Philly Front Office - Phillies

Have Belief in Relief

Around the trade deadline, many Phillies fans were hoping for a bullpen arm. That got us curious – did the Phillies really need another arm for the ‘pen? After the deadline, is their bullpen good enough to win the division?

When it comes to statistics, it is always important to keep in mind whether the stats you are looking at are predictive or not. That is, some stats indicate what has already happened without any indication as to what will happen in the future. Other stats may not be as good as explaining what has happened but are better for explaining what is likely to happen. To determine how the bullpen is likely to fare down the stretch, we want to focus on the second group of stats.

When discussing how to approach answering this question, one phrase kept coming up: “small sample size.” The thing with predictive stats is that they only become reliable with a large enough sample and the stats stabilize. Otherwise…

See, the thing about relievers is that they don’t actually pitch that much. To put it in perspective, Aaron Nola pitched 40 innings in May. Only one Phillies reliever pitched more than 40 innings from Opening Day through the end of July – Victor Arano, with 40.2. When you start trying to analyze 30-40 IP, you run into questions of whether the numbers are reliable enough to have predictive merit. For a reminder of what can happen in 30-40 inning samples, just remember former Phillie Cliff Lee, who was often good for two months of a sub-2 ERA and one month of a 6+ ERA every year. It is very dangerous to read into numbers that simply aren’t stable. For that reason, splitting individual relievers’ innings any further was not an option – there just would not have been any useful data to glean.

To get samples that were large enough to be useful, we used monthly splits of the full bullpen from Fangraphs. 102 IP for March/April, 83 for May, 94 for June, 95 for July. These are large enough numbers that there may be some predictive value within. To do the analysis, I reviewed the following stats:

Wins, Losses, and Saves: While these stats have no predictive value, they actually do help explain fan feelings.

K/9 and BB/9: Strikeouts are good. Walks are bad. Insight!

BABIP (batting average on balls in play): Generally known as the luck stat, pitching BABIP should basically always average around .300. If it drifts far above or below .300, the pitchers were lucky or unlucky. These are the seeing-eye singles, the balls that fall between three fielders, the line drives smacked right at a fielder, and the like. While batters typically have more control over their BABIP, pitchers will basically always come back to averaging .300.

LOB% and HR/FB: LOB% is the percentage of runners who are left on base and for relievers tends to average somewhere in the 72-75% range. HR/FB is the percentage of fly balls that fly out of the park and averages 10%. If either of these numbers are far outside the average, regression to the mean is expected.

ERA/FIP/xFIP: ERA measures what actually happened, while the other two are calculations of what would have happened if the above stats were not far outside the average. They are useful for turning the theory into something more graspable.

The first thing that stands out is that the bullpen had its worst ERA in June and its most losses in July. It is easy to understand why there has been some concern about the bullpen lately. But that does not mean the concern is warranted. In June, despite having their best month by both strikeouts and walks, everything else went wrong. Highest BABIP, lowest LOB%, highest HR/FB. Digging into the game logs, the numbers were destroyed by 5 bad outings: 1 by Jesmuel Valentin (who is not even a pitcher), 1 by Adam Morgan, 1 by Luis Garcia, and 2 by Hector Neris. Beyond those 5 outings, the bullpen was actually quite good. In July, the ERA was great, but the ‘pen had its lowest K/9, second highest BB/9, lowest BABIP, highest LOB%, and lowest HR/FB. The true talent of the bullpen is somewhere between these two months. This is a top 10 bullpen, and by ERA and FIP, they have been a top 5 bullpen more recently. Other than a bad June that is unlikely to be repeated, there simply was no reason to make a major move for a bullpen arm.

Improvement can also come by figuring out who is pitching best and pitching them more, and that has been the real key to their recent success. July saw the return of Pat Neshek and the emergence of Austin Davis. While Neshek will not maintain a sub-1 ERA all season, replacing the thus far ineffective Neris with the very effective Neshek had great results, and giving Davis his first run as middle reliever has greatly improved the middle innings. This has been shown in the first three days of August: the Phillies have identified Dominguez, Neshek, Hunter, Arano, and Davis as their 5 most effective relievers, and those are the 5 who have pitched (with the exception of one batter faced by Loup). Combined, they have given up 1 run in 23 plate appearances. As long as the starters keep going 6 innings, Kapler has found 5 guys who can get the job done as consistently as can be expected and can rely on only those 5.

As a final note, Neris and Ken Giles this season are perfect cautionary tales for teams who lust after relievers at the deadline. Giles was traded to the Astros in a 7 player deal that has since boiled down to Giles for Velasquez. Despite pitching practically identically by FIP and xFIP, two of Giles’ three seasons in Houston were below average due to nothing more than some bad luck. In 2016, his HR/FB spiked to a level not seen before or since. In 2018, his BABIP spiked and LOB% tanked. By FIP and xFIP, he has been an excellent pitcher all three seasons. By actual results, he did not get the job done in two of his three seasons. Hector Neris has a similar story – he was a good pitcher the last two seasons who this season suffered a spike in BABIP and an insane spike in HR/FB – over 30%. Neris has always been prone to giving up HRs, but even his career rate is closer to 13-14%. 30% is a true worst case scenario number, and unsurprisingly, he has not given up a single HR since being demoted. If the Phillies want to add a great bullpen arm, all they have to do is recall Neris, who by xFIP (which normalizes HR/FB)  has actually pitched better than he did last season.

When it comes to the bullpen, it is always important to keep in mind that when dealing with only a couple dozen innings, the vagaries of luck can have a great impact on results. It is more important to focus on what is likely to happen moving forward than what has already happened, and to do that, looking at the underlying numbers is always going to give a better picture than the results will. The Phillies have improved their ‘pen throughout the season by adding pieces along the way (their current big 5 pitched just 15.2 IP in April, 12 of which were Arano) and by identifying the most effective relievers. If they need further improvement, they need look no further than Neris, an elite reliever who just hit an unlucky stretch. As a whole, this is an extremely effective group that can get the job done down the stretch and into the playoffs.

Adam Schorr

Adam Schorr (@BusterDucks) likes advanced stats, perhaps too much. We must go deeper!