the Pirates don't have anyone better, you're completely ignoring his point that many pitchers who put up better numbers can't pitch as many innings. He's pointing out that we tend to evaluate pitchers on quality while ignoring quantity, and noting that quantity has value too.

I tend ignore most quantities of most things if they are bad. I will give you a personal example.  If AirTran offered me free tickets for life (quantity), I would turn them down for Jet Blue (quality), even though I would have to pay Jet Blue.  There is no amount of bad quantity that makes me pass on quality.

Simply because the Pirates situation allow Duke and Maholm to throw lots of innings, doesn't make them any better or more valuable.  If either were on another team, they wouldn't be throwing this many innings whether they were capable of it or not.  I cannot agree that Duke and Maholm have more value because the Pirates lack of pitching allows them to throw more innings.  They are pitching by default not ability.

[epoc]

GoBucs, if you are truly interested in understanding what we are talking about, please re-read the Lookout Landing article that Matt posted, as well as what I've posted in this thread. I am confident that your concerns about context/durability will be obviated. If you still have questions, I will happily answer them. But this thread is not about your subjective opinion on how to value pitchers. It was begun by CentralCABucsFan to discuss statistical markers for starter performance, and that is what the rest of us are talking about.