So it is that a good starting point, given the history most of us share in the game, is to note that professional ballplayers are not Little Leaguers (and when they were, they never had to worry about choking up or not striking anyone out). With a team of 10-year olds carpooling to practice, the primary issues are advancing the game in a timely manner and that everyone gets a good amount of participation. That means the key focus on coaching pitchers is making sure they throw the ball over the plate and not that they miss bats. When they do find the plate, you want the kids to hit it. When both sides are doing this, you've got a good game.

Unfortunately for many of us, this is where most of our coaching in the game ends. In order to separate the rules of value we played under from those in the pros, we have to first have a grasp on where value comes from at the top levels of the game. In doing so, we break players into two distinct categories and look at their values independently. Obviously, these categories are pitchers and non-pitchers (or hitters, as some like to call them, though that is not the whole of their value).

We judge pitchers by how good they are at run prevention. We judge non-pitchers by a combination of run prevention and run creation. From here, we can begin to see where our valuations of both categories differ. The first thing you should notice is that the task of run creation is left entirely to the group of non-pitchers (essentially), but that run prevention is split up between pitchers and non-pitchers (as fielders). The entire spectrum of events on the offensive side of the game is the charge of the hitter, whereas the pitcher's responsibility is bunched toward the fielding independent side of the spectrum.

This means two things. One, since the breadth of pitcher responsibility is spread disproportionately over a tighter range on the spectrum, those areas (the fielding independent stats such as K, BB, and HR) each hold more weight relative to the rest of the spectrum. Two, non-hitters share the responsibility for some of the areas of run prevention as well as controlling all of run creation. This means that there are more areas where non-pitchers can pick up value, and no one area is as essential to their value. They don't even necessarily have to be more than passable hitters if they provide enough value on defense.

Adam Dunn can strike out 200 times in a season and still be a great hitter because there are so many other ways he can add value at the plate. Adam Everett can fail to do anything at all to make him a valuable hitter and still be valuable because of his glove. Adam Wainwright needs to be good at some combination of striking hitters out, limiting walks, and limiting home runs in order to be valuable. If he's not good at those things, there's not enough places left for him to make it up. This is the first reason that strikeouts are more important for pitchers than for non-pitchers.

Of course, preventing walks and home runs is also important. In fact, the concensus among statisticians is that both a walk and a home run have a greater effect on run scoring than a strikeout, and a home run substantially moreso. Run estimators that assign weights to different events reflect this by giving strikeouts less weight than HR and walks. Why all the focus on strikeouts, then?

We can look specifically at a run estimator to see how strikeouts affect value differently from other events as well as how they affect value for pitchers differently than for hitters. For this article, I'll look at Jim Albert's formula using four component rates. The formula is:

-3.2 + 13.2*BB% - 12.3*K% + 40.9*HR% + 24.5*BABIP

I chose this formula over other run estimators for a number of reasons that suit it to our purposes:

-It is easily applied to both pitchers and hitters.

-It breaks down value into the three most important fielding independent events and one rate that lumps fielding influenced value into one category, so we can focus on the fielding independent events.

-It uses strikeouts as one of its components.

-Each successive rate in the formula excludes all previous events from its denominator (K% does not include walks, HR% does not include walks or strikeouts, BABIP includes none of the other events). This means that each rate is reasonably independent of the others. HR/PA or HR/9 will hide some of the value of strikeouts in the HR component because allowing contact on fewer PAs will mean fewer home runs. Each of those rates is really a combination of HR/contact PA (the HR% in Albert's formula) and the amount of contact PAs allowed per AB (K%, essentially). Since we are concerned with the effect of strikeouts, using a formula that isolates strikeout rate as much as possible from other components is preferrable.

As we can see from the formula, home runs are by far the most important component. They're so important, in fact, that it is virtually impossible to make the Majors and stay there for any length of time if you can't keep home runs below a certain level. Anyone who serves up gopher balls at a BP-like rate is going to be weeded out pretty quickly. So there is an upper limit to HR% in the Majors for pitchers. Over a large enough sample, nearly every pitcher who makes the Majors will fall below that limit. There is also a lower limit to HR% (obviously 0 is a limit, but the actual limit is above that because big league hitters are good enough that they will inevitably leave the yard occasionally if they can put the ball in play). The upper limit is close enough to the lower limit that there isn't that much room for pitchers to rack up value.

Walks have a bigger spread than home runs, but there is still only so much room below the average BB% for pitchers to pick up value. Strikeouts, on the other hand, have a larger spread and have a ton of room above average where elite pitchers can gain value over their peers. The following graph shows the spread of each component of the above formula using all pitchers with at least 200 IP since 1993 (845 pitchers). The coloured box represents the spread from the 25th to 75th percentile with a line in the middle for the median. The whiskers above and below each box are the spread up to 1.5 times the spread of the box. Circles represent outliers that fall outside that range.

Strikeouts not only show a greater spread between the second and third quartiles, they also show the most outliers (particularly in the positive value direction-above for strikeouts and below for walks and home runs) as well as outliers the furthest away from the spread. Elite value is easier to come by in the strikeout category. That is, the best pitchers in K% gain more value from being good at striking out hitters than the best pitchers in BB% or HR% gain from being good at preventing those events. By applying the weights from the formula to the spread, we can see how much each spread is worth in run value.

25-75 | SD | |

BABIP | 0.51 | 0.39 |

BB% | 0.36 | 0.26 |

HR% | 0.45 | 0.33 |

K% | -0.75 | -0.58 |

This means that the difference in a pitcher at the 25th percentile in K% and a pitcher at the 75th percentile is .75 runs per game. The difference in pitchers 1 standard deviation apart in K% is .58 runs per game. As you can see, this is a significantly larger difference than the other components. Expanding the spread to include the extreme reaches of each component will further exaggerate the difference because of the additional spread outside the box for strikeouts, so you can see how important strikeouts can be to the value of an elite pitcher. Looking at the extreme values does introduce some problems if we try to use them to quantify the spread, as many of the outliers are relievers pitching under circumstances that inflate K%, but we can still see qualitatively how the extremes are pushed further in K% than in any other component.

Let's look at the same graph for hitters (minimum 800 PAs since 1993, 842 hitters in sample):

and the values of the spreads (NOTE: These numbers are not on the same scale as the pitcher values. They are scaled to fit runs per game, so this is hitter value over roughly 40 PA, which is obviously not comparable to the playing time for a pitcher whose value is scaled to runs per game. That means you can't compare the actual values between the two tables, only how they relate to the other values in the same table.):

25-75 | SD | |

BABIP | 0.71 | 0.54 |

BB% | 0.48 | 0.37 |

HR% | 1.23 | 0.90 |

K% | -0.92 | -0.70 |

HR% shows the largest spread value for hitters. BB% climbs a bit to just over half the spread value of K%. The fielding dependent component, which is more important for hitters since we charge them with all of run creation, also shows a relatively higher spread value compared to K%. We also see significantly more outliers in HR% and BB%, especially in the positive value direction (above the spread), fewer outliers in general in K%, and no outliers in the positive value direction (below) for K%. Elite hitters can pick up a lot of extra value by excelling in components besides strikeouts, and, unlike pitchers, hitters who don't strike out a lot don't gain all that much value compared to those who excel in other components.

Not coincidentally, hitters with poor strikeout rates tend to make up for that with good home run rates. Hitters who lose a lot of value in strikeouts can still be valuable hitters because it is possible to make up more than that value in HR%. For hitters, K% correlates fairly strongly to HR% (r=.59 over the sample). This is not true for pitchers (r=.04). Much of this is do to a selection bias (hitters who strike out a lot are only allowed to play if they can make up that value in hitting for power, whereas pitchers are already limited to a much tighter range of home run rates), but there is something relevant to take from this. Hitters can improve their HR% by deliberately swinging for power, usually at some cost to their K%. In other words, there is often a positive trade off to changing their approach to increase strikeouts. This is not true for pitchers. A pitcher who changes his approach to reduce strikeouts will generally not see a corresponding drop in HR%. For pitchers, K% is either good or bad with little effect on HR%. For hitters, the two are somewhat linked, so that changes in strikeouts for hitters are dampened by a corresponding change in power.

None of this is to say that strikeouts are the only way to measure value for pitchers, nor that they mean nothing for hitters. The proper way to consider strikeouts is in conjunction with other components and weighted appropriately. Once we have the proper weights for each event, we don't need to give special consideration to strikeouts or anything else, so nothing in this article should lead you to quantify pitching success differently from how you normally would (assuming you are doing it properly to begin with); rather, the point is to illustrate why the way we analyze pitchers and non-pitchers gives more significance to the strikeout in pitcher success than in non-pitcher success. Non-pitchers simply have more places to pick up value that render strikeouts less essential. They can pick up more value than pitchers in home run rates or on the value of balls in play (the fielding dependent events involved in run creation/prevention). They have fielding value completely unrelated to hitting. What's more, one's approach at the plate links home runs and strike outs in ways that a pitcher's approach can't so that the cost of strikeouts can be offset by gains in power.

When your team signs a Tim Lincecum or Javy Vazquez, take their value for what it is. Enjoy that they can mow through lineups with minimal contact allowed for the other team to work with rather than worrying about them buying into whatever notions about pitching to contact and using the defense you might have. When they sign a Mark Reynolds or Adam Dunn, instead of getting hung up on the frustration of every plate appearance wasted without making the fielders work, focus on where else they provide value. Look for the damage they do when they do make contact or the advantages they win when they work a walk. The results will be there if you know how to look.

## 0 comments:

## Post a Comment