ALDS Game 2 Preview: Yankees vs. Twins
Curtis Granderson currently serves as the poster boy for why match-up numbers mean little during individual at-bats. Everything about his sixth-inning at-bat last night against Francisco Liriano screamed failure. He was just 4-for-25 lifetime against Liriano heading into the game, and a ground out and a strikeout made that 4-for-27. He also hits lefties poorly. Yet he found a pitch to hit and drove in the go-ahead run. It goes against all the numbers put forth in the ALDS Game 1 preview.
For today’s preview let’s look at different numbers. Rather than look at each lineup’s overall performance against pitchers of the same handedness as they will face, let’s look at similar pitchers based on batted-ball profiles. This season Andy Pettitte allowed 18.4 percent line drives, 43.9 percent ground balls, and 37.6 percent fly balls. Based on that, here are a few comparables:
Brett Cecil: 17.6%, 44.2%, 38.2%
Mark Buehrle: 15.9%, 45.7%, 38.4%
John Danks: 15.7%, 45.4%, 38.9%
There were a couple of others that were close, but we’ll go with these three since they’re all lefties who don’t throw particularly hard. Here is how the current Twins lineup fares against these guys.
It would appear, then, that Minnesota does have a decided advantage when facing left-handed pitchers similar in batted-ball profile to Pettitte. It isn’t the biggest sample, but it’s a bit more comprehensive than just going with their results against Pettitte.
This lineup would appear to hurt the Twins. Gardenhire wants to break up the lefties as to avoid a late-inning Boone Logan encounter, but his best hitters are broken up by ones who fare poorly against hitters like Pettitte. I wonder if he’d consider moving up Michael Cuddyer, considering his excellent numbers against lefties, and softer-tossing lefties in particular.
Carl Pavano has turned into a groundball machine this year, keeping 50 percent of balls in play out of the air. Sticking with right-handed pitchers with a similar batted-ball profile (17.8% LD, 51.2% GB, 30.9% FB), here are his best comparables:
Clay Buchholz: 17.7%, 50.8%, 31.5%
Gavin Floyd: 18%, 49.9%, 32.1%
Rick Porcello: 17.6%, 50.3%, 31.2%
And here is how the Yankees lineup fared against them:
We have a much smaller sample on the Yankees, but it’s clear that they don’t hit as well against hitters of Pavano’s ilk. Maybe that’s because they don’t see them often — Pavano, Floyd, and Porcello are in the AL Central, Porcello and Buchholz haven’t been around for long, and Pavano was on the team’s roster for four years and pitched in the NL before that. Mark Teixeira looks like he can do some damage, but the only other hitter with a SLG above .400 is Lance Berkman, who did it in a mere 15 PA.
Last night the match-up numbers seemed to favor the Twins. They had plenty of hitters in the lineup who fared well against lefties and at home, while the Yankees seemed to perform a bit worse against lefties and on the road. This time around, though, we’re using different numbers, the Twins have a decided advantage. It doesn’t necessarily mean that they’ll hit Pettitte, but it certainly bodes well. That’s about the best we can do with pre-game analysis.
Joe also writes about the Yankees at River Ave. Blues.
All right, I think you are confused (or I am). It seems that it is Berkman that has the SLG above .400 in only 15 PA.
I think he means OBP.
Yeah. An error while glancing on my part.
It’s all good. Point being: the few hitters with good numbers have done it in tiny samples.
You picked a sample size of 3 pitchers who have faced the Yankees, one of which is a Cy Young candidate. For the Pettitte comparables, you picked 3 guys with stats at, or significantly worse than his (career, and especially this season). I like the idea of this post, but I’m concerned that it speaks more about the comparables that you’ve chosen than anything else.
I think a better analysis might be to take the performance of those pitcher and comparing their performance against comparables to those hitters. Then again, I haven’t looked at those numbers, so it may have the same issues (along with significantly more work for you).
I like the concept of looking at similar batted ball types (and handedness), but don’t you have to look at pitch type to use it?
If Felix Hernandez just happened to have the same batted ball profile (he doesn’t), would you be comfortable lumping his stats in to get a Pavano projection?
As an example Bucholz throws 4 pitches FB(94), slider (90), CB(78), CH(82)
Pavano is primarily a 3 pitch pitcher FB(90), Slider (83), CH (81)
They have significant velocity differences on both FB and slider, and Bucholz has a tremendous changeup.
Both Gavin Floyd and Porcello don’t throw nearly as many changeups (Floyd relying more on breaking stuff, Porcello more on the fastball), so I’m not sure those are valid comparisons either.
Again interesting concept in terms of trying to get a larger sample size, but I think you have to have similar pitch type profiles as well for it to be somewhat meaningful.
Why not include K% or BB% in this analysis as well? Buchholz and Floyd strike out 2-3 more batters per 9 innings than Pavano. Cecil, Buehrle and Danks all strike out fewer batters per 9 innings than Pettitte. This difference will absolutely skew your data.
while i like the idea of trying to get larger sample sizes, using comparables has its dangers. for one, as someone above me said, they have different pitch types and thus may be completely different pitchers even with the same batted ball types. second, you picked a cy young candidate as a comparable for pavano, while pettittes comparables have all put up worse numbers than him on the season and on their career lines. while i like the idea of finding comparable pitchers to broaden the sample size, i think you need to look at: handedness, pitch types, velocity, batted ball types, K%, BB%, HR%, and maybe some others that i can’t think of right now. you certainly should be considering more than three pitchers so that one pitcher’s results can’t skew the data massively. still a good idea though to combat small sample sizes.
I agree with the above posters. This is a cool idea, and I’d like it extrapolated more, but in it’s current form doesn’t mean much. I think, as others have pointed out, it’s more important that the pitches are similar, as well as the batted ball profiles.
What would be interesting to me, and what would probably be close to impossible to program, is to just take the pitch, not the pitcher. See how the Yankees do against X pitch with Y speed and Z movement. Then, take the percentages that Pavano throws that pitch, and other pitches, and make a guess at how well they’d do. Like I said, that would be a more ideal, if much harder, way of doing this.
But, if you just stick with comparables, I think more stats should be counted. K, BB, batted ball, velocity, etc. Instead of finding guys who are very close in one aspect, find guys who are the closest overall.
oh i like that idea of using the actual pitch. like a curveball with x vertical break and y horizontal break thrown at z speed. that would be so cool if it could be broken down to that level. unfortunately i don’t think that’s possible to right now… still that’s a really cool idea!
not an unreasonable idea.
None of this matters. The Home Plate ump has once again awarded the Yankees the win.
Go look at whose been getting the outside corner tonight.
Hint…it involves a porn-stache.
PItchtrax FAIL
Keep repeating it in every thread, please. Maybe eventually you’ll convince yourself that the terrible zone was only in place with pinstripes on the mound. Too bad the rest of us have eyes.