Late Inning Leads Are Becoming Less Secure

The playoff race is heating up and for teams still competing for the postseason, the stakes are the highest they’ve been all season. The spotlight shines especially bright on high-leverage relief appearances in this environment. Unfortunately for the Padres, their All-Star closer Mark Melancon took a small step backward in yesterday’s matchup against the Athletics. Melancon entered the ninth with a two-run lead but allowed two runs on three hits and a walk. The A’s went on to win 5-4 in extras. Despite the setback, Melancon has been one of the best closers in the league in 2021, converting 32 of his 37 save opportunities and leading all of baseball in saves. This season, however, has seen a ton of blown leads in late innings. In the past two seasons, save conversion rates have plummeted, diving from a stable range of 66-70% from 2002-18, down to as low as 61.7% so far this season.

At first glance, it’s easy to point to the expansion of the active roster to 26 players and an influx of injuries as the reason for baseball’s poor performance in closing out games. Save opportunities are being distributed much more widely than in the past. The chart above shows that the drop in save conversion rate actually begins in 2019. The days of multiple workhorse closers meeting the 40-save benchmark are gone. Even getting 40 save opportunities has been elusive for all but a handful of pitchers:

2021 Save Opportunities Leaderboard
Name Team G SV BS SVO ROS SVO
Mark Melancon SDP 45 32 4 36 54
Liam Hendriks CHW 47 26 5 31 47
Matt Barnes BOS 43 23 4 27 41
Edwin Diaz NYM 42 23 4 27 41
Kenley Jansen LAD 41 22 5 27 41
Raisel Iglesias LAA 44 22 5 27 41

Just two closers — Melancon and Liam Hendriks — are on a comfortable pace to eclipse 40 save opportunities, with a handful of others right around that number. Compare that to just five years ago: in 2016, six pitchers finished with 40 or more saves and a total of 12 different pitchers had over 40 save opportunities.

The total innings share is starting to shift in favor of more relief pitching than in the past, which has actually challenged the existence of the true closer. There are two big driving forces behind this. First, it’s common knowledge now that starters tend to get hit harder the third time through a lineup, so the call to the ‘pen can come quite early. Second, strikeouts have been trending up in recent years, so pitchers are hitting their pitch count limits earlier in games. With bullpens being asked to do more work more often, relief pitchers’ workloads are being managed very carefully. Most teams are using their 26th roster spot for an additional arm in the bullpen while also freely moving pitchers on and off the roster to allow sufficient rest between outings. So far this season, 370 pitchers have logged at least 10 innings in relief, with an average workload of 29 IP. Through August 4, 2019, there were 355 such pitchers with an average workload in excess of 31 IP. Relievers were making nearly two more appearances in 2019 than their ‘21 counterparts at this stage in the season. These are seemingly small differences, but the impact it has had on the game is significant. This has also trickled into high-leverage situations. Traditionally, the ball was given to a setup man in the eighth inning and the closer in the ninth inning, but these roles are increasingly being shared in many bullpens.

I mentioned earlier that the league-wide save conversion rate is bottoming out. Right now the league is on pace to reach 750 blown saves on the season, which would be a record. In an ideal world, it’s an afterthought that the closer will get the job done. For elite closers this is more or less true, even this season. Josh Hader is nearly perfect on the season, with 22 saves converted in 23 opportunities. But even with Hader, the Brewers as a team have 27 saves in 47 opportunities for a conversion rate of 57.4%.

Focusing on save conversion rate can be a bit misleading, though. A save can only be attributed to the pitcher getting the last out. Any reliever preserving the lead prior to the last pitcher is only awarded a hold. A relief pitcher in a hold situation, however, will get tagged with a blown save if he gives away the lead in that outing. The Brewers are a great example of this. Brent Suter has only had four appearances in which he’s entered the game in the ninth inning or later, yet he has six blown saves. Technically, when he enters the game in the seventh inning with a one-run lead it’s a save situation, but in reality very few pitchers will stay on to close the game and earn a three-inning save. As long as relievers are being used more, the number of total blown saves will increase along with it. The conversion rate doesn’t necessarily indicate that closers are performing worse:

Thankfully using leverage index, we can see how pitchers are performing in save, or save-like, situations. Leverage index measures how critical a situation is in a game. A value of 1.00 is average. Using the FanGraphs Splits Leaderboard tool, we can run pitching splits based on high-leverage (anything with a LI of 2.0 and above) situations. Specifically, running this on ninth inning situations should capture most situations where a team’s closer, or its best available high-leverage arm, is in action:

Relief Pitching FIP, 2015-21
Season Total RP FIP RP High Leverage 9th Inning FIP
2015 3.83 3.25
2016 3.99 3.80
2017 4.16 3.77
2018 4.06 3.94
2019 4.51 3.98
2020 4.45 4.21
2021 4.19 4.16

Pitchers are indeed performing worse in these high leverage situations in 2021, both relative to relief pitching as a whole and also compared to previous seasons. The average FIP for relievers this season is 4.19; it’s an improvement over the last two homer-happy seasons, but still a bit worse than the pre-2019 years. Also notice that the difference in FIP by the pitchers in high-leverage, ninth inning situations is nearly the same as the total relief pitching average. This runs contrary to what previous seasons have shown. Going back to 2015, relief pitcher FIP has increased from 3.83 to 4.19, but for high leverage situations, it’s increased by more than twice as much from 3.25 to 4.16.

Were the closers in 2015 better than the closers this season? Let’s look at the top performers. The top 10 closers in 2015 converted 91.3% of save opportunities with an average FIP of 2.94. So far this season, the game’s top closers are converting 87.9% of opportunities with an average FIP of 3.16. While the FIP is higher by 0.22, it’s reflective of the game as a whole (the average FIP among all pitchers is 0.26 higher, so almost exactly the same). Most of the additional blown saves this season are a function of bullpen usage. The more arms that cycle through the ‘pen, the more less-experienced pitchers who should possibly still be in Triple-A are making appearances, sometimes even high-leverage appearances.

This trend of the bullpen pitching a bigger share of innings will continue for a while. Consider the alternative. When a starter is pitching well, the manager ends up in a difficult situation (let’s all think of Kevin Cash in last year’s World Series). The third time through the order is brutally difficult for starters compared to the first two trips. This season, the league-average FIP for starters on their third time around is 4.66. That’s well above relief pitchers’ average of 4.19. Using league-average values has its flaws, but at a high level it’s safe to say that most starters should be on a short leash once they get through the order twice.

The White Sox capped off last week’s trade deadline with a move to bring in Craig Kimbrel, one of this season’s best closers. Clearly the team believes there is no such thing as enough closers. When high leverage innings come around, they want to have the absolute best options available. Adding Kimbrel to a bullpen that already has Hendriks could be one of the savviest (if most expensive) moves of the deadline. As we near October and the stakes get higher, teams with a reliable bullpen could have an even bigger edge than in the past.





Chet is a contributor for FanGraphs. Prior to FanGraphs, he wrote for Purple Row. When not writing about baseball, he is a data scientist and outdoor sport enthusiast. He can be found on Twitter at @cgutwein.

newest oldest most voted
grandbranyan
Member
Member
grandbranyan

Maybe re-classify anything before the 9th a Blown Hold?

I like the shutdown/meltdown stat a lot, but it’s kind of hard to explain to someone. “Well, you see pa, it’s easy, anytime a reliever increases or decreases his team’s WPA by .06 they are awarded a shut or melt down”.

Also feel like the divvying up of Earned Runs for inherited runners is in store for an update, with each base representing a quarter run (could get more granular & use RE24 or some such if you wanted to get real tricky).

Like Reliver A loads the bases, Reliver B gives up the bomb would be 1.5 ER for Reliver A (.75 for the runner on 3rd + .5 for the runner on 2nd + .25 for the runner on 1st), while reliever B would get 2.5 ER (.25 for the runner on 3rd + .5 for the runner on 2nd + .25 for the runner on 1st + 1.00 for the donger),

Bryz
Member
Member

I had the exact same thought. It’s not fair to attribute a blown save to a pitcher that gives up the lead, but wouldn’t receive a save had he allowed zero runs. A blown hold would be the cleanest way to add context to what really happened, plus it would be easy (well, easier than other options) to retroactively adjust players’ stats from the past.

MikeS
Member
Member
MikeS

Yeah,, I’ve been calling that a “blown hold” with my friends, but obviously that’s a fake stat. It’s possible to have many blown saves in a game, and still to even get a save in that game. If the home team has a one run lead starting the top of the 6th, each half inning has a new pitcher, and each team scores 2 runs in the 6th, 7th, and 8th, then the visitors score 2 in the 9th and proceed to hold the home team scoreless, you will have SEVEN blown saves and one actual save.

I agree with you on dividing up earned runs, too. It’s a lot different allowing an inherited runner on third with nobody out to score compared to coming in with a guy on first with two outs.

Olan
Member
Member
Olan

Maybe say “increase or decrease the probability of winning by 6%”. A bit easier to visualize.