Calculate the hold statistic for evaluating middle relief pitcher performance, and the decent hold for a more insightful analysis.
The mainstream statistics for evaluating relief pitchers are ERA, win–loss, and save. For a middle relief pitcher, there is little chance to get wins, losses, and saves—while the ERA can be misleading for describing pitcher performance. The hold statistic gives some insight into how a middle relief pitcher performs in holding the lead during a game. This hack will discuss why a hold is helpful and how it can be misleading.
The hold is not an official Major League Baseball statistic. Despite this, many media outlets, including the Associated Press (AP) and MLB organizations, use the hold to better evaluate talent. According to the Major League Baseball’s web site definition:
The hold is not an official statistic, but it was created as a way to credit middle relief pitchers for a job well done. Starting pitchers get wins, and closers—the relief pitchers who come in at the end of the game—get saves, but the guys who pitch in between the two rarely get either statistic. So what’s the most important thing one of these middle relievers can do? “Hold” a lead. If a reliever comes into a game to protect a lead, gets at least one out and leaves without giving up that lead, he gets a hold. But you can’t get a save and a hold at the same time.
John Dewan and Mike O’Donnell, the creators of the hold, define it as:
HOLD FOR RELIEF PITCHERS: Credit a pitcher with a Hold when he enters a game in a Save situation, preserves the lead, and does not finish the game.
Currently, there are two definitions: the traditional version of the hold, for which no outs need to be recorded, and the MLB unofficial definition that requires at least one out. Fewer than 6% of the situations in 2004 (121 of 2,114) occurred in which no outs were recorded. To be sure, this hack will show results for both definitions, as well as describe the decent hold—a statistic I created for more meaningful pitching performances. Why was the hold statistic created? It was because of the flaws of using traditional pitching statistics for middle relief pitchers.
Why are relief pitcher statistics flawed? Here are several big problems:
- ERA can be misleading
If the middle relief pitcher pitches only a few innings for the season, there is not a statistically significant threshold for meaningful information. Also, if the relief pitcher strands many runners before the end of the inning, the subsequent pitchers greatly affect the previous pitchers’ ERA by allowing the runs to score (or not).
- Saves can be misleading
Saves are elusive, especially for middle relievers, because the pitcher is at the mercy of the manager to finish the game to get a save. See “Calculate Save Value” [Hack #58] for additional reasons why the save statistic can be misleading.
- Wins can be misleading
A middle relief pitcher must have the go-ahead run score during his pitching performance to get a win. This is out of the pitcher’s hands and is pure luck. Once in that situation, the pitcher must still hold the lead for the win, so some degree of skill is needed, but still…. Also, no weights are assigned to wins and losses. A relief pitcher can pitch eight innings with a one-run lead to get a win, or pitch one inning with a 10-run lead to get a win, all depending on when the single go-ahead run scored.
For example, a middle relief pitcher enters and allows no runs to score in four innings. When the closer comes in, the team has a five-run lead in the ninth inning. The closer then allows five runs to score, but the team rallies to win in the bottom of the ninth. Regardless, the win is not granted to the middle reliever because the closer blew the opportunity.
- Losses can be misleading
Although a pitcher is usually to blame for blowing a lead (which can result in a loss), it is largely out of the pitcher’s hands (or completely out of his hands, if there is a designated hitter) if the team can go ahead and score to avoid a loss.
For example, a middle relief pitcher comes into the game at the start of the sixth inning with a tied game. He allows five runs in the inning, and his team scores six at the bottom of the inning. The relief pitcher is replaced in the top of the seventh, but he still gets the win, all while allowing five runs to score in a single inning!
All of the following can make the hold statistic misleading or flawed:
The hold statistic does not take into account runners on base, which is a major factor in the difficulty of a pitcher converting a hold opportunity into a hold.
The hold statistic does not give credit proportionate to the number of innings pitched. A hold for pitching five innings is credited the same as a hold for pitching one inning, although it is five times as difficult. In the traditional definition of a hold, a pitcher can technically walk a runner and be relieved, yet still get a hold. This happened 121 times in the 2004 season. This understandably upsets many analysts, so some further define a hold so that the pitcher must get at least one out.
The hold statistic does not give credit proportionate to the size of the lead a team has at the start of the hold opportunity. Getting a hold with a five-run lead is much easier than getting a hold with a one-run lead.
The pitcher can pitch effectively, but the team is tied or behind in the score, so he does not get a hold. The pitcher must be put into a game in which his team is winning, which is at the mercy of the manager and how good the pitcher’s team is.
To calculate the number of holds (according to the original creators’ definition of not needing to record an out) for a pitcher, issue this SQL:
SELECT count(*), pitcher FROM fullpbp2004 ON [save opp]="T" AND last_pitcher_flag="F" AND (if (batting_team=0, end_home_score – end_vis_score, end_vis_score – end_home_score) > 0 GROUP BY pitcher ORDER BY pitcher
The code returns the number of holds each pitcher received for the season. For a pitcher to get a hold, he must enter with a save opportunity ([save opp] = “T”). He must not finish the game (“last_pitcher_flag="F”). And he must preserve the lead: if batting_team=0, this is end_home_score minus end_vis_score; otherwise, it is end_vis_score minus end_home_score. This difference must be greater than 0. 0means that the pitcher allowed the tying run to score, and a negative number indicates the pitcher blew the lead and the team is now behind in runs.
A more effective way to define a hold is to require that the relief pitcher record at least three outs, which I devised as a decent hold.
Here is some pseudocode to calculate decent holds. (This code won’t work exactly as written; it requires a “save opportunity” flag to be derived.)
SELECT count(*), pitcher FROM pbp2k ON [save opp]="T" AND last_pitcher_flag="F" AND (if (batting_team=0, end_home_score – end_vis_score, end_vis_score – end_home_score) > 0 AND outs_pitched > 2 GROUP BY pitcher ORDER BY pitcher
Of the 2,114 holds in 2004, 928 (44% of them) occurred with the relief pitcher getting zero, 1, or 2 outs; 121 holds were granted when the pitcher got no outs; 445 with just one out recorded; and 362 with two outs recorded. Thirteen outs recorded is the longest for any hold (Sun-Woo Kim of the Expos pitched the third out in the seventh inning against the Royals with a 2–0 lead—a very decent hold indeed).
- Example 1: Which 2004 teams had the most holds?
For the definition of not needing to record an out, there were 2,114 holds for all pitchers in 2004. The San Francisco Giants led with 106 holds, and the St. Louis Cardinals were a close second with 105. The next team (New York Yankees) drops off at 89 holds.
For the definition of needing to record at least one out, there were 1,993 holds for all pitchers in 2004. The St. Louis Cardinals led with 100 holds, the San Francisco Giants had 90, and the Philadelphia Phillies had 84.
- Example 2: Which 2004 teams had the most decent holds?
There were 1,186 decent holds for all pitchers in 2004. The San Diego Padres led with 59 decent holds, and the Philadelphia Phillies and Houston Astros were a close second with 57. The next team (New York Yankees) drops off at 53 decent holds.
- Example 3: Which 2004 teams had the fewest holds?
For the definition of not needing to record an out, the average team had 70 holds. The Seattle Mariners had just 48 holds, with the Minnesota Twins and Tampa Bay Devil Rays having 50 holds each.
For the definition of needing to record at least one out, the average team had 66.5 holds. The Seattle Mariners had just 42 holds, with the Minnesota Twins, Kansas City Royals, and Tampa Bay Devil Rays having 49 holds each.
- Example 4: Which 2004 teams had the fewest decent holds?
The average team had 39.5 decent holds. The Seattle Mariners had just 18 decent holds, with the Kansas City Royals having 23 decent holds and the Oakland Athletics having 28 decent holds.
- Example 5: Which 2004 pitchers had the largest discrepancy, depending on which definition of holds was used?
Seventy-one pitchers would receive a hold based on the definition stating that they do not need to record an out, but these pitchers would not have received credit for a hold from the MLB definition. Jason Christiansen had seven holds without recording an out, Felix Heredia had five, and four tied with four holds each.
- Example 6: Which 2004 pitchers had the most holds?
For the definition of not needing to record an out, 272 pitchers had at least one hold (47 had exactly one hold and 39 had exactly two holds). Tom Gordon of the New York Yankees led the Majors with 40 holds. Considering that he had only four saves (and six blown saves), you can see how the hold adds to the analysis of middle relievers. Ray King of the St. Louis Cardinals had 35 holds (and no saves and one blown save). Akinori Otsuka of the San Diego Padres had 34 holds (and two saves and five blown saves).
For the definition that requires at least one out, 268 pitchers had at least one hold (47 had exactly one hold and 38 had exactly two holds). Tom Gordon led with 40 holds (same as with the other hold definition). Akinori Otsuka had 34, as did Ray King. Ray King dropped one hold because he did not get an out during one of his appearances.
- Example 7: Which 2004 pitchers had the most decent holds?
Two hundred forty-one pitchers had at least one decent hold (69 had exactly one decent hold and 44 had exactly two decent holds). Akinori Otsuka led with 32 decent holds, Tom Gordon with 28, Francisco Rodriguez of the Anaheim Angels had 27, Guillermo Mota with 23, and Salomon Torres and Chris Reitsma with 21.
—Ari Kaplan