Extra MustardSI On CampusFantasyPhoto GalleriesSwimsuitVideoFanNationSI KidsTNT

Excerpt: The Book

The right -- and wrong -- time to use your ace reliever

Posted: Monday April 17, 2006 10:32AM; Updated: Monday April 17, 2006 10:32AM
Free E-mail AlertsE-mail ThisPrint ThisSave ThisMost PopularRSS Aggregators

By Tom Tango, Mitchel Lichtman and Andrew Dolphin

The following is an excerpt of the recently published THE BOOK -- Playing The Percentages In Baseball, available at InsideTheBook.com, and is being reprinted with permission from TMA Press.

Excerpt: The Book
TMA Press

The usage patterns of relief aces, along with the perceived importance of the bullpen, are continually being re-evaluated and challenged. How early to bring them in (seventh, eighth or ninth)? Should the team be leading, tied, or slightly behind? How long should they keep pitching? These are all questions that confront and confound the manager. And, given that it's only been since the 1990's that there have been commonly accepted responses to these questions, there is probably still some doubt as to what the best answers are. For example, managers generally adopt much more aggressive strategies during the playoffs. Of course, with the season on the line, and with relievers having over three months to recover from any overuse, a manager should approach the playoffs in a different manner.

The Three-Run Save

The primary motivation for modern bullpen strategy (or at least, how a team's best reliever is used) is the save rule. The record books don't make distinctions between the one-run save and the three-run save, but it's rather clear that it's much easier to save a three-run game than a one-run game. Nor do contracts make this distinction. If a reliever has a saves clause, he'd love to get in those three-run games, if only to make padding his saves that much easier. As well, relievers themselves may prefer a defined role that is based on the inning, rather than the leverage of the situation. They can appreciate the fact that there may be a situation in the seventh or eighth inning that can be a turning point for the game, but their conditioning prepares them only for the ninth inning (or perhaps two outs in the eighth inning).

There is also a psychological reason for managing according to the save rule. Blowing a three-run lead in the ninth inning, while deliberately keeping your ace reliever on the bench, would be a disastrous outcome. Media, fans, even a manager's own players and upper management, will question and criticize such a move. Win today seems to be the motivation for the manager. We are not qualified to speak to the psychological impact. Neither, however, are most people. While we should not dismiss such claims, neither should we quickly embrace them without actually studying such claims. Most people are probably risk-averse, preferring a three-month treasury bill to a twelve-month government bond, even if you show them that in most three-month periods, the twelve-month government bond will yield higher returns. The potential for a negative return is enough for people to continue to pour money into such safe investment vehicles as the short-term treasury bills. At the same time, no one will question the manager if he lets his ace sit on the bench at a very critical moment in the game, if that moment occurs in the seventh inning. After all, an even more critical moment may come in the ninth inning. And, what happens in those times when it doesn't come? Perhaps the true motivation of the manager is win today, but don't try to win too early in the game.

How easy are three-run leads, and should you use your best reliever in those situations? Do you want to bring him in, and risk not having him available for the next game; or do you bring in your second- or third-best reliever, reduce your chances of winning this game, and potentially increase your chances of winning tomorrow? Here we are, in the information age, and the managers still operate almost exclusively on instincts. Instincts are valuable if you don't have all the data and time you need to make a decision. Your experience and wisdom will control your decisions when your instincts need to be kicked in. But, the three-run lead? You have plenty of time to formulate your thinking. And, there's plenty of data to assist. Let's take a look at some.


We'll look at all games from 1999--2002 in which a reliever was brought in the ninth inning, with a three-run lead, bases empty and no outs. This happened 1,034 times. And, how often did a team lose such a game? 31 times or 3%. A team leading in the ninth with a three-run lead will win 97% of the time.

Quick, say something. Yes, that's right. They will win 97% of the time, if they use relievers as they have been used from 1999--2002. And, that means bringing in your best relievers, most of the time, to hold onto such a lead. So, it's fair to say that if you split up the 1,034 games between the ace relievers and the rest, then that 97% won't hold for both groups. Right? Let's see.

The first thing we have to do is identify the ace relievers. It should be easy enough: Percival, Hoffman, Wagner, Rivera ... But, let's try to do so in a more systematic fashion. We'll introduce a toy here. It's not something that we can prove mathematically, or that we can use as a tool to identify important aspects of the game. It's just something that tries to quantify something qualitative, while allowing for a reasonable margin for error.


Which relievers have been assigned a high level of trust from their managers? Troy Percival has faced 684 batters (after exclusions of bunts and intentional walks) in the ninth inning with the game tied, or ahead by one, two, or three runs, or in the eighth inning with the game tied, or ahead by up to two runs. He's also done a little mop-up duty, by facing fifteen batters with his team behind by at least five runs. In all, he's faced 896 batters. We can quantify his usage pattern by saying that 76.3% of the batters he's faced were when the manager needed to trust him the most, and 1.7% of the batters were when the manager required almost no level of trust. That difference, 74.7%, is his trust level. Here are the top nine relievers from 1999-2002:

Most Trustworthy Relievers, 1999-2002
Player Trust Level
Troy Percival 75%
John Wetteland 69%
Kazuhiro Sasaki 69%
Trevor Hoffman 68%
Robb Nen 67%
John Smoltz 67%
Mariano Rivera 66%
Jeff Shaw 61%
Billy Wagner 58%

It's a nice little toy, and gives us pretty much the names we expected. It gives you an instant sense of how a manager thought of his reliever. You could compile this historically to get a profile on the manager or reliever. You could change the definition to say "tying run on base, or at the plate," or any combination you want to reflect whatever your idea of trust is. But, we've given a very simple definition, which, for our purposes, yields results fairly consistent with our expectations. You can't ask for more from a toy.

The league-average, for 1999-2002, is 0%. If you see someone below 0%, this means he was used more often during mop-up innings than during late and close innings.

Anyway, back to the matter at hand. The above nine relievers are fairly representative of the idea of the ace reliever. They come in when it counts far more than when it doesn't. How did they do?

They came into the ninth inning with the bases empty and needing to get three outs to close the game 271 times. Their team lost 8 of those games: 8 divided by 271 is ... 3%.

Wait a minute. We just told you that the 1,034 times that any pitcher was brought into this situation, their team ended up losing the game 3% of the time. But, when you have the above studs pitching, their teams also lose 3% of the time? Rivera? Yup, lost once out of 32 times. 3%. Wetteland, Sasaki, Wagner? Yup, yup, yup. Each lost once in 21, 36, and 34 tries, respectively. Hoffman and Nen? 2 for 41 and 2 for 46, respectively, which is worse than 3%. Percival was perfect in 29 tries, and Smoltz in 15 tries. Put all these studs together, and you get 3%. Just like all other pitchers.

Why nine relievers? Why not more? If we extend it to the top 31 relievers, it's 19 for 769, or 2.5%. The rest of the relievers (the set-up guys and any other reliever who found himself in the ninth inning three-run lead, bases empty, no outs situation), were 12 for 265, or 4.5%. If we use this as the definition to separate our ace relievers from other relievers, then we see that the stud closer has a .02 win advantage over the non-closer.

Just for fun, what if we look at all those relievers below the 0% trust level? Those are relievers who came in more often for mop-up duty than closing duty. Surely those guys must have really hurt their team? Here are all fifteen of them:

Least Trustworthy Relievers: Alan Levine, Rich Croushore, Troy Brohawn, Scott Service, Jerry Spradlin, Matt Anderson, Kerry Ligtenberg, Guillermo Mota, Francisco Cordero, Doug Henry, David Lee, Mike Matthews, Lou Pote, Bryan Ward, John Parrish.

Their trust level as a group is minus 22%. These are the 15 guys whom the managers trusted the least, but still allowed them to come in at least once in the ninth inning with that three-run lead. As a group, they only came in twenty-five times. And, how many times do you think the manager regretted that decision? How many times would you guess? Ten times out of 25, they would blow the game? Five? Four? How about once? That's right. Jerry Spradlin came in, and his team lost. The other twenty-four times, their teams won. One divided by twenty-five. That's 4%.

As you can see, we are talking about such a small difference here. Why is that? Why don't great relievers have much more impact than just .02 wins? Because they have a three-run lead with only three outs to go! That is such a large lead in that time frame that even a bad reliever will almost always escape with his team winning the game. As we saw, it's a difference of 97.5% for the ace reliever, and 95.5% for everyone else. Our perception, or our instincts, may make us believe that the difference was much higher. But the actual data show how our minds can play tricks on us.