In certain circles of the baseball internet, there's a long-running joke that Joe Kelly has "great stuff." The humor there is clear: Kelly has elite velocity, averaging 98.1 mph, and he's got a great-looking high-spin curveball, but he's never had the results to match. He's got the tools, he's got the "stuff," and he's long since made the rotation-to-bullpen shift that many failed starters make. It should have clicked by now. It hasn't, stuff or not.
Or it hadn't, at least, until the 2018 playoffs. Kelly threw 11 1/3 innings during Boston's run to a World Series title, striking out 13 while issuing zero walks and allowing only a single earned run. For a pitcher with regular control issues -- his 11.2 percent walk rate was worse than 90 percent of the other pitchers who threw at least 60 innings this year -- it was a stunning turnaround. It was only 11 1/3 innings, sure. But they were great innings.
Six of those innings -- and 10 strikeouts -- came against the Dodgers in the World Series, and they must have been impressed by what they saw, because Kelly is now a Dodger, having reportedly signed a three-year deal worth $25 million to return to his native Southern California and attempt to fill the "2017 Brandon Morrow" role that Los Angeles struggled so badly to replicate in 2018. 
For a fanbase somewhat scarred by failed three-year deals to unimpressive relievers like Brandon League and Matt Guerrier and seemingly apprehensive about this deal, that's really the question here: Can 11 admittedly great innings overshadow the 600-plus up-and-down innings that came before it? The Dodgers appear to be banking that they will.

First off, let's not worry about the contract, because it's exactly what was expected. At FanGraphs, writer Kiley McDaniel projected three years and $24 million before the offseason began, exactly what Jon Heyman had. MLB Trade Rumors suggested three years and $27 million, a nearly identical figure. Three years and $25 million is exactly right, and it's really not that much in a world where Craig Kimbrel reportedly wants north of $100 million. The money is fine. If he succeeds, it's a steal for the team. If he doesn't, it's a drop in the Dodger payroll bucket. 
It doesn't matter what his career stats are, because half of that time was spent trying unsuccessfully to be a starter with St. Louis and Boston. It doesn't matter that he has two career saves, because saves aren't important -- just look at Andrew Miller -- and he's not coming to Los Angeles to unseat Kenley Jansen, anyway. What matters is if the Dodgers think they can take the October version of Kelly and help make that the version they see over the next three seasons -- just like how the Astros have helped talented pitchers improve after arriving in Houston.

Can they? It all depends on if Kelly's October success was small-sample luck, or the results of actual changes. We can say with certainty that he made changes, at least. That's a start. 
Kelly brings an elite skillset 
Obviously, you start with the aforementioned "great stuff," and it's not hard to see what's attracting the Dodgers here. Over 600 pitchers threw at least 100 fastballs, and Kelly's 98.1 mph average velocity was fourth-best. Call it a 99th percentile skill. Over 230 pitchers threw at least 100 curveballs, and Kelly was one of only six to average over 3,000 RPM of spin rate. Call that a 99th percentile skill, too. 
Combine those two skills to find pitchers with similar fastball velocity and curveball spin, and you can't, really. It's difficult to teach velocity or spin, and Kelly is an outlier in both.

(Those two dots nearest him in the upper right are Garrett Richards, who has been so good when healthy that the Padres just gave him a two-year deal despite knowing he won't pitch in 2019, and Thomas Pressly, who, after several unmemorable years in Minnesota, became one of baseball's biggest relief weapons after being traded to the Astros.)
It's easy to see an analytically-inclined team looking at that and thinking that with minor tweaks, they can help him be the next Pressly. 
But if we already saw the beginnings of that ... what changed? How did Kelly manage that October performance in the first place?
What made the October Kelly a different pitcher
Just after the World Series, MLB.com's Mark Feinsand spoke to a Major League scout about whether Kelly's postseason heroics might have increased his market value.
"I think in this day and age, the performance in the postseason won't provide as big a bump as it might have in the past, unless there is data/information that shows a sustainable change to his stuff or how he was using it," the scout said.
That's exactly right. The small sample numbers are nice, but it has to be more than that. There has to be a reason behind it in order to put any stock into it. As you'd expect, there's no shortage of theories here. Two different local reports out of Boston pointed to a meeting Kelly had with the coaching staff during a Sept. 18-20 road trip to New York.
"Basically, went into the clubhouse after the game, said, 'What's going on? What are you feeling?'" said pitching coach Dana LeVangie to the Boston Globe in October. "From that point on, he was willing to buy in to make adjustments. This is where we're at now."
During the same meeting, WEEI reported that a pitch-tipping issue had been identified as well.
It's not hard to see all of those changes happening. First, about the tipping, let's take a look at five randomly-selected Kelly appearances, three from before the coaching meeting, and two after. The only thing they have in common is that they're all at Fenway against righty hitters. See anything notable?

It's pretty clear that Kelly changed the position of his hands as he came to the set position. For most of the year, he would bring them to his chest. After the meeting, and into the playoffs, he would hold them at his waist. If there was concern that runners on base could see his grip and signal to the hitter, or if somehow the batter himself could see what was coming, this is a pretty clear indicator.
Beyond that, it's pretty clear that his pitch usage changed as well, emphasizing the curveball more, and the slider less.
Kelly 2018 regular-season pitch usage
Four-seam: 50 percent
Curve: 19 percent
Slider: 15 percent
Changeup: 11 percent
Sinker: 6 percent
Kelly 2018 postseason pitch usage
Four-seam: 58 percent
Curve: 33 percent
Slider: 0 percent
Changeup: 8 percent
Sinker: 1 percent
While this view doesn't make it clear, the changeup was more prominent in the postseason than it might look, because he had all but abandoned it in the middle of season. (He threw it just 3 percent of the time in June, but 20 percent in September, and it was a big weapon for him in October, as Matt Kemp found out.)

Now, remember when we talked about the characteristics of his four-seamer and curveball above? Those two pitches comprised 90 percent of his postseason offerings, with the slider and sinker disappearing entirely. (As WEEI reported, Kelly replaced the slider with a pitch that used "the curveball grip but with a velocity more resembling the slider.")
So he used the curve more, and he used it differently. During the regular season, that curve would come on the first pitch 28 percent of the time. In the postseason, hitters would see it 43 percent of the time. When he got to two strikes, out came that 100 mph heat, very regularly high in the strike zone, even more than in the regular season.

Put another way, 49 percent of Kelly's two-strike fastballs were 3 feet or higher in the regular season ... and 71 percent of them were in the postseason.
There's an argument to be made that they could have spent more for a potentially more dominant and reliable reliever like Adam Ottavino, though Ottavino had a poor season as recently as 2017. There's an argument for bigger names with longer track records, like Miller or Zach Britton, though both have had serious injury concerns lately. Most of all, there's an argument that 11 1/3 innings worth of very good work just isn't enough of a look to think that Kelly is somehow improved.
Those are all fair and valid and good arguments. But you can at least see what the Dodgers are thinking here. They're thinking that Pedro Baez, Scott Alexander, Dylan Floro, Tony Cingrani, and friends in front of a coming-off-his-worst-year Jansen isn't enough. They're thinking that Kelly has now had three straight years of postseason success -- in 17 2/3 October innings since 2016, he's yet to walk a single hitter, and has a 0.51 ERA -- and they're thinking that the changes he's made and the ones they may yet make will finally unlock all that talent.
Kelly has always had that "great stuff," obviously. The Dodgers are betting they can turn that into great production. It's not hard to see what they're thinking.