Negative Differences in K-BB% with Men on Base: Alex Cobb
Chris Cwik’s piece on Tyler Skaggs and the tough times the southpaw has endured while men are on base prompted me to wonder about that aspect of pitching. Which hurlers have performed differently, at least in terms of some typically advanced statistics, after hitters have reached base? What might we ascertain from what we discover?
Generally, when a pitcher pitches while men are on base, we assume that he works from the stretch. Therefore, if a starter is considerably less effective after a hitter has reached base against him, we may hypothesize that the hurler struggles while he pitches from the stretch. However, that isn’t necessarily true, because the pitcher may do things such as change his pitch mix in order to achieve a different type of result against a future hitter. Still, we can begin to make conjectures.
A snapshot of data like this can be misleading, so it’s important not to draw concrete conclusions. But we may discover a lede or six, a thing or a half-dozen that we wish to explore further. We may stumble upon a buy or sell candidate for fantasy baseball players. Or we may not find much of interest. We may just scratch our heads some more. I do enough of that without the presence of baseball information.
Now that I’ve qualified this thing, let’s get on with a reckless claim.
| Name | Bases Empty | Men on Base | K-BB% diff | ||||||||
| TBF | HR/9 | K% | BB% | K-BB% | TBF | HR/9 | K% | BB% | K-BB% | ||
| Alex Cobb | 225 | 0.68 | 26.2% | 5.8% | 20.4% | 148 | 1.25 | 14.2% | 8.8% | 5.4% | -15.0% |
| Marco Estrada | 314 | 1.95 | 22.9% | 6.7% | 16.2% | 168 | 2.41 | 13.7% | 10.1% | 3.6% | -12.6% |
| Matt Cain | 234 | 1.15 | 22.2% | 7.7% | 14.5% | 140 | 1.51 | 12.9% | 10.0% | 2.9% | -11.6% |
| Tyler Skaggs | 256 | 0.72 | 21.9% | 6.3% | 15.6% | 170 | 0.67 | 10.6% | 6.5% | 4.1% | -11.5% |
| Francisco Liriano | 196 | 0.65 | 26.0% | 8.7% | 17.4% | 194 | 1.16 | 21.7% | 15.5% | 6.2% | -11.2% |
| C.J. Wilson | 290 | 1.11 | 24.5% | 7.9% | 16.6% | 201 | 1.05 | 17.9% | 12.4% | 5.5% | -11.1% |
| R.A. Dickey | 377 | 0.99 | 21.5% | 6.9% | 14.6% | 221 | 1.71 | 15.8% | 11.3% | 4.5% | -10.1% |
I’d been optimistic that Cobb would have a nice turnaround after the All-Star break. He might still, and a 3.76 ERA isn’t terrible. But I hadn’t realized how much he’s struggled while men are on base. His batted-ball data in those situations isn’t much different from that of those with the bases empty. In his last two seasons, in which some fantasy owners have liked him as something close to a potentially undervalued ace, his K-BB% was around 10% with men on base, but that mark has dipped noticeably.
What’s the cause? I’m not sure. But it’s conceivable that the oblique strain that forced the Tampa Bay Rays’ pitcher to miss a month and a half early this season has either lingered and affected him while he’s pitched from the stretch or prompted him to be cautious in that scenario. The more distance Cobb puts between himself and that injury, the better off he’d seem to be. But health issues, whether structural or soft-tissue, can affect players long after they return to action, perhaps for the balance of a season.
Cobb’s start on Wednesday versus the St. Louis Cardinals seems to be encouraging. He didn’t walk anyone, however. He allowed two hits in an inning once but was aided by a caught stealing. When the Cards’ leadoff man reached with a double in the sixth frame, they proceeded to hand the right-hander an out with a sacrifice. Then a fielder’s choice grounder allowed the Rays to throw out the runner who was going from third to home. What might have happened had St. Louis just decided to swing away instead of to lay down the bunt?
It’s hard to say whether Cobb will continue to perform this way while men are on base. The news isn’t immediately encouraging. He spoke earlier this month of his attempts “to get his mechanics right,” as Marc Topkin put it. Mechanical issues sound like a possibly prime reason for struggles from the stretch. Cobb wondered aloud if he hadn’t put enough time into them before he returned to action. Topkin covered the subject again prior to Cobb’s then next start, noting that the righty let an inning get away from him in his previous outing and in general has been unable to get deep into games.
Perhaps the All-Star break has given Cobb the proper time to heal. His owners wouldn’t mind hearing that he didn’t do much more than play catch with his extra four days between starts. A rebound is surely still possible, but latest information suggests that, even if one is likely, it might not be as strong as I’d hoped.
Nicholas Minnix oversaw baseball content for six years at KFFL, where he held the loose title of Managing Editor for seven and a half before he joined FanGraphs. He played in both Tout Wars and LABR from 2010 through 2014. Follow him on Twitter @NicholasMinnix.
Interesting piece. Is 36ip enough of a sample to draw from though? What about compared to his career with men on base?
Good call, indirectly, first off, and thank you. I’ll change from IP to TBF in future posts if I have more findings here.
K% has one of the lowest stabilization points of pitching statistics (seen here). BB% is close, he’s faced 148 in those situations. I’d say we’re close enough, especially since the differences are so drastic. I focused on qualifiers, basically.
As I mentioned in the post, in his last two seasons, he displayed notably better prowess in the situation in K-BB%.
I’m far from a scout or even anyone qualified to comment on pitching mechanics with great authority, but if you watch Cobb pitch, you’ll see a clear difference between how he pitches in the wind-up as compared to out of the stretch. His delivery from the windup is slow – almost delayed – where he briefly pauses mid-delivery with one leg lifted and sustained in the air before he delivers. Needless to say, his delivery from the stretch is quite different.
The broadcasters last night – I think the Cardinals telecast (I was watching on MLB.tv and flipping through a few games at the time) made a point to discuss this and how he’s widely considered to have one of the strongest “bottom halves” (i.e., legs) of all pitchers. Obviously there’s no statistical support for that, but he does put a lot of weight and balance on his grounded leg before release and it would seem to support that theory.
As to whether or not that affects the numbers listed above, I don’t know. I would guess it would help to see if there’s any differences in velocity and movement between when he pitches from the wind-up vs. when he pitches from the stretch. If he truly is generating a lot of his power from his legs, it could be neutralized somewhat by pitching out of the stretch without his delayed leg lift prior to releasing the ball.
Is your K-BB% diff stat consistent from year to year? Like will it always be the same guys or is it something that varies a lot?
I haven’t examined. I would imagine as with many differences like this, it depends on the pitcher.
It’s lead, not lede. “Lede” is sometimes used as a non-standard spelling of “lead” in journalism to describe the snazzy opening bit used to draw someone into a story (e.g., “burying the lede”). But as you are using it, meaning a tip or potentially promising path (e.g., a detective “following a lead”), it is never spelled “lede.”
Ha. How did I know that your a-hole-ry would extend to spelling as well…
It’s not spelling, it’s people preferring to use the obviously more pretentious spelling of something because, well, they’d rather be pretentious than use the conventional spelling to communicate something. Ask yourself, what motivates someone to use “lede” instead of “lead” and you might get why that peeved me.
My mistake. Thanks for pointing that out. It was just a mistake. I think evo was out of line to call “a-hole-ry,” but your assumption that I’m pretentious because of my use of a word isn’t exactly classy.
Please keep the constructive comments coming, at least.
There is another obvious and plausible explanation for differences in pitching with men on/bases empty, which should apply leaguewide I would presume. If the bases are empty, you probably just got at least one person out in a row in the at bats preceding this one (the exception being a home run). If there is someone on base, then at least one other batter has reached base against you recently. There are then several extremely plausible reasons why that means you’d probably do worse, statistically, with a man on than with bases empty on the following batter:
1. Stuff and command being correlated batter to batter, game to game—like it or not, pitchers aren’t the same random-pitch generator 30 starts a year, 100 pitches a game…
2. If you’ve given up baserunners, you’re more likely to be facing a better offense or the AL vs. NL, and relatedly:
3. If you’ve given up baserunners recently, you’re more likely to be in the meaty part of the lineup, and therefore the next player is more likely to be effective against you.
The normal split (all pitchers last 2 seasons) is ~10.5% K-BB for men on, vs. ~13.5% K-BB for men on. So when looking for outliers who are bad with men on, the baseline is about -3%.
I meant to say: 10.5% for men on base, vs. 13.5% for bases empty.
That’s not valid because it’s not a same pitcher comparison. You’d have to do the calculations for each pitcher, and take the average of that. Throwing everyone into the same mix with league stats means you’re comparing, on average, better pitchers with worse pitchers.
If you do that within-pitcher comparison, for qualified starting pitchers, the split this year is 13.63-11.72 = 1.92%.
Only taking qualified starting pitchers introduces some bias, however if you do what this author did—just lump everyone into the same boat, regardless of sample size, and discuss the biggest outlier—this is a major problem.
A routine mistake on fangraphs these days—see the piece of Nick Castellanos from earlier this season. One should expect the biggest outlier of this sort of analysis to systematically be a player with a lower sample size—this is the nature of variance.
Hey, Peter: no shit. Hence, my use of “tildas”.
And that “bias” introduces almost no error. You’re concerned about whether a normal split is 2% or 3% when we are looking at guys with a 10%-15% split? Makes sense. Love that you proceed to use only this half-year’s stats to calculate the apparent “true” value of 1.92%. You sure it wasn’t 1.94%, btw? What if we used the second half of 2007? Learn the appropriate use of precision if you really want to add value here.
We could have begun a civil conversation here, and maybe just talked about the nuances of this article. I’m not sure who you’re trying to impress with your name-calling…I was the entire audience for your comments, because no one else is reading this days-old article. I’m a pretty good audience too, I might add, as comments sections go. Now that audience is zero. Congratulations.
I think that Peter 2 is a solid audience and commenter, for what it’s worth. Thanks for the extensive back-and-forth on this.
Perhaps I’m wrong to think that something may be going on in a given year in this area, but I think given the way these windows change so frequently and that the rates stabilize quickly that there’s something to glean or at least hypothesize.
Great input, thanks.
From a quick test I just did, the K-BB splits look to be fairly flaky from year to year. I took all qualified starting pitchers from 2010-11, and looked at their aggregate K%-BB%. Then for same pitchers, I checked their rates from 2012-2013 (min. 100 IP with men on). For the 56 qualifying pitchers, the 10-11 to 12-13 correl. was 0.19. This is not meant to be a precise, scientific study (survivorship bias, etc. etc.); rather, just a quick check that suggests runners on base/off base splits for pitchers’ K%-BB% rates are not very stable over time.
I’d certainly welcome a proper study on this topic to see at what sample size, if ever, the splits stabilize.
FWIW, the correl. using xFIP splits is (unsurprisingly) similar: 0.24.
Even if K% and BB% are fairly non-noisy individually, when you subtract them the noise (variance) will add. So in other words, even if after X number of total batters faced you expect to be able to trust K% and BB% individually, it may take much longer to trust their difference.
We’re not concerned with how noisy K% and BB% (or xFIP) are; they are about as stable as pitching stats can be. We’re concerned with whether splitting a pitcher’s performance into two categories and concluding that skill is what makes them different is a reasonable assumption.