Talking Sabermetrics: Postscripts on Astros' Veras and Humber Acquisitions

We have had several articles about the Astros' waiver acquisition/contract signing of Philip Humber and the free agent signing of Jose Veras. I don't want to be repetitious of the evaluations in these articles. But a couple of recent sabermetric articles may give us a glimpse of the Astros' thinking behind the two pitching acquisitions. So, here are a couple of quick points about each player's 2012 season.

Humber and kwERA

Glen DuPaul at Beyond the Boxscore wrote about the 2012 starting pitchers who are likely to improve or decline based on the pitching metric "kwERA." This metric was devised as a quick measurement for pitching performance, but DuPaul has previously found that kwERA is surprisingly effective at predicting starting pitchers' future performance over the remainder of a season or for the next season. The kwERA metric gives equal weight to strike out % and walk %, and is scaled to ERA. This seems similar to FIP and x-FIP, except that it ignores HRs. DuPaul's previous Hardball Times article suggested that kwERA is better than FIP and x-FIP at predicting future performance, and concluded that simple is better in this case.

Based on the gap between ERA and kwERA in 2012, DuPaul shows Humber as no. 2 on his list of starting pitchers most likely to improve in 2013. Humber's kwERA gap is 2.03 runs and is second only to Jake Arrieta of the Orioles. Humber's kwERA was 4.41 in 2012. This may not seem like an impressive "target" for regression. But keep in mind that the average American League's starter ERA was 4.37 in 2012. Humber had a bad year, but he pitched like an average starter in the AL, according to kwERA. If you consider the fact that Humber pitched in the higher run environment of the AL, it's possible that he pitched better than several of the pitchers in the Astros' rotation. It's also worth noting that Humber's kwERA is almost the same as his 2012 SIERA (4.48), which is also a predictive pitching stat. Obviously, there are no guarantees here, but it seems like a worthwhile gamble.

The Astros' pick up of Humber is a similar gamble to the Pirates' signing of Francisco Liriano, who also is on DuPaul's list of pitchers most likely to improve in 2013 based on kwERA. The Liriano signing inspired Fangraphs to write about the "slow death of ERA" as a front office evaluation method.

What is the kwERA for the Astros' pitchers with at least 60 IP? Here it is, courtesy of the rotovalue web site.

Astros kwERA

Lopez (traded) 3.24

F. Rodriguez 3.42

Veras (signed) 3.84

Norris 3.87

Lyles 4.32

Humber (signed) 4.41

Harrell 4.44

Keuchel 5.43

It's hard to say why kwERA may be better than FIP and x-FIP in predictive ability. We usually think that more information (such as the inclusion of HR rates) is better. But that may not be the case, particularly for short term periods. One possible reason: HR/Flyball and HR/9 are relatively volatile stats. Yes, both HR statistics have some year-to-year correlation, but the correlation on a year over year basis is 40% - 50% less than the year to year correlation of K% and BB%.

Jose Veras

My last Talking Sabermetrics article discussed DuPaul's findings that K rate is the best predictor of relief pitcher performance, with particular reference to the prospect of a F-Rod rebound. Only a few days later, the Astros signed Jose Veras, who brings a 2012 K rate of 26.3% to the table, highest among current Astros' relievers. As you may recall, one of the reasons for relying on K rate in evaluating relief pitchers is the fact that K rates stabilize in a smaller sample size than most statistics. Relief pitcher results tend to be volatile from year to year, mostly due to typically small sample sizes.We can speculate that the Astros' evaluation of Veras is consistent with heavy reliance upon K rate for rating relief pitchers.

Here at TCB, we had some discussion of Veras' walk rate weakness and his potential for high leverage pitching. Around the same time, I found BtB writer Bill Petti's spreadsheet of pitching statistic correlations. I came up with a couple of correlation nuggets which might be relevant to Veras.

First, year to year K rate correlation is 20% higher than BB rate correlation. Maybe Veras' K rate, his positive attribute, is more certain than his BB rate, his weakness. Perhaps there is a little bit of hope for some improvement in his BB rate.

Second, K rate is highly correlated (.47) with a player's future Wins Probability Added (WPA), while BB rate has a lessor correlation (-.07) with WPA. WPA is moderately correlated with next year WPA (.33). Granted, these correlations are for starting pitchers, so I am making the leap that a similar relationship exists for relief pitchers. Change in WPA is the input for calculating reliever shutdowns and meltdowns; therefore, WPA is particularly relevant to relievers. And it makes sense that pitchers with high K rates can get out of jams more readily, thereby leading to the relationship between K rates and WPA. For what it's worth, Veras had a high WPA last year (1.31) and has posted positive WPA for all but one year in his career. Take these as interesting observations, rather than predictions.