Saturday, February 28, 2015


There has been much ado--mostly in the form of noise--about the fact that assigning wins to individual pitchers is a flawed process. Much of this ado (and much of it is, in fact, about nothing) is tied to an ideology that has developed from the value measurement systems that are being forced down the public's throat by emboldened cultists wishing to actualize the work of Bill James (who would merely make a corrective notation in the historical record) by actively rewriting the official stats to suit their own desires.

Step right up for those Radioactive Tango Love Pies™!!!
Call us curmudgeonly (and believe us when we say that we've been called much worse...), but we think that it might make sense to quantify the extent of the flaw before mounting a mouth-foam crusade to toss away the historical record. In the rush to judgment and the desire to own a mandate to interpret history, these folks (as usual, a number of them aligned with the purveyors of the Tango Love Pie™) have proven to be eerie precursors of the current United States Congress, a large faction of which hope to hijack history as well as the government. They both share the same strange obsession: to unilaterally declare something utterly irrelevant and bankrupt, only to follow by attempting to replace it with something that is, in fact, far worse than what they were criticizing in the first place.

The first step to a semblance of sanity with respect to the assignment of pitcher wins is to actually anatomize the flaws and then determine the rate of their occurrence. As is so typical of the "neo-post-neo" faction now searching for market inefficiencies in the degree of tension in an athlete's jock strap, the analysts favor "big data" without actually synthesizing any of it.

We see, for example, that a number of scoring quirks existed early in the twentieth century that assigned a handful of wins to a starter who hadn't pitched five innings. And we see a few stray instances of inconsistent judgment calls by official scorers. When we add this up over the long history of the game, however, we see that these types of glitches account for 0.5% of the total games played.

Viva la revolucion, n'est-ce pas??

But there is an area where wins are assigned with a through-the-looking-glass quality. These occur directly in conjunction with inefficient relief pitchers who surrender a lead and receive a win when their team retakes the lead while the pitcher who was lousy is still the pitcher of record.

This occurrence is not accurately quantified in any official way thus far; there are only surrogates for it that do not directly address the actual frequency. Over at Forman et fils, they list the number of times the starting pitcher is "in line for a win" but the team goes on to lose--an interesting stat in its own right and one that might assist in understanding the efficiency of a team's bullpen, but one that doesn't get to the root of our issue at all since we are looking for "wins while pitching badly that were stolen from someone who pitched better."

There is one statistic that can get us to where we want to go, however. That stat is the "blown save."

It's a stat that is overlooked, even scorned, due to two factors: 1) its lack of historical pedigree and 2) due to its odd lack of precision that creates "save situations" in the sixth, seventh and eighth innings. But it's precisely that lack of precision that affords us insight into the situations where what we've characterized (back up top in the title...) as "uggly relief wins."

In other words, wins as a result of blown saves.

So--how many of these are there? In 2014, there were a total of 59 "blown save wins"--wins awarded to relievers pitching badly enough to relinquish a lead, and then benefit from a go-ahead rally while they were still pitchers of record.

That works out to 2.4% of all wins for the 2014 season.

So what that means is that the cause celebre, this blight of all blights, constituting--apparently--the sellout of truth, justice and (God help us...) the American way, is focusing on a method that is perfectly reliable upwards of 95% of the time.

Reassigning 59 wins from relievers who've found this annoying little loophole seems a lot more reasonable than developing overwrought automated systems that reassign up to ten times as many wins in any given season.

Perhaps we need more history and more context? Would it be valuable to know if the percentage of "uggly relief wins" has changed over time? And maybe useful to have a sense of how the changes in the usage of the bullpen may (or may not) be affecting blown saves/blown save wins and this purported "conceptual crisis of the win"?

Well, of course we do. And Forman et fils is the place to acquire it. We spent some time (when we had it--as you may have noticed, we're not here a lot at this time because we have many, many other things on our plate...) looking at the data. And we've compiled a summary chart that gets us to the root of the matter.

The chart (above) starts with the actual number of blown saves in a season. We've condensed the data to reflect how the essential pattern has evolved. Accelerated bullpen use took hold in the mid-to-late 80s and was exacerbated by the offensive explosion in the 90s: we reach a peak at the very end of that decade. Things have declined a bit since, but seem to have plateaued.

Blown Save Wins (BlSvW) have also descended since the late 90s, and the percentage of "uggly relief wins" (UGG%) has declined back to pre-offensive explosion levels.

One of the other things that we thought might be significant here was to see if the ever-increasing use of relievers was having an effect on "uggly relief wins," particularly in terms of how long a reliever pitches in these. What's striking here is that from 1949 to the present, the vast majority of "uggly relief wins" come from pitchers who pitch at least one full inning (89% of them, in fact) while the percentage of blown saves that are an inning or more in duration is consistently between 50% and 60%.

That means that there's something structural about how "uggly relief wins" manifest themselves that resists any perturbing effect by the increasing number of relief appearances being made.

And interestingly, the percentage of blown save wins (same thing as "uggly relief wins," just in case that wasn't clear) in appearances equal or greater to a full IP is dropping. (That can be seen in the far right column, the one marked B!Sv1+% (yes, that "!" was supposed to be an "l"...fat fingers uber alles!) One wonders that if offense continues to decline, whether "uggly relief wins" will continue to drift downward.

So should we worry about the "uggly relief win" and how it has ruined the use of pitcher wins? No, of course not. But we don't expect this finding to gain much traction with the ideologues, who would like to take away texture and shape and all of the "imprecision" in value assessment that they imply, and try to do so with the zeal of a score of possessed mothers compelled to throw out bathwater and baby. (Give us spots on our apples, and leave us the birds and bees, already...)

Finally,  here's an astonishing fact related to blown saves in general that just dropped out of the data collection effort. It turns out that, over the last twenty years at least, the blown save--this is in general, now, for all outings identified this way, not just those that become "uggly relief wins"--is accompanied by an exceptionally high stolen base rate. It's not uncommon for the success rate in steals during blown saves to be upwards of 85% (in 2013, there were 91 SB, 14 CS in blown saves, or 86% to the good).

What does that mean?? Hard to say. But it's strange, and interesting, and more worthy of some fuss and fol de rol than the so-called "crack in the earth" purportedly produced by the fact that pitcher wins are not a perfect laboratory product.