BALTIMORE -- Ever since arriving in Baltimore on a Minor League free agent deal last winter, Matt Harvey has spoken about wanting to reinvent himself, diving into the analytics of modern pitching in ways he didn’t need to when he was a phenom fireballer, learning to survive less on stuff and more on guile and the learning curve it would require. Twenty-three starts into the season, that process isn’t quite complete. But evidence indicates it's trending in the right direction.

It revealed itself over the course of five innings of Wednesday's 5-2 loss to the Tigers, despite Harvey suffering his first loss in five second-half starts. Allowing three runs, including Miguel Cabrera’s 499th career homer, Harvey watched his second-half ERA climb to 1.64 while he pitched in and out of traffic opposite Tarik Skubal, striking out five. The Orioles managed little more than Anthony Santander’s fourth homer in three games against Skubal and four relievers, dropping their seventh straight game.

• Box score

It wasn’t the sharpest version of Harvey, but it wasn’t the combustible version he was for much of the first half, which he finished with the highest ERA among all MLB starters. And it revealed a key reason for his general turn of fortune since the All-Star break: Harvey is a junkballer now.

“We’ve been talking about getting into a mode of pitching more and using all the weapons, really mixing in the curveball more and changeup more, really trying to change speeds and keep guys off balance,” Harvey said. “It hasn’t been easy, the whole thing, but it’s something we’ve been working on.”

Said O’s manager Brandon Hyde: “He’s come a long way since earlier in the year.”

So, what happened? What’s different with Harvey? Five starts is still objectively a small sample, but it’s also nearly a full month of data and shows a marked change in his approach and execution. Let’s dig into the numbers:

Is it … stuff?
Yes and no. Harvey isn’t throwing any new pitches, and his velocity is down slightly across the board, but not significantly. The difference is the spin Harvey is generating on his breaking balls, specifically his curveball. Harvey’s spin averaged 2,579 rpm on Wednesday, up 114 from his season average. This comes after it jumped 160 rpm in his previous start and 97 rpm in the start before that.

All told, 19 of Harvey’s 20 spinniest curves have come in his past four starts. As a result, entering Wednesday he was getting four more inches of vertical break on the pitch than he was in June and six more inches on average than in May. He’s also getting four times the horizontal break with the pitch compared to any other point in his career.

Harvey’s curveball, average horizontal break by season
2012-13: 0.5 inches
2015-17: 0 inches
2018: 1 inch
2019-20: 3.5 inches
2021: 12 inches (entering Wednesday)

“He’s executing pitches much better than he was in the first half and mixing pitches much better,” Hyde said. “I see him making a lot more quality pitches. The changeup was really good tonight, the slider was really good tonight. It’s unfortunate he hung that curveball to Miggy, but he made pitches all night long.”

Is it … pitch mix?
Yes. Harvey threw four pitches -- both fastballs, his changeup and curve -- at least 20 percent of the time each on Wednesday, while mixing in his slider 11 percent of the time. That’s notable because only from 2017-18 did he even throw three pitches at least 20 percent of the time, and during those years, 80 percent of his pitches were either fastballs or sliders. Even as he incorporated his two-seam fastball more earlier this season, Harvey was still throwing roughly 80 percent fastballs and sliders through May.

Now, that’s all changed. Harvey has sliced his fastball usage nearly 20 percent since June, doubled his curveball usage since April and is now throwing all five of his pitches each at least 15 percent of the time. That’s something he only came close to doing once over a full season, in 2015, when he was at the height of his stardom.

“It’s a trust thing,” Harvey said. “It’s learning how to use them in certain counts, learning how to not revert back to my old ways of overpowering guys with fastballs and sliders. Getting in tough situations and feeling comfortable using those pitches, with runners on and behind in the count.”

Is it … luck? And is it sustainable?
All that remains to be seen. Harvey’s peripherals are good: he’s pitching to essentially the same strikeout rate in the second half as he was in the first, while slicing his walk rate 28.5 percent and trimming his home run rate nearly in half, from 3.2 percent to 1.8 percent.

The eight hits allowed on Wednesday could be cause for alarm, but Harvey limited the damage by retiring Harold Castro to leave the bases loaded in the first and third innings. That’s crucial, given that opposing batters hit .361 with a .405 BABIP against Harvey with runners on in the first half. Entering Wednesday, they were hitting .182 with a .190 BABIP against him in such second-half situations (all per FanGraphs) until the Tigers went 3-for-11 with runners on.

Defense is another factor. The Orioles’ defense is tied for the AL’s worst by Outs Above Average, and that certainly played a part in Harvey’s first-half struggles. But with Richie Martin and Jorge Mateo up the middle and Austin Hays now entrenched in left field, Baltimore’s current unit is as athletic and capable as it's been all year.