It's probably tempting to say "Dallas Keuchel is back."

After two straight seasons with an ERA under 3.00, the southpaw disappointed with a 4.55 ERA in 2016. But through four starts this year, his ERA is down to just 0.96.

As impressive as his first 28 innings of 2017 have been, they have featured A-list stars of Small Sample Size Theater, like a .194 BABIP and 99.0% strand rate. We probably shouldn’t get too excited about his ERA just yet (though it has been backed up by a 3.03 FIP and 2.89 xFIP).

Another reason why it might not be wise to declare Keuchel “back" is that it's not clear he actually “went” anywhere. Last season, he did post a 111 ERA-, but also pitched to a 92 FIP-, 83 xFIP- and 85 DRA-. His run prevention numbers were hurt by a 68.4% strand rate and 16.4% home-run-per-fly-ball rate, a pair of stats that are generally the product of luck or randomness.

However, Keuchel’s peripherals and soft-contact rates did drop off a bit last season. While his ERA was inflated by random factors, it is still probably safe to say his 2016 marked a real decline from his 2014-15 performance.

Those two years were special in part because of extreme ground-ball rates, a small number of walks allowed and a ton of weak contact, and that is precisely what we have seen so far in 2017.

Killing Worms

One of the constants of Keuchel’s career has been generating a ton of ground balls. Since arriving in the big leagues in 2012, his 59.0% ground ball rate leads all 179 qualified starters.

From 2014-15, he posted a rate above 62.0%, but dipped back down to 56.7% last season (relatively speaking, at least; he still ranked second among starters last year). In his 28 innings this season, Keuchel has upped his ground-ball rate to an eye-popping 70.0%.

This would put him on pace to break the record for highest recorded ground-ball rate (dating back to 2002), which was set by Derek Lowe in 2002 (67.0%; San Diego’s Clayton Richard is also on this pace). This would be good news for Houston since that's the ideal type of contact for a pitcher to give up -- hitters can't leave the yard and they rarely go for extra-base hits.

Last season, big league hitters posted a .230 average and .249 slugging percentage on ground balls (good for a .209 wOBA and 29 wRC+). On fly balls, they hit .224 with a .674 slugging percentage, which translated to a .370 wOBA and 140 wRC+.

While it is unlikely for Keuchel to sustain such a high percentage over the next few months, his ground-ball rate does seem to be indicative of a tweaked approach.

For most of his career, Keuchel has been primarily a sinkerballer, relying on a pitch that routinely produced a ground-ball rate above 65.0%. He is still throwing his sinker more than his other pitches this year, but his usage has dropped somewhat to accommodate for more four-seam fastballs. He is still only throwing the fastball 18.8% of the the time, but this would be a career high.

Per Brooks Baseball, his sinker is producing a 78.6% ground-ball rate, which would also be a career high. That's not unexpected, though -- sinkers tend to produce grounders. The production from his four-seamer is more eye-popping, as the pitch is generating a staggering 88.2% ground-ball rate.

His fastball never came close to reaching this insane grounder rate, as the pitch had a ground-ball rate under 44.0% in four of his first five big league seasons.

The pitch does seem to be getting more vertical movement, as indicated by this chart at Brooks, but this probably does not tell the whole story. It also doesn’t explain why his sinker is getting so many more grounders, as that pitch hasn’t changed much in terms of movement.

The answer comes from the fact that Keuchel is throwing dramatically lower on all his pitches. He is actually literally off the chart here.

The southpaw has allowed only 0.64 home runs per nine innings this season, which is nearly half the big league average of 1.17. This low rate is in spite of a worse-than-average 14.3% home-run-per-fly-ball rate -- when the ball is almost never in the air, it is hard to have a home run problem.

Going Low

Keuchel is working so low that he is actually throwing out of the strike zone with great frequency.

He is throwing just over 31.5% of his pitches in the zone, per PITCHf/x data at FanGraphs, which is a career low. His career average is 42.9% and the big league average is 47.8%, while he was at 42.7% in 2016.

Jeff Sullivan wrote about this at FanGraphs earlier this week, hypothesizing that this approach may be a way to compensate for reduced velocity. Opponents are still chasing these pitches out of the zone, which has helped produce a 21.4% strikeout rate, the second-highest of his career. The strikeout rate is still below the league average, but combining it with his low home-run percentage and a 5.8% walk rate has produced a 78 FIP-.

The low zone rate should also help Keuchel manage contact, since the league is posting an 82.4 mph average exit velocity on pitches outside the strike zone and a 90.5 mph average exit velocity on pitches in the zone.

Keuchel's 87.6 mph average exit velocity ranks 24th of 63 pitchers who have allowed at least 50 balls in play, while he's second in soft-contact rate and has the third-lowest hard-hit rate, per FanGraphs. He also has a .242 xBABIP per Andrew Perpetua’s xStats.org (xBABIP combines exit velocity with launch angle and correlates with BABIP at a greater rate than exit velocity or hard/soft contact rate), suggesting his low BABIP isn't just a fluke.

xBABIP itself isn’t a great predictor of future BABIP (nothing really is), but its cousin, expected slugging (xSLG), does a better job predicting future slugging percentage. This is more good news for Keuchel, whose xSLG is just .284, putting it right in line with his real slugging percentage of .263.

While his ERA will eventually rise, Keuchel’s process has certainly been solid and we should expect the good results to continue.