Last season the Brewers shattered the major-league club record by striking out 1,399 times, eclipsing the mark of 1,268 set by the 1996 Tigers. As power-hitting has come to dominate the game, increasingly more players are swinging for the fences with whip-handled bats that generate tremendous speed over the plate. Coupled with changes to the strike zone in 2001, the result for the Brewers was lots of home runs and lots of strikeouts.
Consider the all-time list in strikeouts as a percentage of batting outs (defined as at-bats – hits + sacrifice hits + sacrifice flies), which is dominated at the top by recent teams:
Year Team SO% --------------- 2001 Mil 33.23 2001 SD 30.45 2000 StL 30.35 1994 Det 30.18 1996 Det 30.09 1998 Ari 29.43 2000 Mil 28.91 1998 Chi 28.72 1999 Pit 28.62 2000 Fla 28.45
Despite one third of their batting outs being strikeouts, the Brewers outscored five National League clubs, all of which struck out hundreds fewer times. In addition, the Padres, who with 1,273 strikeouts also surpassed the 1996 Tigers, finished sixth in the league in runs. The notion that strikeouts are worse than other batting outs is ingrained in baseball lore. The rationale is that a team that strikes out a lot suffers by failing to put the ball in play, thereby reducing opportunities to advance base runners with outs and to force errors by the fielders. Using a number of approaches, this theory was put to the test using the statistics from the 1,122 team seasons since 1955.
The first test was to calculate the correlation between SO% and runs per 27 outs (calculated as at-bats / hits + sacrifice hits + sacrifice flies + caught stealing + grounded into double plays).
Correlation measures the association of two variables on a scale of -1.00 to 1.00. A negative number indicates a relationship where one variable rises as the other one falls, while a positive number indicates a relationship where both variables rise and fall together. SO% and R/27 were used rather than raw strikeouts and runs in order to adjust for unequal opportunities among the teams. Also, strikeouts as a percentage of batting outs has the added benefit of inversely expressing what percentage of its batting outs a team put in play. The correlation revealed a relationship between SO% and R/27 of 0.19, a weak positive relationship.
That SO% and R/27 move together should come as no surprise. SO% also shares a weak positive relationship of 0.25 with slugging average, and slugging average shares an extremely strong positive relationship of 0.93 with R/27. Hitting for power positively relates to striking out, but it also positively relates to scoring. The correlation yielded nothing to indicate that teams that strike out often score fewer runs than teams that frequently put the ball in play on outs.
The second test involved a regression to determine whether strikeouts are more costly than other batting outs in creating runs. A regression analyzes how a set of independent variables such as hits, walks and outs contribute to a dependent variable such as runs. A regression using singles, doubles, triples, home runs, unintentional walks plus hit by pitch, intentional walks, stolen bases, caught stealing, strikeouts and other batting outs (calculated as at-bats – hits – strikeouts + sacrifice hits + sacrifice flies) resulted in a value of -0.101 runs for strikeouts and -0.095 runs for other batting outs. In other words, strikeouts were 0.007 runs more costly than other batting outs.
A separate regression separating grounded into double plays from other batting outs yielded slightly different results: strikeouts cost -0.105 runs, other batting outs cost -0.093 runs and GDPs cost -0.522 runs. This widened the gap between strikeouts and other batting outs to 0.013 runs but also revealed the cost of putting the ball in play when a double play results.
At an additional cost of 0.013 runs, the 1,399 strikeouts by the Brewers reduced their scoring by about 17.6 runs compared with making other batting outs. The Brewers also grounded into just 102 double plays, fewest in the National League. Not including the Brewers, the average National League team struck out 1,101 times and grounded into 122 double plays. Had the Brewers matched those averages, they would have scored about 3.7 additional runs by striking out 298 fewer times but cost themselves about 10.2 more runs by grounding into 20 extra double plays.
The threat of a GDP is the offset to the benefit of putting the ball in play on outs. A double play not only produces an extra out, it also erases a base runner. A correlation demonstrated a weak negative relationship of -0.21 between SO% and GDPs as a percentage of base runners (calculated as hits + walks + hit by pitch – home runs – caught stealing). Teams that strike out a lot might not advance runners on outs very often, but they also burn base runners on ground balls less frequently. GDP% also bore a weak negative relationship of -0.20 to R/27, the mirror opposite of the positive relationship of SO% to R/27.
The third test relied on the Tech-1 version of Bill James’ runs created formula, first published in the 1985 Baseball Abstract, which incorporates hits, unintentional walks, intentional walks, hit by pitch, total bases, stolen bases, caught stealing, sacrifice hits, sacrifice flies, GDP and at-bats but not strikeouts. (James has since created a version of the formula that includes a slight penalty for strikeouts.) The purpose of this test was to measure the difference between every team’s runs created and its actual runs scored. If strikeouts reduce a team’s scoring efficiency, presumably the difference between runs and runs created would reflect this since the runs created Tech-1 version excludes strikeouts.
For all teams since 1955, runs created was very accurate, differing from runs by an average of 18 runs, or 2.7 percent. The correlation between SO% and the percentage difference between runs created and runs yielded an almost non-existent relationship of 0.02. At least two explanations would account for the lack of a relationship. Either SO% has no perceptible effect on scoring, or the runs created formula is insufficiently precise to measure the difference. The impact of SO% on scoring might fall below the 2.7-percent margin of error in runs created.
The fourth test also used runs created as well as sifting the teams with the lowest and highest strikeout totals from the rest of the population. There were 183 teams at least one standard deviation below and 195 teams at least one standard deviation above the set’s average SO%. If the theory holds, the teams with a low SO% should have tended to overperform according to their runs created, while the teams with a high SO% should have tended to underperform according to their runs created.
The low-SO% teams actually underperformed their runs created by an average of five runs, or 0.7 percent, but the high-SO% teams underperformed their runs created by a greater margin of an average of 12 runs, or 1.7 percent. The difference between the two groups was thus seven runs, or 1 percent, in favor of the low-SO% teams.
Further refinement to include only teams that were below or above the average SO% by two standard deviations yielded 19 low-SO% and 27 high-SO% teams. The extreme low-SO% teams overperformed their runs created by an average of four runs, or 0.7 percent. The extreme high-strikeout teams underperformed their runs created by an average of six runs, or 0.9 percent. The difference here, in small sample sizes, was an average, before rounding, of 11 runs, or 1.8 percent, in favor of the extreme low-SO% teams. Perhaps at these ends of the spectrum the detriment of striking out frequently was revealed.
Here is a comparison of the characteristics of the low-SO% and high-SO% teams:
1 StDev SO% R/27 GDP% ------------------------ Low SO% 17.07 4.29 7.19 High SO% 26.13 4.65 6.57
2 StDev SO% R/27 GDP% ------------------------ Low SO% 14.81 4.38 7.06 High SO% 28.65 4.71 6.48
The final test employed another Bill James creation, similarity scores. In the 1986 Baseball Abstract, James published an article examining whether the spectacular base running of the 1985 Cardinals, who stole 314 bases, had a significant impact on their scoring. Using similarity scores, James devised a list of 10 teams in baseball history most similar to the Cardinals in almost every aspect except stolen bases in order to observe whether and how much the Cardinals outscored those teams.
Similarity scores express the degree of likeness between two players or teams. A pair of identical teams will have a similarity score of 1,000. To calculate similarity scores for teams, start with 1,000 and subtract one point for the difference of every one game, 20 at-bats, one run, one hit, five doubles, 1.5 triples, 0.5 home runs, two RBI, 2.5 walks, 20 strikeouts, three stolen bases, 1/4 points of batting average, and 1/2 points of slugging average.
For the purposes of this test, runs and RBI were omitted from calculation of the similarity scores (since a difference in runs was the phenomenon to be observed), and the difference in strikeouts was added rather than subtracted from the similarity scores in order to identify teams with many fewer strikeouts than the Brewers. Also, hit by pitch were counted as walks, and one point was subtracted from the similarity scores for the difference of every 1/4 points of on-base percentage. The following are the 10 most similar teams to the Brewers:
Year Team Avg OBP Slg 2B 3B HR R BB HBP SO SB Sim ------------------------------------------------------------- 2001 LA .255 .323 .425 264 27 206 758 519 56 1062 89 950 1985 Det .253 .318 .424 254 45 202 729 526 27 926 75 948 1996 Det .256 .323 .420 257 21 204 783 546 29 1268 87 908 1964 Min .252 .322 .427 227 46 221 737 553 44 1019 46 895 1987 SF .260 .322 .427 274 32 205 783 511 39 1094 126 893 1987 Bal .258 .322 .418 219 20 211 729 524 22 939 69 893 1963 Min .255 .325 .430 223 35 225 767 547 41 912 32 877 1986 Min .261 .325 .428 257 39 196 741 501 37 977 81 875 1987 Min .261 .328 .430 258 35 196 786 523 38 898 113 870 1966 Atl .263 .326 .424 220 32 207 734 512 40 913 59 868 ------------------------------------------------------------- Average .257 .324 .426 245 33 207 760 526 37 1001 78 930 ------------------------------------------------------------- 2001 Mil .251 .319 .426 273 30 209 740 488 72 1399 66 1000
One problem with this approach is that no team is extremely similar (similarity score greater than 950) to the Brewers, and only three teams are very similar (similarity score greater than 900). On average, however, these teams have numbers comparable, or slightly superior, to those of the Brewers. The difference in walks is offset by the difference in hit by pitch. The similar teams do have a slightly higher OBP, but also grounded into an average of 12 more double plays. If GDPs are subtracted from baserunners, the OBP difference is cut in half. The Brewers struck out 39.8 percent more often and were outscored by 2.7 percent.
Perhaps the strikeouts made the difference of 20 runs in scoring. If so, this is a small negative given the huge difference in strikeouts. Indeed, the cost appears to be about one run per 20 strikeouts. This figure of 20 runs is greater than the 11-run difference estimated above between the extreme low-SO% and high-SO% teams above and the 3.7-run difference predicted by the regression between the Brewers and the average National League team. Overall, the data seem to indicate that the negative effects of strikeouts are minor at best.
Thus, while there may be reason to conclude that plentiful strikeouts are detrimental to a team’s offense, there is no basis for saying that reducing strikeouts is critical to a line-up’s success. Particularly if curbing strikeouts involves removing power from the batting order and causing more groundballs that result in GDPs, such a move appears likely to be harmful to a team.