Taking Another Look at the Pitching Staff
About a month ago, the Pittsburgh Pirates had one of the best starting pitching rotations in the league. They were ranked 4th overall, thanks to an impressive stretch where they got at least six innings and two or fewer runs from each starter every time out. I looked at how good the pitching staff actually was, and found that we could expect some regression going forward with pretty much every pitcher. After looking at every pitcher, I concluded with the following:
I don’t think we needed this analysis to point out that the pitching staff isn’t as good as what we’ve seen so far. The Pirates starters as a group have the 4th best combined ERA since the end of April. They’re in the same category as teams like the Phillies and Giants. But the difference is the Pirates don’t have a rotation full of aces like Roy Halladay, Cliff Lee, and Cole Hamels, or Tim Lincecum and Matt Cain.
Looking at the above, the Pirates do have some good pitchers. All of their guys have been performing in the 3.70-4.20 ERA range. Having five pitchers who can do that every night is huge. The problem for the Pirates going forward is their offense. They have been winning games because their pitchers have been putting up unreal numbers, which has made up for their lack of hitting at times. When the pitching staff starts to regress, they will need the hitting to pick up the slack. Based on the above analysis, the Pirates can expect 6-7 innings and three earned runs allowed per start. Unfortunately, that might be too much for the offense on some nights.
Odds are that the pitching staff will regress. We’ve seen some of that already with struggles from Morton and Correia in June. The idea of regression is that these guys aren’t going to go out and have a great start every time. We’re not going to see them allow two or fewer runs in each start. Eventually they’re going to get lit up for a start or two. That’s why it’s so important for the Pirates to address their hitting needs. They can’t count on the pitching staff to put up great numbers every start, which is what we’ve been seeing all year. Therefore, the offense needs to show some improvements, so that they can take advantage when their starters do put up good numbers.
Since that time, we have seen some regression from the starters. The biggest areas of regression have come from Kevin Correia and Charlie Morton. Jeff Karstens and Paul Maholm have seen a few rough outings, but still have numbers that can be considered lucky. James McDonald was the only pitcher who was performing at an unlucky rate, mostly due to his first four starts of the season. He’s now to the point where his current production matches his expected production. Let’s take a look at each starter once again to see exactly where they stand one month later.
ERA: 3.27 (Career: 4.32, Last Month: 3.08)
K/9: 5.62 (Career: 5.58, Last Month: 5.29)
BB/9: 2.97 (Career: 3.03, Last Month: 3.32)
HR/9: 0.62 (Career: 0.82, Last Month: 0.55)
BABIP: .275 (Career: .309, Last Month: .253)
Strand: 75.1% (Career: 70.8%, Last Month: 74.5%)
HR/FB: 7.4% (Career: 9.9%, Last Month: 6.3%)
Analysis: Maholm has improved his strikeout and walk numbers, which is good to see, as that means he’s putting fewer runners on, and relying less on his defense. He’s seen a regression in his BABIP and his HR/FB ratio, which explains the bump in his ERA. As of last month, he was pitching at a pace for a 4.15 ERA. His xFIP has improved to the point where his current performance warrants a 4.01 ERA. There’s still some regression that is to be expected, mostly with his strand rate. He won’t continue stranding 75% of his batters every time out, and he should start seeing more batted balls falling in to play.
xFIP: 4.01 (Last Month: 4.15)
ERA Since Last Update: 3.98
ERA: 2.49 (Career: 4.36, Last Month: 2.55)
K/9: 4.90 (Career: 4.58, Last Month: 5.29)
BB/9: 1.71 (Career: 2.49, Last Month: 1.64)
HR/9: 1.28 (Career: 1.31, Last Month: 1.55)
BABIP: .245 (Career: .279, Last Month: .240)
Strand: 84.5% (Career: 70.5%, Last Month: 88.0%)
HR/FB: 12.2% (Career: 10.4%, Last Month: 14.4%)
Analysis: Since the last update, Karstens has seen a drop in his walks and home runs, ad a jump in his strikeouts. His BABIP, strand rate, and HR/FB ratios have all regressed some. The HR/FB ratio should have brought a higher HR/9 ratio. Instead, Karstens lowered the number. A big reason for this has been his increased ground balls as of late. In his last four starts, Karstens has averaged 12 ground balls a game, up from 7 per game in his previous starts. He’s still got a lot of regression ahead of him, especially with the strand rate. There’s no way he can sustain a rate of stranding 84.5% of his base runners.
xFIP: 3.94 (Last Month: 3.81)
ERA Since Last Update: 2.25
ERA: 3.80 (Career: 5.28, Last Month: 3.80)
K/9: 5.63 (Career: 5.92, Last Month: 5.29)
BB/9: 3.88 (Career: 3.85, Last Month: 3.53)
HR/9: 0.38 (Career: 0.88, Last Month: 0.28)
BABIP: .317 (Career: .319, Last Month: .318)
Strand: 71.7% (Career: 65.3%, Last Month: 70.2%)
HR/FB: 7.4% (Career: 11.0%, Last Month: 5.5%)
Analysis: Morton has actually been pretty consistent. He had a 3.80 ERA last month. He’s had a 3.80 ERA since the last update. He has a 3.80 ERA now. His xFIP is 3.88, so his ERA is deserved. He should see a bit of regression with his home run totals, but he’s an extreme ground ball pitcher, so the regression shouldn’t impact him too much. He’s been seen as a disappointment, although I think that stems from the Roy Halladay comparisons. They look the same when pitching, but Morton is not Halladay. Right now he’s performing with a 3.80 ERA, which is in line with what you can expect from a number two starter in a weak rotation, or a number three starter in a strong rotation.
xFIP: 3.88 (Last Month: 3.71)
ERA Since Last Update: 3.80
ERA: 4.71 (Career: 4.59, Last Month: 3.74)
K/9: 4.64 (Career: 6.32, Last Month: 4.54)
BB/9: 2.06 (Career: 3.36, Last Month: 1.99)
HR/9: 1.33 (Career: 1.10, Last Month: 1.04)
BABIP: .295 (Career: .299, Last Month: .272)
Strand: 68.1% (Career: 71.5%, Last Month: 70.9%)
HR/FB: 11.6% (Career: 10.4%, Last Month: 9.0%)
Analysis: Last month I wrote that Correia’s BABIP would regress from .272, which would end up hurting him. Not only did the BABIP regress, but his strand rate and HR/FB rates also regressed, to the point where they are unlucky. Correia was a guy who should have had a 4.15 ERA, but was lucky with a 3.74 ERA. He went from that to his current state, a guy with a 4.71 ERA who should have a 4.22 ERA. He’s been unlucky the last month, and should bounce back to the 4.20 ERA range.
xFIP: 4.22 (Last Month: 4.15)
ERA Since Last Update: 9.53
ERA: 4.23 (Career: 4.02, Last Month: 4.40)
K/9: 7.57 (Career: 7.76, Last Month: 7.24)
BB/9: 4.01 (Career: 4.05, Last Month: 4.60)
HR/9: 1.34 (Career: 0.96, Last Month: 1.17)
BABIP: .310 (Career: .309, Last Month: .316)
Strand: 78.6% (Career: 75.4%, Last Month: 76.4%)
HR/FB: 11.5% (Career: 8.4%, Last Month: 10.0%)
Analysis: Last month I separated McDonald’s first four starts from the rest of his season, due to his injuries and the fact that he wasn’t fully stretched out when Spring Training broke. He’s still putting up strong numbers since those first four starts, with a 3.01 ERA, including a 3.22 ERA since the last update. He will regress from those numbers, and should end up in the 4.00 ERA range going forward. That’s better than his full season pace, which is in line with his current 4.23 ERA.
xFIP: 4.28 (Last Month: 4.50)
ERA Since Last Update: 3.22
What Does Regression Mean?
When I talk about regression, I usually get the same comments.
“Maybe something has changed with (Insert pitcher with high Strand rate) that has allowed him to pitch better with runners on base.”
“(Insert star pitcher who has performed well for years) has gotten by with a BABIP that is lower than .300. Why can’t (Insert pitcher who has a career .300 BABIP and a season BABIP of .250) do it too?”
“Maybe (Insert pitcher with a low HR/FB rate) is doing something to keep his fly balls in the park.”
Let me just say:
1. No. There is no justifying an extreme strand rate. It’s luck. You can’t do what Jeff Karstens is doing and continue to strand 85% of your base runners. It defies logic. If Karstens is doing something to pitch better with runners on base, then why isn’t he taking that same approach to keep runners off the bases?
2. No. If a guy has a career BABIP in the .300 range, and he’s having a lucky season with a .250 BABIP, you can’t compare him to an ace who has a BABIP of .277 for his career. Also, BABIP is something that’s the same for the majority of pitchers. Roy Halladay has a career .292 BABIP. Most starters usually end up in the .290-.310 range in their careers. They can have a good season, or a bad season, but most seasons they fall in this range. Suggesting someone might be the exception to this rule is just denying that the player is having a lucky season.
3. No. There is nothing a pitcher can do to control where his fly ball ends up. Across the league, an average of 10% of fly balls leave the yard for home runs. A pitcher can’t control how deep a ball is hit, or where a ball is hit. He can place pitches and hope for results, but in order to have full control he would need to make every pitch perfect, and would need to control how the batter swings at the ball.
On that last note, I will say that the park can impact this stat. Look at Paul Maholm. He’s got a 7.4% HR/FB ratio. That’s in line with his ratio over the last few years. My guess for why he’s been able to put up a “lucky” HR/FB ratio? He’s a left hander. That limits home runs against left handers, and because he pitches most of his games in PNC Park, which is unfriendly to right handed hitters, he probably sees some fly balls that would have left most parks, but were held in by PNC’s spacious left field.
When I talk about regression, I’m not saying that the pitcher will start bombing until his season ERA matches the expected ERA. I’m saying that the pitcher is getting better results than his numbers suggest he should get, and that, going forward, he can’t expect the same results. I thought Pat had a great explanation for regression:
Remember that regression to the mean doesn’t mean that if you flip a coin ten times and it lands heads ten straight times that the odds are now in favor of you flipping ten straight tails. It means that the odds are still 50/50 that on the eleventh flip you’ll get tails and that if you flip that coin 1,000 times, you’re just as likely to get a run of ten straight tails as you are ten straight heads.
Let’s apply the coin flip theory to Jeff Karstens. He’s stranding 85% of his runners so far this year. Nothing will change that fact, just like you can’t change the fact that a coin lands on heads ten times in a row. However, that doesn’t mean Karstens can expect an 85% strand rate going forward, just like it doesn’t mean the coin will continue to land on heads every time. At the same time, it doesn’t mean Karstens will regress to a number much lower than the league average of 70%, thus making up for the good luck he saw earlier in the year. That’s in line with the “the coin will now land on tails ten times in a row to make up for it landing on heads ten times in a row” thinking.
Karstens is expected to have a strand rate around 70%. Just because he’s been lucky this season, doesn’t mean his odds change. It’s the same as a coin flip. Landing on heads ten times in a row doesn’t mean a coin flip now favors heads more than 50% of the time. So going forward, Karstens can expect a 70% strand rate, which means that the other 15% of runners he was stranding will score. That will drive up his ERA, making it impossible to maintain his unreal numbers this season.
The same goes for all of the other stats above. Paul Maholm will see a regression with his BABIP. James McDonald will see a regression with his strand rate. Kevin Correia should actually see some improvements with his HR/FB ratio. Charlie Morton will see a regression with his HR/FB ratio. That doesn’t change what they’ve done. It just means that, going forward, we can’t expect the same results that we’ve seen from most of these guys all year. That’s what I’m talking about when I talk about regression. I’m not saying that their season numbers will meet their xFIP. I’m just saying their future numbers are likely to be in that xFIP range.