Saving Face: Reconsidering relief pitching

By Jim Poserina
published in the Baseball Research Journal (No. 33)

I want to thank all the pitchers who
couldn't go nine innings, and manager
Dick Howser who wouldn't let them go.
— Dan Quisenberry, on winning the
1982 Rolaids Relief Man Award

As baseball grew over its first half-century and the manner in which it is played has evolved, new statistics have appeared and vanished. But only one statistic, barely two generations old, has single-handedly altered the way the game is managed, influences the spending of millions of dollars, and determines the roles to be played by almost a quarter of a team's roster. It is the only major statistic to have been introduced since the end of the Dead Ball Era. It is also an extremely poor yardstick of the very performance it is intended to measure: the save.

In this paper, I will discuss the origin of the save rule, chronicle how the save is unique among statistics in how it dictates strategy, and explore how it unfairly rewards some relievers while leaving the majority in obscurity. I will present a new system that evaluates middle relievers and closers on the same scale and also does not credit pitchers for saving a game that is not in any jeopardy in the first place. I evaluate relief pitching in the 2004 season and apply this system to every postseason relief appearance from the first by Bucky Veil in 1903 to the last by Keith Foulke in 2004.

The data was collected by analyzing the individual game situations for each of the 2,428 regular-season games in 2004 and for each of the 4,233 all-time postseason relief appearances. Data sources included the websites of ESPN (which contained a surprisingly high number of errors and inaccuracies), Fox Sports, CNN, Yahoo!, NewsOK.com, and Retrosheet.

Origin of the Save

The idea for a statistic called a "save," specifically intended for relief pitching, began in 1952. Three National League executives began unofficially awarding a save to any pitcher that finished a winning game and was not the winning pitcher, regardless of the score.

The first formula for the save was written by the legendary sportswriter Jerome Holtzman as he sat on the Chicago Cubs' team bus outside St. Louis's Chase Hotel in 1960. Holtzman wanted a way to credit Cub relievers Don Elson and Bill Hendry, who were routinely protecting late-inning leads in statistical obscurity. Holtzman showed his formula to Cubs manager Lou Boudreau, who thought it a good idea, as did J. G. Taylor Spink, editor and publisher of The Sporting News and future namesake of the Hall of Fame's award for baseball writers. Spink began publishing the unofficial stat and awarding an annual trophy to the top reliever in each league. After a decade of lobbying by the Baseball Writers Association of America, the Scoring Rules Committee formally adopted the save in 1969, with a few minor changes from Holtzman's original formula. This was still the era of the four-man rotation where the complete game was the exception rather than the rule.

The save is defined as follows in Rule 10.20:

Credit a pitcher with a save when he meets all three of the following conditions:

1. He is the finishing pitcher in a game won by his club; and
2. He is not the winning pitcher; and
3. He qualifies under one of the following conditions
1. He enters the game with a lead of no more than three runs and pitches for at least one inning; or
2. He enters the game, regardless of the count, with the potential tying run either on base, or at bat, or on deck (that is, the potential tying run is either already on base or is one of the first two batsmen he faces); or
3. He pitches effectively for at least three innings. No more than one save may be credited in each game.

Relievers Today

The save is quickly becoming as ridiculous as measuring batting performance based solely on the number of plate appearances. Today, a manager is expected to designate a closer and use him in one of the save opportunity situations, typically that mentioned in condition 3(a). This is so that the closer can accumulate saves during the course of the season and point to that total when it is time to renew his contract. There is also a sort of baseball urban legend about managers receiving angry calls from their closers' agents the morning after a different reliever was summoned from the bullpen in a save situation.

Because of this phenomenon, managers tend to put some of their bullpen decisions on autopilot because the ninth inning is made the exclusive dominion of the closer. It doesn't matter if the team is in a jam in the seventh inning; the closer can't come in because the manager has to save him for the ninth. It scarcely enters into the decision that if the team doesn't survive the seventh-inning crisis, the ninth may well not even be a save situation at all.

Another nickname for the closer is the fireman, one who comes in during an emergency to extinguish a fire. Because of how this statistic has altered the conventional wisdom, the designated closer may rarely be the fireman any longer.

Consider the following situations, in each of which a call to the bullpen is made:

1. The Yankees are leading the Giants 4-3 in the bottom of the eighth inning. San Francisco has just loaded the bases on two singles and a walk, and Barry Bonds is at the plate with nobody out.
2. The Cubs and Cardinals are tied 5-5 in the bottom of the eleventh inning. St. Louis is threatening with nobody out, runners on second and third, and Albert Pujols due up.
3. The Dodgers are leading the Brewers 4-1 in the top of the ninth inning, and three September call-ups are due up for Tampa Bay.

In the first scenario, a Barry Bonds grand slam would put the Yankees in a three-run hole. In the second, anything from a passed ball to a sacrifice fly would end the game. In the third, Los Angeles could give up two runs and still win the game. Now assume that each pitcher called in from the bullpen strikes out every batter he faces, and is lifted for a pinch-hitter in the following inning if necessary. Which pitchers will get a save, and which pitchers actually saved the game?

The influence of the save on bullpen usage reaches its most absurd level in extra-inning games on the road. Assume a tie game in the bottom of the tenth inning: One false move by the visiting reliever would immediately end the game. Say the reliever gets out of the tenth inning unscathed, and his team scores three runs in the top of the eleventh. When the closer enters the game in the bottom of the inning, he has a three-run margin of error, meaning that he could allow two upper-deck home runs and his team would still come away with the W. The ace reliever did not enter the game until his team had the lead, while a less effective and more inexperienced pitcher was summoned from the bullpen with no margin for error. To an uninformed observer this would seem extremely counterintuitive. But under Rule 10.20, the tenth inning is not a save situation and the eleventh inning is.

Intuition, or at least common sense, should tell the manager that his best relief pitcher should come into the game in the tenth inning where there is no margin for error, and to save his less effective reliever for the eleventh inning and its two-run safety net. But this intuitive reasoning is ignored by followers of the Cult of the Save, and often with disastrous results, as Alex Gonzalez showed the Yankees in Game 4 of the 2003 World Series. Shaky Jeff Weaver (5.99 ERA, 5.25 K/9IP, .320 BAA) was handed the ball in a do-or-die situation while the near-automatic Mariano Rivera (1.66, 8.02, .235) sat in the bullpen because it was not a save situation, and the rest is history.

More games are won or lost in each the seventh and eighth innings than in the ninth. Yet no pitcher that leaves a regulation game while it is still going on can be credited with a save, regardless of when the game was actually "saved," if it was at all. For a game so intertwined with numbers, there are no prominent statistics at all for the workhorses of the team, especially since starting pitchers of recent vintage are leaving the game so much earlier: the middle relievers.

We hear it so often during broadcasts that we don't even think about it. "They're bringing in Kevin Gryboski to get out of this eighth-inning jam so he can hand the ball off to John Smoltz in the ninth." One image that this conjures is that of Kevin Gryboski rushing down a football field, breaking three tackles, shaking off the safety, and jumping over a diving cornerback, only to lateral the ball to John Smoltz on the one-yard line so that Smoltz can proudly march it into the end zone and record a touchdown. Similarly, Smoltz is credited with saving the game and Gryboski is ignored. Rarely is the question asked why, if Smoltz is the better player, the ball was handed to Gryboski at all.

In an attempt to somehow remedy this, some media organizations have been tracking an unofficial statistic on their websites called the hold. This goes part of the way towards creating an honest assessment of the roles of all relief pitchers, but it is still leaves a clunky system very much in place. A pitcher can enter a three-run game with two outs and nobody on, walk five batters in a row, leave the game with his bases-loaded mess for the next pitcher to clean up, and receive credit for a hold.

It was only a handful of years ago that baseball even started officially charging pitchers with blown saves. The relievers' postseason honor, the Rolaids Relief Man Award, now includes in its formula a one-point bonus for a "tough save," defined as one in which the closer entered the game with the tying run already on base. But what is the counterpart of the tough save, the "easy save"? And is an easy save a game that was really "saved" in the first place? Apparently not, since less than eight percent of saves during the 2004 season were considered tough saves.

In light of these distortions, I have devised a system that more effectively measures the performance of all relief pitchers, and not just those with big numbers under the "S" on the backs of their baseball cards. The bullpen is no longer the Siberia to which underperforming pitchers were exiled, at least to the extent that it was in years past. The relief corps will become even more important in the future now that the terms "pitch count" and "Tommy John surgery" have firmly embedded themselves in the baseball lexicon, replacing other quaint, antiquated phrases like "complete game" and "250 innings per year." Now that Dennis Eckersley has joined Rollie Fingers in the Hall of Fame and Bruce Sutter and Lee Smith are waiting in the wings, it's time to reform the way we measure relief pitchers and discard the artificial statistic that creates contrived heroes.

But don't take my word for it.

Just ask former Dodger reliever Mike Marshall: "If you're pitching with a three-run lead and you get to start an inning where there's no trouble to start with, that's not that big of a deal. Just about anybody on the ballclub should be able to do that."

Or Fingers, on Gagne's streak of 84 consecutive converted saves: "If he would've been in situations like myself or [Goose] Gossage when we were pitching three or four innings to get a save, he might not have gotten that many in a row."

21st-Century Statistics for 21st-Century Bullpens

After watching relievers like Eric Gagne and Mariano Rivera pile on save after save for closing out games that really didn't need rescuing, while uncelebrated middle relievers were stuck with the real dirty work, it became clear to me that modern bullpens had long outgrown the simple, misleading statistic of almost two generations ago. The save just wasn't doing the job anymore, and the longer it continued, the more late-inning disasters would occur as the team's best reliever sat in the bullpen waiting for the ninth inning to come around, and the higher the dollars-to-innings-pitched ratio of the chosen few would climb.

My aim was to devise statistics that are only invoked in situations where there is an immediate potential crisis. Just as the word "relieve" means "to free from distress" and "to rescue from a siege," I considered situations wherein the incoming pitcher would not only save the game, but save the exiting pitcher.

The statistics that I have created are the Saved Lead, Blown Lead, Saved Game, and Blown Game.

One drawback is that at first glance, this new system isn't as simple as the existing rule insofar as determining what constitutes a save situation. From the standpoint of effectively measuring and ranking all relief pitchers together with one formula, one advantage is that the new system isn't as simple as the existing rule. While this system looks much more complicated and unwieldy than it actually is, if the save is going to be redefined, it might as well be done right.

The short version of the system is as follows:

• Award a saved game when a pitcher comes into a game in the sixth inning or later, with the tying run on base or at bat, with a lead of at least one run, does not give up the lead, is the finishing pitcher in the game, and is not the winning pitcher. A pitcher can't get a saved game in a tie game because that would make him the winning pitcher. This is analogous to the current save.
• Award a saved lead when a pitcher comes into a game in the sixth inning or later, with the tying run on base or at bat, regardless of the score, and gets out of the inning without giving up either the lead or the tie. This is analogous to the current hold.
• Charge a blown game whenever a pitcher comes into a game in the ninth inning or later, regardless of the score, and gives up any kind of walk-off situation before he gets out of the inning. By definition of walk-off, only a visiting pitcher is eligible for a blown game (i.e. one can't give up a walk-off double in the top of an inning).
• Charge a blown lead when a pitcher comes into a game, regardless of the score, and gives up either the lead or tie before he can get out of the inning, except when if he would qualify for a blown game.
• These four stats are only invoked when a pitcher enters the game in or after the sixth inning, and only consider the pitcher's performance during the inning in which he enters the game.

In rulebook-speak, that would appear like this:

1. For a pitcher to be credited with a Saved Game, all of the following conditions must be met:
   1. His club is leading when he enters the game, and
   2. he enters the game in the sixth inning or later, and
   3. the potential tying run is on base or at bat, and
   4. he records the third out of the inning in which he enters the game, without giving up the lead, and
   5. he is the finishing pitcher in the game, and
   6. he is not the winning pitcher.
2. For a pitcher to be credited with a Saved Lead, he must not qualify for a Saved Game, and must meet all of the conditions in either of the following scenarios:
   1. His club is leading when he enters the game, and
   2. he enters the game in the sixth inning or later, and
   3. the potential tying run is on base or at bat, and
   4. he records the third out of the inning in which he enters the game, without giving up the lead, and
   5. he is not the finishing pitcher in the game.
      * OR *
   1. The score is tied when he enters the game, and
   2. he enters in the sixth inning or later, and
   3. he records the third out of the inning in which he enters the game, without allowing a run.
3. A pitcher may be credited with both a Saved Lead and a win, or both a Saved Lead and a loss.
4. For a pitcher to be charged with a Blown Game, the following conditions must be met:
   1. The pitcher's team is the visiting team, and
   2. he enters the game in the bottom of the ninth inning, or the bottom of any extra inning, and
   3. he allows the winning run to score before he can record the third out of the inning.
5. For a pitcher to be charged with a Blown Lead, he must not qualify for a Blown Game, and must meet all of the conditions in either of the following scenarios:
   1. His club is leading by any margin when he enters the game, and
   2. he enters the game in the sixth inning or later, and
   3. he gives up the lead before recording the third out of the inning in which he enters the game.
      * OR *
   1. The score is tied when he enters the game, and
   2. he enters the game in the sixth inning or later, and
   3. he allows a run before recording the third out of the inning in which he enters the game.
6. A pitcher may be credited with either both a Blown Game and a loss, or both a Blown Lead and a win, or both a Blown Lead and a loss.
7. A pitcher may not record more than a total of one Saved Game, Saved Lead, Blown Game, or Blown Lead in any one game, but Saved Leads and Blown Leads may be awarded or charged to more than one pitcher in the same game.
8. Any pitching performance in a subsequent inning shall have no bearing on the assessment of a Blown Game, Blown Lead, Saved Game, or Saved Lead, with the exception of the game-finish requirement for the Saved Game.
9. For the purposes of calculating the relief pitcher award, the point values shall be as follows: Saved Game +5, Saved Lead +4, Win +1, Loss -2, Blown Lead -3, Blown Game -4. A pitcher shall receive points for any win or loss recorded during any relief appearance, regardless of when a pitcher entered a game or how long he pitched in the game.

All of these rules share the same common thread: the pitcher comes into the game in a crisis, one wherein either the lead or the W hangs in the balance. This is why in games where the pitcher's team is ahead, the tying run must already be on base or at bat in order to qualify for either of the two "new saves." A situation where the tying run is on deck will not be a new save opportunity since the game is not in immediate jeopardy; the current hitter would have to somehow reach base in order for the on-deck batter to come to the plate still representing the tying run, and there's something just fundamentally wrong about crediting a pitcher for getting out of a jam that he himself helped create.

Currently, a pitcher can enter the game with the tying run on deck or even still sitting on the bench and qualify for the save. If the tying run is not on base, and not on deck, and may still be in the dugout, then the game isn't really in jeop-ardy and should not be said to be saved. If the lead is jeopardized, then it is of the pitcher's own doing by allowing the inherited runners to score as well as a few of his own. Right now, a pitcher can enter a game where his team is up 5-0 with the bases loaded and two out: Imagine he allows a grand slam, walks two batters, hits a bat-ter, and then gets a fly out on the warning track. He will have allowed four runners to reach base, four runs will have scored after he entered the game, and yet he will still be credited with a save. It sure won't feel like a save to the previous pitcher to whose record three of those runs were charged.

If a pitcher comes into a game with a one-run lead and allows the opponent to tie the game, he is charged with a blown save. But if he gave up a lead of six runs instead of one, he would not receive a blown save. In essence, the current save rule penalizes a pitcher for giving up a small lead but not for squandering a big lead. This alone should be reason enough to reconsider the save rule.

There are several major departures from the conventional save under my sys-tem. More than one pitcher can get one of the new saves in the same game. One pitcher can get both a new save as well as either a win or a loss in the same game. A pitcher is eligible for a new save when he comes into a tied game.

One delicate situation is that in which a pitcher is credited with a Saved Lead, only to give up the lead in a subsequent inning. In this situation I have de-cided not to charge the pitcher with a Blown Game or Blown Lead and allowed the Saved Lead to stand. Once a pitcher has saved the lead, it is now the responsibility of the offense to score insurance runs. If a starting pitcher gives up a lead, he like-wise would not be charged with a Blown Lead each time. In the same manner, a reliever entering his second inning of work would, if he gives up the lead, be eligible for a win, loss, or no-decision. He will have done his job insofar as "saving" the game is concerned: preserve either the lead or the tie, and hand the ball off to the offense to either break the tie or extend the lead. Three outs later, when he retakes the mound, he will already be an ensconced participant in the game and will no longer be representing the cavalry riding in to save the day.

To put it another way, this is exactly what already happens with pinch-hitters whose team bats around: he is considered a pinch hitter only the first time he comes to bat. If he comes to bat a second time in the same inning and gets a sin-gle, it is not considered a pinch-hit. Similarly, I am evaluating a reliever solely on his role as a "pinch-pitcher."

Table 1. The 2004 Reliever Rankings

Only pitchers that recorded at least one of the following statistics are listed: Win (in relief), Blown Save, Blown Lead, Loss (in relief), Hold, Saved Game, Save, Saved Lead, Blown Game.

Leader in BOLD, tie in Italics:

Does Rolaids Spell Relief?

Only fifteen pitchers were in both the top 30 of the Rolaids rankings and in the top 30 of my rankings for 2004. This isn't surprising given that I do not dis-criminate against middle relievers.

Mariano Rivera had the highest Rolaids ranking for all of baseball for 2004, finishing with 157 points: 4-2-53 with 4 blown saves and 2 tough saves. He finished 15 points ahead of National League champion Eric Gagne (7-3-45, 2 BS, 3 TS). How did they fare in my spreadsheet?

My system gave Rivera 95 points, based on a record of 4-2, 6 saved leads, 3 blown leads, and 16 saved games. This placed him sixth among full-time closers and eighth overall. The award for best relief pitcher of 2004 goes to Gagne: 7-3, 12 SL, 5 BL, 21 SG.

Living in the New York area, I mentioned this project to several baseball fans, and narrowly escaped being burned at the stake for daring to suggest the he-retical thought that it is possible for there to be a reliever that is better than Mariano Rivera. The reason Mo comes up number eight validates my thesis that the save rule unfairly rewards closers. Rivera came in from the bullpen with:

• the bases empty in 48 of his 57 save opportunities (80.7%),
• the tying run on base or at bat in only 19 of those 57 chances (33.3%), and
• the tying run on base or at bat only three times in 41 save opportunities with leads of 2 or more runs (7.32%) .

Comparing this to other top closers in the rankings, we find that the games Rivera saved were actually in jeopardy relatively rarely. This hurt Rivera significantly in the rankings, since no points are awarded for closing out a game where the tying run was neither on nor up when a pitcher is called from the bullpen. Rivera appeared in 38 such games; none of the full-time closers ranked above him recorded more than 28. Rivera's ratio of save opportunities that came with the bases empty was 10 percentage points (80.7%-70.5%) higher than the Major League average, and his proportion of save chances where the tying run was on or up when he came in was 20 percentage points (33.3% to 53.6%) lower than average.

Table 2: Save opportunity analysis for the top six closers of 2004

Table 3. 2004 Inning-by-Inning Breakdown

Table 4. Team Rolaids Leaders That Did Not Lead Their Own Teams in the Rankings

Table 5: All-Time Postseason Reliever Rankings

Only pitchers that recorded 10 or more points, -10 or fewer points, or at least three of the following statistics are listed: Blown Lead, Saved Game, Saved Lead, Blown Game.

Leader in BOLD, tie in Italics:

Conclusions

Statistics are measurements of game performance. If game strategy emphasizes accumulating a statistic like the RBI, the intent is to actually increase the runs on the scoreboard first and record the RBI on the stat sheet second. It's not the K on the scorecard that a pitcher wants as much as being another one twenty-seventh of the way toward a victory. That he is credited with a K is simply a bonus.

Experience, especially over the last 25 years, has shown that virtually all managers, at one time or another, will allow their game strategy to be dictated by the desire to have one player accumulate statistics. Unlike the RBI or the K, it is the statistic itself and not the game event that it represents that is the primary goal. If this weren't true, then these opportunities would not be reserved for one specific pitcher. It is a statistic, moreover, that unlike virtually every other does not represent a single action in the course of a baseball game and, it must be said, whose definition is both artificial and arbitrary.

Game-threatening situations come and go while the closer, who is supposedly the most lights-out reliever in the entire bullpen, just sits there. The bullpen phone never rings because it is not a "save situation." Even if it is technically a save situation, the manager will often not make the call because that would require the closer to pitch more than one inning. The younger, more inexperienced pitchers are summoned forth when the game is really on the line in the seventh or eighth inning. The closer yet waits in the hope that his team can get out of the jam so that he may be handed another cupcake three-run-lead bases-empty bottom-of-the-order-due-up save. Of the 1,230 saves recorded in 2004, only eight were awarded to pitchers who came into the game with the tying run on base or at bat and pitched more than two innings, common a generation ago but almost unthinkable today; only Esteban Yan accomplished the feat more than once. Almost 93 percent of saves in 2004 did not qualify as "tough saves."

It is quite clear that the save rule has outlived its usefulness, at least in its current form, and that it is in dire need of redefinition. The practice of reserving who is supposedly the team's ace reliever for situations that do not require an ace, just so the closer can accumulate statistics, needs to be rethought as it is often not in a team's best interest. Of the top 25 closers in the Rolaids rankings, it turned out that a full 40% of them had teammates that proved to be better able to handle the tough situations anyway.

Until the save rule is changed, managers will allowing the closers to keep piling on the stats in safe situations that could reasonably have been handled by the setup man that pitched the eighth (and who will continue to receive an unofficial consolation stat that no one cares about anyway), while not coming into the game at times that you shouldn't trust to the setup man. This system effectively quantifies all relief pitching, which can no longer be said for Rule 10.20.

Note: There were a few printing errors in the published version that weren't there in the last proof I got from the editor. Click here for details.