Of all the things that didn’t go right for the 2017 Pittsburgh Pirates, their utilization of a young pitching staff was not one of them.
While the Pittsburgh Pirates staff had its hiccups — an ineffective Tyler Glasnow, or a generally tough start to the season from Chad Kuhl and Trevor Williams, for example — those sorts of issues are to be expected from a starting rotation where only 2 of the 7 different pitchers to start this season had an entire Major League season under their belt.
Now that these pitchers have their first full year nearly complete, we can look back on their season numbers and see which pitchers posted legitimate performances and which were impacted by other circumstances.
To do this we first need to build a model around Earned Runs as a function of things the pitcher can, theoretically at least, control.
We do this in order to control for things like the defensive play behind a given pitcher on a given team, or the size of the park that the pitcher pitched in, and other factors outside of the pitcher’s control. In order to build this model we will make a few assumptions.
- We will assume Pitchers have some control over strikeouts, walks, and HBPs; which are by and large true assumptions, certain batters can control their strikeouts or walks, but those batters are fewer and further between relative to the number of pitchers who do control these factors.
- Next we will assume that pitchers have control over the type of contact that they can induce. For instance, in recent years the Pittsburgh Pirates have made a big push for their pitchers to induce more weak contact and give up more ground balls up by keeping the ball down in the zone. In our model this assumption will manifest itself in including the number of GBs, FBs, LDs, and IFFBs, a pitcher gives up.
- A final assumption we will make is that base runners are random events; in other words the effect of one hit or walk occurring doesn’t impact the probability of the next hit or walk occurring. Perhaps a clearer way of putting this might be that a pitcher doesn’t let the fact that there is a base runner on change their ability to get future batters out. This isn’t to say that hits can’t be strung together, but only that hits being strung together are due to random chance and not a pitcher becoming flustered. This is a weak assumption as there is an obvious difference, even mechanically, from nobody on to the first base runner as Pitchers have to change their mechanics from the full windup to the stretch, meaning that the probabilities are likely affected. Additionally, it does seem that some pitchers become flustered with base runners on and pitch worse because of it; as such we will relax this assumption later on, but for now it will work.
Our model then involves Earned Runs as a function of GBs, FBs, LDs, IFFBs, SOs, BBs, and HBPs. Here’s what that regression using the pitching data for all 2017 pitchers looks like:
This gives us an Expected Earned Runs (xER) equation of:
Thus, if we are given the various inputs, the answer we get is the expected number of earned runs a pitcher will give up if they had a league average defense behind them. This is an accurate model, correlating at about 97%.
We can also turn these expected earned runs into Expected Earned Run Average (xERA) by just plugging them into the ERA formula for earned runs. This helps us compare different pitchers on the same terms, since the variations in team defense etc. are controlled for in this analysis.
Starting pitcher performance
Here’s a list of the Pittsburgh Pirates Starters ranked by xERA
There are several takeaways from this list. The first is that Trevor Williams almost certainly has an inside track on a spot in the starting rotation next year. His ERA and xERA are very similar, suggesting that the numbers he puts up are the numbers he deserves to have. That is not to say that he can’t improve, just that any improvements he makes will likely mean sustainable numbers going forward.
With Nova the conclusion is more difficult to come to.
The second takeaway is that Cole, Williams, and Taillon are all above the league average in xERA of 4.42 for starters. Nova is very close to league average, and Kuhl is below average. This is despite the fact that all but Taillon have actual ERAs below the league average of 4.49. This means that the Pirates starting rotation has been above average this season, but perhaps is due for some regression going forward.
The next takeaway is the differences in ERA and xERAs of Ivan Nova and Chad Kuhl. This difference suggests that Nova and Kuhl are due for some regression, meaning that they will likely give up more runs going forward. There is another possibility that they violate the assumption that base runners are random occurrences, meaning that these two actually pitch better with base runners on, or that they bear down on batters when the going gets tough. This would mean that they give up less runs with runners on than the rest of the league does.
I have a tougher tme believing this second explanation as it really relies on something that I have not seen, anecdotally, with Kuhl. It seems that Kuhl can actually become more flustered with base runners on, rather than less, meaning this model would predict him having an xERA lower than his actual ERA. Since the opposite is true, this suggests that Kuhl has actually been the beneficiary of some good luck this season.
With Nova the conclusion is more difficult to come to. Perhaps he really does violate this third assumption and pitches well with runners on; however, as of late, he has been pitching with a lot of runners on, and has been looking even worse than before. Again, while this is only anecdotal evidence, we would expect his xERA to be less than his ERA, however, since the opposite is true, we would expect his ERA to continue to inflate until it is closer to the 4.50 mark.
The recommendations for the two are much different. In Kuhl’s case I would suggest moving him to the bullpen. A 4 pitch pitcher that can touch 100mph coming out of the pen is a dangerous weapon that any team would like to have. Additionally Kuhl in the bullpen would likely be an upgrade over the likes of Joaquin Benoit, Daniel Hudson, or Wade LeBlanc.
In Nova’s case, the recommendation is contingent on what the other options are. As it stands right now Nova is the #4 starter behind Cole, Williams, and Taillon. Tyler Glasnow is almost certain to get another shot at a starting role, and depending on what he does over the next few weeks and spring training next season, Nova could become a #5 starter. With Steven Brault and Nick Kingham also getting looks as a starter, Ivan Nova could relatively easily be the odd man out next season.
If this is the case, or the Pirates just prefer to give the starting job to a young guy like Kingham, who may be out of options for 2018, then the Pirates should seriously consider trading the reasonably priced veteran elsewhere for some kind of return this offseason.
The last note from this list is the near 2 run difference between Glasnow’s ERA and xERA. Having such a stark contrast between the two suggests that Glasnow may have been the victim of some bad luck, likely combined with a breaking of the 3rd assumption as he may have allowed for the pressure of the situations to get to him, leading to additional runs. Hopefully, Glasnow’s additional time in the minors, and appearances in the higher leverage playoffs, has helped him fix the latter problem, making him a legit starting rotation option going into 2018.
Now what about the relievers
* denotes a free agent at the end of the 2017 season
The problem with analyzing this is that most of the young arms have only performed over small sample sizes so we can’t obtain any statistically significant findings. That being said there are a few takeaways from this list.
For context the league average ERA for relievers is 4.18 and for xERA is 4.38
First and foremost Rivero is utterly dominant; I don’t think that this is a surprise to anyone that has watched him pitch this year, but his ERA and xERA are the 3nd best in the NL among relievers, making him undoubtedly elite.
Second is that Kontos seems like a solid pickup. Over the entire season, Kontos has a 3.39 xERA. I doubt his current 2.57 ERA with the Pirates will remain that low, but a 3.40 ERA bullpen arm is a definite upgrade from where the Pirates were earlier this season.
Third is that Hudson, LeBlanc, and Benoit have been bad, and it isn’t because of bad luck. All three pitchers have high ERAs and similarly elevated xERAs, suggesting that it hasn’t been bad luck that has resulted in them not being effective, but rather just bad pitching. None of these pitchers are anywhere near league average ERA, with Hudson and LeBlanc at least being relatively close to the xERA average.
The final takeaway is that Schugel and Neverauskas have benefitted from some good luck, and look solid, Brault and Santana have not been pitching well so far, but only Schugel has done so over a long enough period of time that we can reasonably predict that this is the kind of pitcher he will be going forward. Projecting any of these pitchers with accuracy would be difficult over such a small sample size, but all of them have displayed tools that could allow them to be solid Major Leaguers.
xERA tells us that the Pittsburgh Pirates starting rotation is actually in decent shape for next year. Having a lot of depth at pitching is never a bad thing, particularly when most of that depth is made up of average to above average starters. The bullpen is clearly weak; only 4 of the 9 listed have above league average ERAs and only 3 have above average xERAs. Moving Kuhl into the pen and adding bullpen help this off season should help solidify this weakness and, while not necessarily making the relief a plus for the Pirates, will at least stop it from costing them wins.