Wednesday, May 21, 2014

THE STRAIGHT DOPE ABOUT STARTING PITCHERS & TIMES THRU THE BATTING ORDER

We've been hearing a lot about the "crisis" in starting pitching over the past ten years, during the apogee of the neo-sabe onslaught that pushed its way into baseball's front offices at the turn of the millennium.

Starters ostensibly could not go around the batting order more than twice; they were simply getting killed as they began that third go-round (batters 19-27, or what Forman et fils breaks out as "3rd PA in G as SP").

The crisis was so great that even esteemed, veteran analysts such as Craig Wright were convinced that shorter appearances by starters and an enlarged corps of relief pitchers to close things out would be the only way to adjust for the ongoing offensive onslaught.

But the data behind such claims was always sketchy at best. And, in the "post-neo" age, which took hold just as offense began to decline toward the levels we've seen over the past 3-4 seasons, we're still skeptical that such a "crisis" actually existed. It's more likely that the hitting philosophy that drove the offensive explosion was simply countered by a new wave of starting pitchers who could up the ante on the "stand and crank" approach that had shattered home run records and produced high run scoring levels.

Here we present a few charts, based on the "times through the batting order " data for starters that recently became readily available in the Play Index at Forman et fils. The data represented in these charts will clarify the claims of a "crisis" in starting pitching as related to the "times through the batting order" formulation.

First, here's the chart that shows the relative percentage of plate appearances (PAs) that starting pitchers face in the 2nd, 3rd, and 4th time through the batting order as compared with the first time.

Looking here, you can see that there is indeed a "crisis"--but it's to be found in the 4th time through the order, not in the 3rd. Since the mid-70s, the percentage of PAs the fourth time around has dropped from 40% of the batters faced in the first time around to around 6%.

The fourth time around, in case you need some numerical orientation, accounts for batters 28-36 in any given game.

As you can see, this goes hand in hand with the precipitous decline in complete games that occurred with a very similar descent line from 1974 to the present.

How about performance? Is there a significant degradation in performance when we get to the later times through the batting order? Are we in the middle of a crisis?

Looking at the data, and risking a certain amount of derision by using ERA as the performance measure, we see that there was a crisis all right--but it happened in the late nineties.

ERA the third time around in the batting order ran between 90-99% of ERA during the first time around until the mid-nineties, when it took a big drop, down to around 85%. It's since righted itself, and has been consistently around 90% ever since.

So something did change, but the trend did not continue. We are not in the midst of a performance crisis in the third time through the batting order.

What is astonishing, however, is the data for the 4th time around. It actually shows us where the crisis in starting pitchers relative to complete games occurred, and why/how expanded relief innings came into play.

You can see it in the the mid-80s, where there's a downturn in relative ERA in the 4th time around that bottoms out in 1987. At this point, relief innings as a percentage of total innings begin to rise, and the removal of starters who struggle in the 4th time around produces a "survivor bias" where those who do make it into batters 28-36 actually exceed the overall performance levels for batters 1-9.

That "survivor bias" also took a hit in the same time frame that the ERA data for the 3rd time around hit its bump in the road. As Dick Enberg would say: oh, my.

But then the chart takes a surprising turn, with "survivor bias" for the 4th time around gaining momentum again (around 2004), and then suddenly jumping through the roof in the past three seasons.

And the raw data for ERA (table at right) supports this, by showing that 4th-time around survivors have been beating the overall 1st time around performance every year since 2007, and doing so quite handily over the last three seasons (including our partial current season in 2014).

Such data puts the lie to the idea that starters should invariably get pulled in the eighth or ninth innings because they are much more likely to give up runs than a relief pitcher. It's clear that, in general, managers and pitching coaches have a good sense of which starters can be permitted to continue deep into the game (i.e. the 4th time around), and the results in the above table bear that out.

Let's finish with a performance component that many believe is directly correlated to the changes in offensive levels over the past few seasons--strikeouts. Here we have a chart that shows the K/9 averages (again, relative to the first time through the batting order) for the 2nd, 3rd, 4th time around.

This chart might be said to show us how much of a starter's "stuff" remains as he goes through the game.

And this chart shows that as power pitching has come to the fore (beginning, in fact, around 1960 and slowly building over time to the astonishing crescendo that we're witnessing today), the ability of pitchers to hold their K/9 levels in subsequent times through the batting order has been increasing.

Now, the figures are not dramatic. As we can see, there was a lull in the late 70s (you might call it the last stand of the finesse pitcher...) but since then, the gains in K/9 rate retention in later times through the order have been consistent.

There is certainly no existence of a dramatic downturn in performance in later times through the order here, no matter how you measure it. The "crisis" referred to is one that occurred ten to fifteen years ago and is no longer a feature of the data.

(Now the "injury crisis" is another matter. That we'll deal with a bit later on...)