Prospectus Hit List: R... (06/19)
39 comments have been left for this article. BP Comment Quick Links
| |
|
|
June 21, 2009 Prospectus Idol EntryBabe Ruth the PitcherOne of my favorite essays on any topic is Nate Silver's "Is Barry Bonds Better Than Babe Ruth?" from Baseball Between the Numbers. Rather than rehashing the same tired arguments about how much harder it was to hit a home run in Ruth's time or how much better the competition against Bonds was almost a century later, Silver uses a variety of metrics to demonstrate how the two players would actually perform on a level playing field. The two players are very close until Ruth's pitching is taken into account. While his hitting is examined in great detail, his pitching is only valued via NRA. NRA has nothing to do with rifles; it actually stands for Normalized Runs Allowed, which basically takes Ruth's run prevention skills from the Teens and tells us what they would look like in a more typical "all-time" Major League season, adjusting for his park and league along the way. An NRA looks like an ERA in a normalized environment. An average pitcher will have an NRA of about 4.50. While I don't find fault with any of Nate's conclusions, NRA does have some inherent issues that could make it a faulty metric when we want to see the quality of a pitcher's work. It adjusts the runs to more recent times, but doesn't take the player's peripheral stats into account. In other words, if a player had incredibly good luck piled on top of good defense, NRA would like him a lot more than it should. All of this got me thinking about Babe Ruth and how he would really fare against not just a level playing field, but also the players of today. In order to know how Ruth would fare we have to first become acquainted with Davenport Translations, or DTs. If you've ever walked into a sports bar, you probably heard some drunk guy shouting about how steroids are ruining the "statistical integrity" of the game. Minding your own business, you knew with certainty that the game never had the type of statistical integrity the fellow was shouting about. A home run in 1920 was different than one in 1960, which in turn is different from one hit today. DTs account for just that. They convert the numbers for all players in all time periods to tell us how they would perform in a neutral environment. There is a DT for ERA, but it is essentially very similar to NRA-it doesn't account for the pitcher's peripheral numbers. Fortunately, those peripheral numbers are also included in the DTs. K/9, BB/9, and HR/9 are the central stats that the pitcher has control over. They are also the stats we need to look at to find out how Ruth would have pitched today. We'll concern ourselves with the years 1915 through 1919 when Ruth did the vast majority of his pitching. He had 3.70 K/9, 3.22 BB/9, and 0.09 HR/9. Anyone who's looked at a box score of a game that Mike Pelfrey has pitched knows a lot about these numbers. Having a similar number of walks and strikeouts is bad, but on days when the pitcher in question doesn't allow any home runs, they look like some kind of all right. Ruth didn't allow homers because of the era he pitched in, but a look at the Davenport Translations of these numbers from Ruth's DT page tells us that he wasn't as bad of a pitcher as the Ks and BBs seem to suggest. His average DT rates for the five years we are concerned with were as follows: 5.56 K/9, 3.46 BB/9, and 0.99 HR/9. Of course, the homers shoot up because the neutral environment reflects how much power has gone up in the years since Ruth pitched, and strikeouts come up as well. Ruth pitched at a time when pitchers didn't get high K rates. One reason is that they were expected to throw hundreds and hundreds of innings. For example, the Babe threw 317 in 1917. We couldn't expect him to keep up a high K rate any more than we might expect Mariano Rivera to keep up his if he had to throw 6 innings every time he came out. The DT rates exist on a level playing field. It is a sort of imaginary "all-time" season where the average pitcher has rates of 6.00 K/9, 3.00 BB/9, 1.00 HR/9, and a 4.50 ERA. We can use these numbers to compare whichever pitcher we're looking at. In this case, Ruth has walk and K rates that aren't bad but still aren't as good as an average pitcher's rate. Although the average DT rates represent those of a typical environment, I'm not sure where we can find a real season exactly like it. Nevertheless, rates in the last ten years of Major League Baseball aren't entirely dissimilar: 6.40 K/9, 2.91 BB/9, 1.04 HR/9 and a 4.17 ERA (I am only focusing on starting pitchers who logged at least 150 innings here since their job is more similar to Ruth's). Clearly, the numbers aren't that far off, except maybe the ERA, which doesn't concern us since we'll figure out our own from the peripheral stats. The first thing we need to do is convert Ruth's numbers from the neutral setting to the contemporary setting, i.e. the last ten years. For the mathematically interested: the ideal way to do this is through the use of standard deviation. But in order to do that we need a season very similar to the neutral setting, and I don't have one on hand. That said, the neutral stats are similar enough to the actual stats that the quick and dirty method of "splitting the difference" will suffice and give us numbers that aren't far off. For example, we saw that the average K rate changed from 6.00 in the neutral DT environment to 6.40 in the last ten years. We can assume that hitters strike out a bit more often these days (see Howard, Ryan), but pitchers are also probably more talented and Ruth's level of talent would remain the same. So we'll split the difference and bump up his K rate by 0.20, half of the full 0.40. Ruth gets an altered line of 5.76 K/9, 3.41 BB/9, and 1.01 HR/9. These numbers aren't sexy and the Babe certainly wouldn't be confused with a Cy Young candidate, but pitchers like Jeff Suppan and Miguel Batista have made careers out of lines like this, and it is certainly better than replacement level. Let's take an even closer look to see what kind of ERA the Sultan of Stock Pitching would have in the years 1999 to 2008. Our typical ERA is 4.17 so we'll start there. Last week, I went over the value of preventing home runs. Ruth's home run rate is not only estimated without the benefit of groundball rates, but it's also so close to average that it's not going to make much difference, in terms of accuracy, to apply it here. However, in the spirit of being thorough while assessing a pitcher's ERA, we'll apply his home run rate and find that he comes down to 4.13. The methodology I used to find how much a good HR/9 lowers an ERA may also be used to find how much it is affected by Ks and walks. Using the pitchers in the last ten years that we are comparing Ruth against, each K/9 below or above average tended to respectively increase or decrease an ERA by 0.17. For instance, if Ruth had 5.40 K/9, exactly one less the average of 6.40, we'd raise his ERA by 0.17. As it stands, he has 0.64 K/9 less than the typical starting pitcher so his ERA is raised by 0.11 to 4.24. Each BB/9 will tend to increase an ERA by 0.29. Ruth would walk 0.50 hitters more than average according to our conversion, so he gets a final lift of 0.15. Ruth's ERA stands at 4.39. The Bambino might have been what we like to call an innings eater: he eats up innings and vomits runs, but at least he makes it through enough of them to give the bullpen a rest. Of course, we looked at all Ruth's pitching years as a whole. Let's see how he would have done as a contemporary pitcher in each of the individual seasons he pitched: Year K/9 BB/9 HR/9 ERA 1915 6.41 3.50 1.20 4.56 1916 7.10 3.34 0.74 3.73 1917 5.91 3.17 1.03 4.30 1918 4.01 3.16 1.17 4.82 1919 3.56 4.33 1.04 5.05 Looking at him on a year-to-year basis, a clear pattern comes into focus. Ruth had a nice year in 1916, but the following years don't reflect a positive trend. Even with a healthy conversion rate, his strikeout rate dips into the Carlos Silva danger zone. Meanwhile, the walks and home runs don't do him any favors either. Earlier, I mentioned Ruth's NRA and DT ERA and how they convert his actual ERA, while ignoring peripheral stats. Here, we find that the above ERAs, based on his translated peripherals, are higher than the two former metrics suggested, even despite the lower baseline ERA in the last ten years. The fact that Ruth's translated earned run averages come up better than they should indicates that Ruth's ERA was better than it should have been in the years he actually pitched. In other words, his pitching was somewhat overrated even in his own time because of good luck and defense. Ruth famously quit pitching so he could hit every day, but his would not be a memorable pitching career whether or not it had continued. If he pitched like this in the twenty-first century, he'd have a job, but by his fifth year, he'd be borderline and would probably be looking at the end of a shorter career than he might have liked. In his day, he would have had a longer pitching career that stretched into the twenties, but he wouldn't be a very notable figure in sports history. Good thing he could hit a little bit too.
|
I really like how easily Brian broke down some really heavy statistical concepts into chunks. It took him a while to explain NRA, but by doing so slowly, methodically, and (!!) entertainingly, I was able to not only follow along, but barely notice that I was absorbing this. He gets a bit too dense for me (and yes, I'm dense on subjects like this) in the middle, but he couldn't keep going as slow as he had without ending up about 5,000 words. There's some wobble from the strong start but he ends with a strong conclusion that isn't just strong, but supported well.