The Mismatch That Game Six Improbably Isn’t
Remember that Josh Tomlin isn’t even really supposed to be here. Not that there’s anything wrong with him, but the Indians like Corey Kluber, and they like Trevor Bauer. They liked Danny Salazar, and they liked Carlos Carrasco. If the Indians had their druthers, Tomlin, perhaps, would be a bullpen long guy right about now. Maybe he would’ve been left off the roster entirely. Not only will Tomlin now start a game that could deliver the Indians a World Series championship — he’s going on short rest. Don’t lose sight of how the Indians are a miracle.
Of course, by name value, the Game 6 starter matchup is frightfully uneven. The Cubs are happy to be going with Jake Arrieta, because a year ago, he was maybe the best pitcher on the planet. Tomlin, meanwhile, recovered from shoulder surgery before making 10 starts. This year, Arrieta took a step back, but Tomlin basically lost his rotation spot. Go off perception, and it feels like the Cubs have a great chance of extending this all to seven. Anyone who knows anything would rather have Arrieta on the mound.
Arrieta, see, is the more talented pitcher. He’s the tougher pitcher to hit. He has higher-quality stuff. The edge Arrieta doesn’t have is in recent results. In what amounts to the most recent history, Tomlin has done a better job of pitching.
I see you over there. I see you preparing to call me an idiot. This is literally Jake Arrieta against literally Josh Tomlin. Arrieta projects for a 3.08 ERA. Tomlin, 4.48. Those are numbers we have on this very website, and it would be irresponsible of me to ignore them. I don’t mean to ignore them. By highlighting them here, I’m not ignoring them. In the longer-term, of course I see Arrieta as the better starter. What’s more relevant, though, is who’s better right now. That’s a lot less clear.
I have for you two tables, informed by FanGraphs and Baseball Savant. The tables compare Arrieta and Tomlin, and the first one shows numbers through early September, when Tomlin temporarily lost his job.
| Pitcher | IP | ERA- | FIP- | xFIP- | Exit Velo | Launch Angle | Whiff% | Strike% |
|---|---|---|---|---|---|---|---|---|
| Jake Arrieta | 174.0 | 69 | 83 | 89 | 87.3 | 8.5 | 10% | 63% |
| Josh Tomlin | 148.1 | 114 | 122 | 97 | 88.9 | 11.4 | 7% | 68% |
I think that’s about what you’d expect. Tomlin gets the edge in strikes, and the gap in xFIP- isn’t huge. But Arrieta is tougher to make contact against, and Arrieta demonstrated that he could reduce contact quality, whereas Tomlin for a while resembled a home-run machine. There’s an enormous gap in adjusted ERA. There’s a similarly enormous gap in adjusted FIP. Tomlin got the boot for a reason. And though this Arrieta wasn’t 2015’s Arrieta, this version was fine. A few too many walks.
The second table shows numbers since Tomlin returned to the rotation. The playoffs are included here, and for both pitchers, we’re looking at seven starts.
| Pitcher | IP | ERA- | FIP- | xFIP- | Exit Velo | Launch Angle | Whiff% | Strike% |
|---|---|---|---|---|---|---|---|---|
| Jake Arrieta | 40.0 | 108 | 97 | 95 | 87.5 | 13.9 | 12% | 61% |
| Josh Tomlin | 41.0 | 41 | 65 | 93 | 86.0 | 5.5 | 8% | 68% |
Arrieta remains tougher to actually hit, but the difference in strike-throwing has widened. And, over the seven starts, Tomlin’s allowed one home run. In his first 25 starts, he allowed 35. You can see some supporting evidence elsewhere. Relative to earlier, Tomlin’s average exit velocity allowed has dropped by three ticks. And his average launch angle allowed has dropped by a significant six degrees. By exit velocity, Tomlin has jumped from the 57th percentile to the 91st. By launch angle, he’s jumped from the 49th percentile to the 87th (assuming simply that lower is better). Josh Tomlin’s been getting weaker contact and grounders, while Arrieta has had increasing problems consistently finding the strike zone.
It never makes sense to ignore a projection. Again, I do believe that Arrieta remains the superior pitcher. But I think we have Arrieta in a minor relative rut, while Tomlin is on a hot streak. My sense is that pitchers are more able to make changes than hitters, and so projections might apply less often. Arrieta has had issues with his mechanics for some months. His location has suffered as a consequence. And Tomlin has altered his own approach of late, demonstrating another thing pitchers can do to at least temporarily shed their track records. As chronicled by August and Mike Petriello and others, Tomlin these days is featuring a ton of curveballs, which he’s been able to keep down.
Tomlin has also moved away somewhat from his cutter, favoring a two-seamer. Last game, actually, the cutter came roaring back, but the four-seamer nearly disappeared. The bigger, general point: Tomlin has elected to pitch differently, and hitters have hit him differently. I doubt it’s completely a fluke.
Tomlin remains a guy with a small margin of error. That’s something I’m sure he knows about himself. Arrieta gets a little more wiggle room, and for all I know, something will have clicked mechanically between his last start and this one. Even in his last start, he didn’t allow a hit for several innings. Jake Arrieta isn’t some garbage starter, and the Cubs are happy to have him, even when he’s not so precise.
But the way Tomlin’s gone, I’m not convinced this is an awful mismatch. Not in Game 6, and not with Arrieta having to face so many switch-hitters and lefties. Against Tomlin, you can put the bat on the ball, but more of those balls are now headed to the ground. Against Arrieta, contact can be tricky, but you can wait out pitches outside of the zone, and he’s likely to issue his walks. These are two very different starters, and they’re going to pitch with two very different strategies. I’m not sure one is all that much better than the other, simply in terms of this particular game day, so don’t let your brain get ahead of itself.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
Seems like the main problem for the Cubs is that Tomlin is a bit of a matchup nightmare for them with well-located breaking balls. Then again, Arrieta had a no-hitter through 5 innings last time out…
We all know what Dave thinks about hot streaks and ruts :/
I don’t see whats so crazy about ignoring 7 months of Heywards mechanics being out of whack and thinking he’ll come around in a World Series elimination game
There’s ruts then there’s mechanical issues, to me two totally different things ruts can be broken in a game or two the other takes an offseason
I really struggle with these situations. MGL and other have done such a good job training me to avoid biases by using projections as a starting point that I rarely take much comfort in something like Tomlin’s last two months or even the possible repertoire edge.
As a Tribe fan, I certainly want to believe he’s refreshed, healthy, changed, can throw the right off speed pitches to the Cubs or whatever, but history suggests a lot of those hopes end up being dashed on the hard rocks of regression to something closer to the unexciting true talent level we previously estimated.
Well, as a Cubs fan, I have no idea what we’ll get from Arrieta and no confidence in the Cubs ability to hit Tomlin (especially his curveball) in the 4 or so innings before Andrew Miller and company take over. You should probably just enjoy tonight’s game as your team is winning the World Series.
Well if you say they can’t win then I guess it’s decided. Too bad, I was looking forward to having some hope tonight and tomorrow.
The issue isn’t only hot and cold. It’s also that as dave cameron’s article yesterday shows, the secret seems to finally be out that the cubs can’t hit curve balls and tomlin is throwing a ton of them.
A small detail is the fact that Schwarber hit .125 against curveballs in 2015. We can reasonably expect Tomlin to try to exploit Schwarber’s struggles against them. Furthermore, Schwarber’s three hits have come off of hard pitches; Sinker(His double), Four Seam Fastball(RBI Single), and Cutter(RBI Single). His being a more pull happy hitter will likely mean outside Curveballs, if I had to guess.
How often did he lay off bad curves to draw walks? That seems like an important piece of the pie, as well.
Good Point, here it is:
http://www.brooksbaseball.net/plot_h_profile.php?s_type=2&gFilt=&pFilt=CU|CS&time=month&player=656941&startDate=01/01/2015&endDate=01/01/2016&minmax=ci&var=swing&balls=-1&strikes=-1&b_hand=-1.png
Tomlin’s ability to induce ground balls has added benefit given Cleveland’s infield defence and his own defensive abilities as a former infielder. Opponents have hit .213 when the ball is on the ground (MLB average .246), and the figure would be likely lower with Tomlin on the mound.
I agree there’s something different about Tomlin lately (and Arrieta), and it’s worth noting and not just pointing to the overall projections as the end all be all.
However, the one thing that really isn’t mentioned here, and it really is a complete unknown, is how Tomlin will throw on short rest. AFAIK, this is his first ever short rest start, and in general (aka, non-Kluber division), that usually means less than great things. Maybe his breaking pitches won’t be as sharp, maybe his command is a little off. I’m sure Tito will have a short hook, however, and he has an extremely rested pen and lots of options.
This should be somewhat mitigated by the shortness of his last start (58 pitches, I believe). But yeah definitely a short hook, especially with Salazar and Merritt as options, if it’s clear he doesn’t have his best stuff.
Tomlin actually pitched on 3 days rest in his second major league start in 2010 (July 31st). He came off a 93 pitch outing to throw 83 pitches in 5.1 IP against Toronto, and only gave up one run. The Indians won that game but he didn’t get the W. That was so long ago that it probably doesn’t really matter as far as the specific stats go. But generally, it seems like it shows he has the mentality for it.
In addition to that, he’s pitched out of the pen a couple times on short rest. In 2014, he made a start on 5/17 with 95 pitches then came in to finish the last 3 innings of a 13 inning game on 5/21, throwing 49 pitches and striking out 6, only giving up one run (a HR).
He threw a couple shorter outings from the bullpen on short rest after a start later that year as well. One on 2 days rest, and another on 3. Didn’t give up a run in either, though he only pitched 1.1 and 0.2 innings in those games.
Or: Baseball
This could be a matchup of Clayton Kershaw vs. Alfredo Simon and there’s STILL a decent chance that Kershaw would lose. There’s enough variance in every aspect of play that random shit just rules the day more often than we like to talk about here at Fangraphs.
Throw in the fact that we like to overanalyze playoff matchups as if every little difference between two players is a fundamentally defining precursor to the ultimate result of the entire game…
This very site has the odds at roughly 60/40 in favor of the Cubs. This is not so far removed from a coin flip that ANYTHING is truly deterministic.
It’s all very silly 🙂
I don’t really believe he is a changed pitcher. the cubs have now seen him and probably hit him better. arrieta definitely is the better pitcher. you just don’t really know what arrieta you get this year. he can throw a no hitter but he also has walk-a-thon days.
Chicago definitely should be the favorite though, I would think 65-35 in their favor.
You should keep telling everybody that Tomlin is terrible before each of his starts. You did before his last start. Eventually he’ll have a bad outing and you’ll be “right.”
Ubaldo Jimenez was on a hot streak too… the tricky thing about pitchers with bad stuff is that these hot stretches can disappear in a flash, and then it’s 6-0 opponent. Who knows what’ll happen, if anyone did, they’d own a few dozen islands by now. But Vegas has the Cubs substantial favorites on the road for a reason. Because Arrieta is substantially more likely to pitch better.
Did I say 6-0? I meant 7-0, but hey, Tomlin did give up 6 of them.
Failed to mention short rest tomlin… Think that’s very important with a command pitcher.
Hasn’t really been an issue for him in the past, when coming in for another start or to relieve on short rest after a start. Pretty small sample and anything can happen, but they had him in for a short outing last time so I would think it helps.
Hate to sounds like a broken record, broken record, broken…. Sorry….
Projections include all the recent stuff of course which changes the projections, sometimes considerably, from prior to the recent stuff, so it’s not like projections ignore the hot or cold recent performance.
More importantly, projections IN ADDITION to including recent performance in the sample/body of performance which allows us to estimate a player’s current true talent, weight more recent performance, for example, the last 3 months gets weighted around twice that of the 3 months before that, in order to account for the fact that players’ talent does indeed change over time. We never know how much changes in performance reflect true changes in talent and how much is just random fluctuation although we can with some mathematics analysis estimate that.
Anyway, many writers, including this one, want us to consider weighting recent performance even more than the projections do in their models. That is the clear inference of this and other similar articles, right?
So they are in essence saying that the models are wrong – that the models don’t properly weight recent performance. Well that’s wrong.
You see these models are based on extensive analysis of real live performance of thousand of players over many years. The weightings they come up with reflect what actually happens in real life. If, as these writers constantly suggest, “We are not giving enough weight to recent performance.” were true, then the models would reflect that. The models are built on what actually happens when, for example, a pitcher pitches at 3 RA9 for 3 years and then suddenly pitches at 3.5 or 4.0 (or 4.5) RA9 for 1 month, 3 months, one season, etc. They figure out how to properly weight that recent performance SUCH THAT THE PROJECTIONS END UP EQUALING THE ACTUAL PERFORMANCE IN THE DAY, WEEK, MONTH, ETC. FOLLOWING THAT RECENT PERFORMANCE.
So until these writers can actually show us some evidence that somehow these projections are not built on sound models (which is unlikely) I ask that they please stop wring things like, “Well the projection says this, but…”
If brilliant people like Jared (Steamer) and Dan (ZIPS) though or knew from their research that they were not weighting recent performance enough (and it is easy to find out) they would simply weight recent performance more. That fact they they don’t pretty much means that their models are correct and that it is INCORRECT to put more weight on recent performance, as these writers continually suggest.
It is human nature to want to do that – to incorrectly put TOO MUCH weight on recent events/performance as indicative of the future. It is called recency bias and is common and powerful. But sites such as these, and writers such as these are supposed to point out these biases and correct or debunk them, not encourage and purvey them.
Now obviously within that set of players who are projected as X there may be a small subset who have gotten injured or recovered from an injury or significantly changed an approach, etc., such that we may need an alternative projection model. But I stress that we must NOT consider this as the default. It is not. By definition it must be rare and it must not be insignificant. If it weren’t then the models would not fit the actual data. Because it is the default the burden is on the person questioning the projection to provide evidence that this player in question is one of the rare exceptions to the general rule that we weight recent performance at X (e.g. last 3 months is twice as important as 3 months before that) to already reflect the fact that players do change their talent due to many factors. In other words, the basic projections, as I explained, is already accounting for some degree of talent change otherwise there would be NO weighting for recency. So to give the recent performance additional weight requires a fair amount of evidence that a change in talent is a significant part of the change in performance as opposed to random fluctuation.
OK, flame away…
One more thing. Extreme recent performance is NOT any more evidence that we must give that recent performance more weight! The weighting and the fact that we obviously include that recent performance in the projection model already accounts for that. For example, say a pitcher is 3.0 RA for his whole 5 year career and then for the last 3 months he pitches absolutely horribly, say, at 5.5 RA9. You’ll say, “Well surely that alone is evidence that something is wrong mechanically or there is an injury.” You would be right! But, there it is still NOT correct to give more weight to that poor performance. By including it and weighting it more heavily that the prior 3 months and prior 4 or 5 seasons, our projection will already reflect that he may be injured or mechanically unsound. His projection will probably go from like 3.3 RA9 to 4.0 RA9, a significant change. But, we are still want to use his prior 4 1/2 years (or whatever) to account for the fact that he may NOT be injured (it surely is possible that the 5.5 RA9 was a fluke) or that he is slightly injured or that he is injured but HIS injury is not going to affect his performance as much as the same injury from a pitcher who was not as good as this one prior to the injury, or that the original 3.0 RA9 was lucky and he is really a 4.0 pitcher (or 4.5 or 3.5, etc.).
So what I’m trying to say is that yes extreme recent good or bad performance DOES suggest a change in talent more than small changes in recent performance BUT the projection models like Steamer and ZIPS already account for that with their weighting schemes, so there is NO need for additional weighting for more extreme recent performance. Again, if there were, the models would be different (they would have a less linear adjustment).