|Maury Wills and the Value of a Stolen Base|
By David Smith
The stolen base is an exciting play which has many supporters in the ranks of managers as well as fans. However, the true effect of base stealing activity, including the negative components of caught stealing and pickoffs, has not received detailed analysis. There are three main reasons for the lack of detailed study. First, the effect or value of a particular base running event depends a great deal on the exact situation at the time: men on base, score, and number of outs. These considerations are less important when analyzing the effects of batting results. Or, to put it another way, there's nothing at all like an "average stolen base." Second, base running is relatively infrequent when compared to other offensive activities, even for the most active running individuals and teams. Therefore, there are not a lot of events to analyze. Third, any consideration of base running must include indirect and often subtle effects. For example, one must consider if a steal of second decreases the ability of the following batter to get a hit because the first baseman is no longer holding the runner (stealer) on. The greatest of these indirect effects is of course the intangible of upsetting the pitcher by diverting his attention from the batter.
The study described here is based on detailed analysis of virtually every plate appearance for Maury Wills in 1962, 1963, and 1965 with the information coming from the scorebooks of Frank Finch, the Los Angeles Times reporter who was the Dodgers' official scorer during that time. This examination of a large number of events for one of the more significant base runners of all time allows some meaningful conclusions to be drawn concerning stolen bases.
Base running events are defined here as stolen bases (SB), caught stealing (CS), and pickoffs (P0). These are summarized for Wills in Table 1. The numbers in parentheses in this table are the numbers of events actually analyzed. The differences from the true totals arise because the Finch scorebooks are missing 4 games in 1963 and 5 games in 1965. I was able to examine 307 of Wills' 322 base-running events for these three seasons. Times picked off is not an official statistic, and the 21 pickoffs analyzed here were those found during this study. Some readers may question the inclusion of pickoffs, but they are base running events which have much the same effect as caught stealing. In Wills' case it must also be noted that 10 of his 21 pickoffs would have been counted as caught stealing under an addition to rule 10.08 made in the early 1970's. This rule now says that if a runner is picked off and tries to advance, he is charged with caught stealing.
Table 1. Summary of Wills' base running events.
Table 2 presents an analysis of Wills' steal opportunities and events divided into games with events and games with no events. An opportunity is defined here as his being on base with an open base in front of him. This definition rules out situations in which Wills would have to have been the trailer on a double or triple steal. In the three seasons analyzed, he never was the trailer on a multiple steal. Table 2a contains the data for games in which no event (SB, CS, or PO) occurred, as well as the number of chances per game. In these games 20% of his initial opportunities were of third base or home. Table 2b presents the same data for games in which Wills had a base running event. Only 10% of these opportunities were of third and home. This table also totals the opportunities and gives the frequency with which an event occurred. For example, in 1962, he had 244 total steal chances and 1 20 total events for a frequency of occurrence of .49. In other words, something happened 49% of the time Wills had a chance in 1962.
Table 2. Wills' steal opportunities.
a) Games with no event.
Games Chances Chances/Game
1962 87 88 1.01
1963 83 97 1.17
1965 72 42 0.58
242 227 0.94
b) Games with event.
Chances Total Freq. of
Games Chances /Game Events Chances Event
1962 78 156 2.00 120 244 .49
1963 51 92 1.80 59 189 .31
1965 86 165 1.92 128 207 .62
215 413 1.92 307 640 .48
Several interesting points come from Table 2. Wills' total events and event frequency were quite different for the three seasons. In 1963 he had a series of leg injuries and missed 29 games. These injuries are clearly reflected in the low values for the games in which he did play. The event frequency of .62 for 1965 is very high. He started running a lot from opening day with five steals in the first two games.
After 116 games Wills had 77 SB and 20 CS compared with 60 SB and 7 CS at the same point in 1962. Wills missed game 117 and only appeared in game 118 as a pinch hitter due to a badly bruised leg. It is likely he would have stolen more than 104 bases in 1965 without the injury. It is also interesting to note that in the games with no event (Table 2a), Wills had about one chance per game. In the games with events (Table 2b), the average is nearly two chances per game.
These observations confirm an obvious expectation: more chances lead to more base running. A final point related to event frequency can be seen in Table 3. These data show that in 1962 Wills ran more often late in the year as he approached Cobb's season record.
Table 3. Event frequencies for Wills during 1962.
Month Chances Events Frequency
April 29 10 .34
May 42 21 .50
June 44 17 .39
July 38 12 .32
August 42 24 .57
Sep.-Oct. 49 36 .73
244 120 .49
Since the object of baseball is to win games by scoring runs, it is logical to look at these two areas in some detail. Each of the 307 events was examined with respect to the play by play of the inning in which it occurred. For each steal a decision was made as to whether or not the steal allowed the Dodgers to score a run they otherwise would not have scored. Similarly each CS and PO were examined for potential lost runs. Some of these decisions were easy and some were quite difficult. A few examples may be helpful.
1) May 1, 1962, Chicago at LA, 5th inning, score tied 2-2.
Drysdale leads off with single.
Wills forces him at second.
Wills steals second.
Gilliam singles Wills home.
Moon grounds out.
Snider strikes out.
Without the steal, Wills would have gone to third (or second)
on Gilliam's single, where he would not have scored on
the next two outs. Here the steal gave LA a run.
2) June 16, 1965, San Francisco at LA, 7th inning, score tied 1-1.
Ferrara grounds out.
Koufax grounds out.
Wills singles and steals second.
Gilliam singles him home.
Parker strikes out.
Again, Wills would not have gotten past third without the
3) September 25, 1963, New York at LA, 3rd inning, LA ahead
Koufax grounds out.
Wills singles and is out stealing second.
W. Davis singles.
Moon singles Davis to third.
T. Davis lines out.
Without the CS, Wills would have scored and the inning
would have continued.
After all the decisions had been made, each game was analyzed with respect to the runs which were gained or lost by Wills' running. If a Wills steal were responsible for a run in a game which Los Angeles won by one run, then he was credited with a half win, since without the run the game would have continued into extra innings, and the Dodgers might still have won. However, if the game were an extra inning game and a regulation inning steal led to a run, then Wills received a full win, since without the run they would have lost in 9 innings. Wills was credited with a half win if the Dodgers won an extra inning game by one run which only scored because of his stealing, since they still might have won in a later inning. The same logic was applied to CS and PO in reverse to determine half losses and full losses. These analyses are summarized in Table 4. It may be seen that over the three years examined, Wills' running gained the Dodgers a net of 35 runs and was directly responsible for a net of seven wins.
These results may be compared with the predictions of Pete Palmer's linear weight model which says that steals are worth 0.2 runs each, caught stealing (and pickoffs) are - 0.35 runs each, and the scoring of 10 additional runs over the course of a season will lead to one additional win. The model predicts 21 additional runs from Wills' running (238 SB X 0.2- 56 CS X0.35 - 21 PO X 0.35), which should have led to two additional Dodger wins. The results in Table 4 do not fit well with these predictions. Some of this disagreement may result from the relatively small data base used: three seasons for one man. But it must also be remembered that the model was derived by considering average performances, and the Dodgers of the early 1960's were not average teams. Not only did they win many games, but they won many close games due to their excellent pitching staff. This pattern can be seen especially in 1963 and 1965 when they were 6th and 8th in scoring but a dominant first in ERA while winning two pennants. To such teams, each run is extremely valuable and likely to be protected. Such a contention is supported by the 1965 data in which a net of only three additional runs is related to 21/z extra wins.
Table 4. Runs and wins from Wills' base running.
SB CS PO Net
1962 +27 -2 -l +24
1963 +13 -5 0 +8
1965 +18 -12 -3 +3
+58 -19 -4 +35
1962 +3 0 0 +3
1963 +1½ 0 0 +1½
1965 +4 -1 ½ +2½
+8½ -1 -½ +7
The winning of seven additional games by the scoring of only 35 extra runs through Wills' running may mean that a run scored as a result of a steal is more valuable than a run scored directly by batting. There are two aspects to this idea. First, Wills was a leadoff hitter and likely to be on when the team's good hitters were up; therefore his steals would be more likely to lead to runs than if he batted low in the order. As a consequence, running events could be increased in value to perhaps 0.25 runs for a steal and -0.40 runs for caught stealing (and pickoff), leading to a prediction of 29 runs from Wills' running. Second, as seen in the score distribution data of Table 5, Wills was most likely to run when the score was tied or the Dodgers were ahead by one run (58% of all his events). Therefore, runs from stealing were more likely to be scored when they were really needed. He also stole more often with 0 and 1 outs than the model expects, so perhaps only 6 or 7 additional runs are needed for a win if the runs result from stealing, leading to a prediction of 4½ to 5 additional wins from Wills' running. These modifications bring the model into closer agreement with Wills' actual effect.
Table 5. Score differential for Wills' steal opportunities and events (1962, 1963, and 1965 combined). Score differential is Dodger runs minus opponent runs at the time.
There is another obvious factor in base running which requires attention. That is, before a man can steal he must be on base. The act of getting on base is in itself an important offensive action of benefit to the team. The overall value of Wills to the Dodgers can be seen in the data of Table 6. (The values in this table do not add to the official totals for Wills for these seasons because the Finch scorebooks were incomplete and in some cases contained minor inaccuracies.) It is clear that Wills' presence made the Dodgers a better team. Table 6 shows that the Dodgers won more games when Wills was active on the bases, even when he was caught stealing or picked off. They did much worse when he failed to run at all. Wills' batting average and on base percentage are closely related to the Dodgers winning percentage. To a degree this correlation can be attributed to intangibles, but it cannot be denied that most teams will do well in games in which their leadoff hitter has an on base percentage of over .400.
Table 6. Wills' batting record and Dodger winning percentage. Combined for 1962, 1963, and 1965.
It is also true that games in which Wills did well are games in which his teammates were likely to do well, perhaps because the opposing pitcher was having a bad day. Wills had a phenomenal offensive record in the five games between September 6 and September 10, 1962. During this brief interval he had 13 hits in 16 at bats including.7 straight hits at one point, and stole 12 bases. The Dodgers won three of these five games.
The effect of Wills' running on the performance of those batting immediately behind him is addressed in Table 7. About 85% of these appearances were Jim Gilliam, with most of the rest being Wes Parker (in 1965). The most striking feature in this table is that batting average rose from .267 with no one on to .322 when Wills was on first, but dropped to .189 if he stole. A possible explanation of these changes is that the first baseman holds the runner on, creating a hole on the right side, but that he plays off the bag after the steal and covers more territory. Finch indicated the direction of hits in his scorebooks, so it is possible to estimate this point by totaling the singles to right field. Obviously not all of these singles were in the
Table 7. Performance of men batting behind Wills, cumulative for 1962, 1963, and 1965.
*Percentage of hits which were singles to right field.
**At the time these men completed the plate appearance.
Based on analysis of 403 plate appearances in 1962,1963,1964,1965.
same place, but a large change should be detected. Table 7 shows that the percentage of right field singles increased with Wills on first from 18% to 24%, and declined again to 19% after a steal. This change, although consistent, is not sufficient by itself to explain the batting average change. When a base stealer is on first, pitchers tend to throw more fast balls to help the catcher's chance of throwing out the runner. This pattern of pitching is of course also understood by the batter, whose average should be increased. This point can be extended by examining pitch-by-pitch box scores of 96 Dodger games (24 each in 1962, 1963, 1964, 1965) which I recorded during those seasons. These data (in Table 7) show that the fewest pitches per appearance occurred when Wills was on first, but did not steal. This observation probably reflects the large number of fast balls in this situation as discussed above.
Further analysis reveals, perhaps surprisingly, that walks were much more frequent following a steal, causing on base percentage to be depressed only slightly in comparison to the batting average decline following a steal. The pitch data show a dramatic increase in the number of pitches per plate appearance for those following Wills after a steal of second. This can be most easily explained by noting that these batters (usually Gilliam) took an additional one or two pitches (including pitchouts) while waiting for Wills to steal. The consequences are that walks are more likely and that the average number of strikes is increased. Pete Palmer has shown that batting with two strikes leads to a greatly reduced average.
The final aspect of Wills' base running to be considered is the effect he had on mistakes by the opposition. Table 8 summarizes the number of bases Wills advanced via balks, wild pitches, passed balls, errors on pickoff attempts, and errors following stolen bases. For the three seasons analyzed, Wills accounted for over 20% of all such extra bases the Dodgers received, even though he represented less that 13% of total Dodger base runners. It is clear that his presence on the bases increased the chance of the opposition making these mistakes. This feature and the batting performance discussed in the previous paragraphs are examples of what are often referred to as "intangibles." Complete understanding is not possible, but it is encouraging that play-by-play analysis can give us quantitative evaluations of points which are often in the realm of guesswork and intuition.
Table 8. Bases advanced by Wills on opposition mistakes.
Wills' base running was very important to the Dodgers of 1962, 1963, and 1965. Pete Palmer has shown that in a given season only about 10 batters in the whole league contribute as many as three wins. Wills' running was directly responsible for 7 wins over the three seasons, 3 of them in 1962. In addition his overall contribution was increased by factors related to his base running, namely his batting ability and effect on the opposition's concentration. Wills was highly successful in his running attempts. If his PO are omitted, then his success frequency was .791. Even if POs are considered to be CS, then he was successful .739 of the time. The major league average is usually around .60. Leonard Koppett, writing in The Sporting News (Feb. 10, 1979), has observed that Wills caused a significant change in the way major league baseball is played. He was one of the first men to be allowed to run whenever he wished, and the opposition knew it. The analysis reported here demonstrates that through direct and indirect effects, tangible and intangible, the stolen base is a valuable play, as exemplified by the record of Maury Wills.