Exploring Kolby Allard’s early success

Kolby Allard wasn’t supposed to be on the Texas Rangers’ major league roster right now. He wasn’t supposed to make the Opening Day roster.

Allard was optioned to AAA Round Rock just about a week before Opening Day, which is not something that happens to a player expected to be on the Opening Day roster. He was literally the 26th — and last — player to be announced as having made the Opening Day roster, the announcement literally coming the morning of Opening Day. His making the team owed a lot to Joely Rodriguez needing to start the season on the injured list, and the decision to have Brett Martin start the season on the injured list, despite his having pitched in several games the final week of camp. I suspect the Rangers were also scouring the waiver wire at the end of camp, looking at pitchers who were potentially going to be available for cash considerations or the like that teams didn’t have room for, and if someone the Rangers liked enough to take a flyer on had come available, Allard likely wouldn’t have made the roster.

But make the roster he did. He wasn’t used much — after pitching for an inning on Opening Day, finishing out the Rangers’ 14-10 loss at Kansas City with a 1 IP, 1 run, 1K, 1 home run outing, he didn’t make an appearance again for almost two weeks. Allard pitched a scoreless ninth on April 13 at Tampa Bay in the Rangers’ 8-3 win. Three days later, he struck out five while retiring all nine batters Baltimore Orioles batters he faced in relief of Mike Foltynewicz.

Allard continued to be used sparingly, in a long relief role, after that — he didn’t pitch again for a week, had three outings in a row with four days between each appearance, went a week before his next appearance after that, then pitched in relatively short stints three days later and four days after that, his final relief appearance coming on May 22 in an 8-4 win against the Houston Astros.

At that point Allard had made ten appearances on the year through Game 48 of the Rangers’ season. They were all in relief, the Rangers were ahead when he entered the game only once, and they were generally in the middle innings. He was being used as a middle/long man up to this point.

That said, Allard was also performing well in that role. He was sporting a 3.15 ERA through 20 innings, had struck out 23 of the 86 batters he had faced, had walked just three, and had an 11% swinging strike rate. Allard, the guy who backed into an Opening Day roster spot, had excelled — and was getting another shot at a spot in the rotation as a result.

Allard has now made three starts since being moved into the rotation, including getting the nod in Saturday’s 12-1 win against the Dodgers. He has gone 14 IP in those three outings, putting up a 2.57 ERA, striking out 14 and walking 3. That’s good!

Allard’s success this year is all the more notable because of how much he struggled in 2020. Seen as the team’s most major league ready young starting pitcher after a nice nine start run to end the 2019 season after he was acquired from the Atlanta Braves, Allard had eight starts and three relief appearances in 2020, but only logged 33.2 IP. He allowed an unsightly 29 runs, walking 20 batters and striking out 32, resulting in a 7.75 ERA.

The turnaround has made me want to know what the difference is between 2020 Kolby Allard and 2021 Kolby Allard, and so I started poking around in the stats and the numbers.

The first thing that is worth noting is that the underlying data doesn’t suggest there’s the enormous difference in performance that this year’s 2.91 ERA compared to last year’s 7.75 ERA would suggest. There is a 25% improvement in K rate — 26.6% this year, compared to 21.1% in 2020 — and the walk rate has been cut in half, from 13.2% last year to 6.5% this year. But Allard’s HR/9 is basically unchanged, despite his FB% dropping from 48% last year to 38% this year — he’s actually giving up more HR/FB than he did last year, when his 8.5% HR/FB rate was lower than normal.

The upshot is that Allard had a 4.71 FIP in 2020, and a 5.85 xFIP — not great, but not as bad as the ERA would suggest. Allard has a 3.30/3.58 FIP/xFIP this year, which is basically in line with his actual results.

And it isn’t just a matter of him getting hammered with loud contact on balls in play last year driving up the ERA — Allard had an xERA of 4.80 in 2020, compared to 3.56 this year. Allard’s BABIP is almost identical — .284 in 2020 compared to .281 in 2021. The big ERA/FIP split from 2020 wasn’t a result of an elevated BABIP, as is often the case.

No, the biggest explanation for the difference in ERA between 2020 and 2021 is what he did with runners on base. In 2021, Allard has a LOB% of 83.3% — well above average. In 2020, the percentage of runners Allard left on base was just 49.6%.

That 49.6% mark is incredibly low. Out of 158 pitchers in 2020 who threw at least 30 innings, Allard ranked 158th in LOB%. The closest pitcher to him was Jon Gray, at 54.4%. Only nine pitchers other than Allard had less than 60% of the runners they allowed left on base.

The 49.6 LOB% from 2020 is historically, aberrationally low. I went back on Fangraphs and the last time prior to Allard in 2020 that a pitcher with at least 30 IP had a LOB% below 50% was in 2015, when Mike Morin of the Angels was at 44.4%. Prior to THAT? Bobby Howry, in 2010, when he had a 49.2% rate.

That sort of extreme LOB% is going to generally be driven by dramatically different results with no one on compared to runners on base and...yep, that’s what we see in 2020. Here are Allard’s splits with no one on, a runner just on first, and runners in scoring position last year:

No one on: .147/.273/.227, 14.8% walk rate, 25% K rate, .192 BABIP (88 PAs)

Runner on first: .273/.351/.545, 8.1% walk rate, 13.5% K rate, .269 BABIP (37 PAs)

RISP: .500/.556/.727, 14.8% walk rate, 18.5% K rate, .588 BABIP (27 PAs)

What the hell?

I thought maybe there was a particular issue pitching out of the stretch, or rushing his delivery while trying to hold the runner at first, but that isn’t really borne out here — the walk and K rate are both down, and the BABIP is below normal (though he did allow two of his four home runs on the year with a runner just on first base). (Allard also didn’t allow any stolen bases last year, and had one runner thrown out trying to steal, for what it is worth).

No, its the RISP numbers that are through the roof — primarily because of BABIP.

This year, though?

No one on: .198/.235/.383, 4.7% walk rate, 31.8% K rate, .240 BABIP (85 PAs)

Runner on first: .300/.364/.350, 9.1% walk rate, 22.7% K rate, .400 BABIP (22 PAs)

RISP: .241/.313/.379, 9.4% walk rate, 15.6%, .292 BABIP (32 PAs)

Interestingly, Allard has allowed all four of his home runs on the year with no one on — that, combined with the higher K rate with no one on, makes me wonder if he’s challenging batters more when the bases are empty, and working down and trying to get more ground balls when there are runners on.

In any case...Kolby Allard had a really bad ERA last year, much worse than the underlying rates would suggest, because he was uniquely awful with runners in scoring position. Why? I don’t know. For whatever reason, that hasn’t been the case this year, and given that 2020 was a historical outlier in regards to performance with RISP, I’d suggest that its more likely than not that was an aberration rather than a true reflection of his ability.

That being said, even taking that into account, Allard is performing better this year. His K rate is up, his walk rate is down, his xwOBA of .299 is better than last year’s .333, his xERA of 3.54 is better than last year’s 4.80. So what is different between last year and this year for Allard that explains those improvements?

Is there something different with his pitch mix? He’s throwing his curveball slightly more often and his changeup slightly less often this year, per Statcast, and there’s also a slight increase in fastball usage and decrease in cutter usage, but for the most part, he’s throwing his same pitches just as often as he did last year.

Is it that he has the platoon advantage more often in 2021? Well, he’s faced slightly more lefthanders in 2021 than he did in 2020 on a percentage basis...but lefties hammered him in 2020, for a .263/.349/.553 slash line. He’s actually been worse this year against righthanders than he was in 2020, allowing a 709 OPS this year against a 666 OPS in 2020. But his slash line against lefties this year is .158/.220/.289. He has 25 Ks and 6 walks in 98 PAs against righthanders in 2021, compared to 12 Ks and 3 walks in 41 PAs against lefties.

Expected numbers correlate there as well. Allard has a .307 wOBA and .334 xwOBA against righthanders in 2021, compared to a .302 wOBA and a .320 xwOBA against them in 2020. Against lefthanders he had a .379 wOBA and a .366 xwOBA in 2020. Against lefties in 2021, however, its a .226 wOBA and a .218 xwOBA.

So both the expected and actual numbers show a dramatic improvement between 2020 and 2021 in Allard’s performance against lefthanded hitters. What if we drilled down, got more granular, and looked at how lefthanded hitters have done against Allard with runners on base between 2020 and 2021?

And...bingo. Allard had a .599 wOBA, and a .459 xwOBA, against lefthanded hitters with runners on base in 2020. This year, in 2021, its a .255 wOBA and a .285 xwOBA. That’s compared to .205/.293 wOBA/xwOBA in 2020 with no one on, and .203/.166 in 2021, for lefthanded hitters.

Of course, that still doesn’t explain why there is that improvement from last year to this year. And yes, you can say that with so few plate appearances and so small of a sample size we are slicing things too fine, and it may just be sheer chance and random fluctuations, and that may be the case. But even if it is sheer chance I want to know what, by sheer chance, he’s doing better, or hitters are doing worse, that have resulted in those random fluctuations.

Allard’s four seamer has been more effective this year, per Statcast — .290 wOBA/.287 xwOBA versus .318/.333 last year. Allard is also, though, throwing his four seamer less often against lefties — 42.6% of the time this year compared to 50.2% of the time last year, with the difference being an increase in curveballs thrown to lefties, from 16.2% to 23.9%. And the four seamer is also much more effective this year against lefties — from .404/.415 in 2020 to .244/.153 in 2021, including a 45% K rate this year and a 5% walk rate, compared to 26.7% and 20.0% in 2020. Allard has actually been slightly less effective with the four seamer against righthanders in 2021 compared to 2020, but the substantial improvement against lefties more than outweighs that.

Allard has a cut fastball he is throwing almost exactly as often this year compared to last year, and he is a little less effective with it in 2021 compared to 2020, with the drop in effectiveness being due to lefties actually hitting it a little better in 2021 than they did in 2020.

Which brings us to Allard’s curveball. That’s the pitch he’s throwing more often, particularly against lefties, this year. And that’s where we are seeing a huge improvement from 2020:

Curveball results in 2020:

.396 wOBA, .347 xwOBA, 13.6% whiff percentage, 6.1% putaway percentage.

Curveball results in 2021:

.118 wOBA, .142 xwOBA, 31.3% whiff percentage, 20.6% putaway percentage.

Damn. That’s dramatic. And while the improvement is against both lefties and righties, the change against lefthanded hitters is particularly dramatic. After a 14.3% K rate and a .453/.264 wOBA/xwOBA split in 2020, Allard has a 42.9% K rate and a .000/.115 split in 2021.

Did his curveball change between 2020 and 2021? Is it a different pitch in shape?

Not really. His curve has averaged 76.7 mph both season. His curve has essentially the same amount of vertical drop as last year (almost exactly average), though it has much less horizontal break than last year. going from breaking 5.2 inches horizontally on average in 2020 (which is 4.4 inches less than average) to 3.7 inches horizontally on average in 2021. More break is generally a good thing, so this would seem to be a step backwards for Allard’s curve.

So what is the difference with Allard’s curveball this year, compared to last year? Command. I can’t embed the images, so you will have to follow this link to see what I mean.

See where it says “Statcast Pitch Arsenal? Under there are heat maps showing where Allard has located his pitches in 2021. You see that dark red dot on the outer edge of the plate to a lefty, and just below the zone? That’s where most of his curveballs are being located this year. Its the perfect location to get a hitter to chase, as it comes in looking hittable and over the plate, and then dives just out of the strike zone.

Now click on the menu to change from 2021 to 2020. See the difference in the curveball location? In 2020, he was consistently leaving his curveball in the center part of the plate and in the strike zone. That’s not a chase pitch...that’s a pitch that is going to get a pitcher chased from the game.

And that seems to be the biggest difference in what Kolby Allard is doing in 2021 compared to 2020. He’s commanding his curveball and has turned it into a chase pitch, whereas last year he was leaving it out over the plate to get hammered. That’s allowed him to throw the curveball more often, particularly against lefties. And that has helped make his four seamer more effective.

Can he continue his success? I don’t know. But I am willing to wager that continuing with this level of success will require him to continue commanding his curveball as effectively as he has this season, and continuing to make it look like his four seamer coming out of his hand.