clock menu more-arrow no yes

Filed under:

The Reliable Reliever

New, 14 comments

A quality management approach...

MLB: Philadelphia Phillies at Arizona Diamondbacks Mark J. Rebilas-USA TODAY Sports

Kevin Ginkel’s story was a terrific story in 2019. In 2018 he all of a sudden got everyone’s attention in the minors, racking up strikeouts for Visalia and Jackson (A+ and AA). In 2019 he continued that trend in Jackson and, especially, in Reno (AAA). Once he made his MLB debut, he continued to impress.

Those days seem long gone in 2021. In 2020 he was a disaster in a shortened season and this season he has had some severe ups and downs. He started well, in his first 9 games, and then gave up multiple runs. That has led to some despair on the SnakePit.

Ginkel also led to a tiny discussion on here on reliable relievers. What is reliable? I explained my thoughts to Jack via e-mail and he encouraged me to write an article on it and share my view.


What is reliable?

First of all, it is important to define what reliable is. Cambridge dictionary defines reliable as:

Someone or something that is reliable can be trusted or believed because he, she, or it works or behaves well in the way you expect.

What is key in this definition is: it works or behaves in the way you expect. It implies that you and me will have to agree on what we expect in order to name something reliable.

We also have to agree on what can be considered a “good” performance, because the definition tells us that something that is reliable has to behave well. In other words, we cannot deem a pitcher who continuously gives up 9 runs as reliable, because that would be a performance considered insufficient for a pitcher in general.

But, on the other hand, a reliable reliever does not necessarily need to be a reliever that is “lights out”. You might have a reliable car that has brought you from A to B for many years with hardly or no breakdowns. You might have a reliable colleague who never falters and always provides good work. You might have a reliable handyman that fixes everything in your house or apartment. But they are not necessarily stars: if they were, they might not be with you.

Defining a reliable reliever

Those of you that are familiar with project management and/or manufacturing processes might know quality management processes. When I think of a reliable reliever, I think of a reliable process. How could we define this reliable process and measure its quality?

In quality management control, the control chart is a very popular way to measure how a process evolves over time. You set a time period and then plot the data.

In my definition of a reliable reliever I look at his outings. A season is the time period and the outings are a unit. It does not matter if an outing is 0.1 innings, 0.2 innings or a complete inning, because an outing is a process. The defence behind him does not matter, the park he pitches in does not matter...at least it does not matter for this quality control process because it forms part of the environment the process (=the outing) runs in. This goes a bit against the quality management theory, because to compare the performance of a process over time, it should always be executed in the same environment (but this is my article and we are just having fun here - I can understand if your definition of “fun” is different than mine).

The runs (not earned runs: just runs) a reliever gives up are the data. They could be considered as defects. The mean of the defects should have a value of 0 and 1 in baseball. The more runs a reliever tends to give up, the higher his mean would be. The defects themselves, however, should hover around the mean, being 0 and 1. The more zeros, the better the reliever, but more ones does not necessarily disqualify a reliever either.

Kevin Ginkel vs Taylor Clarke: first matchup

The funny thing about comparing Kevin Ginkel to Taylor Clarke is that, in my opinion, I consider Taylor Clarke a very reliable reliever, unlike Kevin Ginkel. I was happy (and a bit relieved) to see this as well in the data.

I took data until last Saturday (June 5). Kevin Ginkel had pitched in 27 outings and Taylor Clarke in 26 outings.

What do we see here? Well these are the outings of Taylor Clarke (left) and Kevin Ginkel (right) in 2021 and how many runs they gave up in each. Kevin Ginkel’s outings have a lot of peaks, which shows he is not reliable. Taylor Clarke’s outings have less peaks and the runs he gives up hover more around the mean.

But I can understand that you consider this plotting a bit “too easy”. The work does not stop here.

An example of a reliable process: Kenley Jansen outings

Let us take Kenley Jansen as an example. He was the Dodgers’ closer in 2019. At times he brings the Dodgers’ fans to despair, but he is a very reliable reliever. These are Kenley Jansen’s outings in 2019:

There are some peaks in this process that in control charts could be considered as “out of control”. There is a formula for that, but I will not bother you with it. Normally, you would look at those data points that are considered “out of control” and analyse them, in order to bring a process back into control. I will not do that, as we have more advanced statistics to do that for us (but it is probably about park factors, control, velocity and stuff or a combination of it).

But I do wish to point out that it is also very common in data analysis to get rid of outliers, as they can be considered abnormalities. I will do this as well, but not with any strong theory behind it. I just say that “a reliable reliever can have 4 outings that can be considered outliers”: they do not form part of the process.

If we apply that thought to Kenley Jansen’s outings in 2019, his stat line turns into this:

Now a Kenley Jansen process has turned into just one multiple run outing and all others hovering around the mean. Even if Kenley Jansen would have given up more 1 run games, he would still enter my definition of a reliable reliever.

Kevin Ginkel vs Taylor Clarke: second matchup

Having eliminated some outliers, Kenley Jansen has become the definition of a very reliable process. What if we remove some outliers from Kevin Ginkel and Taylor Clarke? Kenley Jansen in one season had 62 outings. Ginkel and Clarke are almost at half of that, so I will remove two outings and consider them outliers. Now we get to see the following processes:

Taylor Clarke now looks like a reliable reliever. He does not put up the amount of zeros Kenley Jansen put up in 2019, but the performance of his outings hover around the mean without any abnormal peaks. That is just fine. In Kevin Ginkel’s case though, there are still a couple of outings with multiple runs, so his bad outings are part of the process: the process is out of control and that is what makes Kevin Ginkel unreliable.

It even visualises the stressful situations, since there are a lot of continuous up and downs: if I were a cyclist I’d rather take the Taylor Clarke tour than the Kevin Ginkel tour.

Was Fernando Rodney reliable?

Who does not remember the Fernando Rodney experience in 2017? Do you think he is reliable? If you look at his outings in 2017 in general, he put up a lot of zeros. That is a big argument in favour of him. After removing 4 outliers (right), he still has 3 big peaks. I would say Rodney was not a reliable reliever, because if he would blow it, most of the times he would blow it up big time. It was the Fernando Rodney experience.

Was Greg Holland reliable?

Greg Holland had his redemption for the Colorado Rockies in 2017. He had a long stay in free agency in 2018 and signed late with the St. Louis Cardinals. His stay there was one to forget. He was picked up by Washington and found somewhat of his groove again. In 2019 he was a Diamondback. He was good for the first couple of months, but was DFA’d in August after a couple of sorrowful outings. Greg Holland made 40 appearances for the Diamondbacks so we allow 3 outliers to be removed. That leaves him with just 1 outlier in 37 games. Perhaps Greg Holland’s performance could have gone even more downhill in August (although in these graphs downhill means mountains), but according to my view and until the cutoff, he was a reliable reliever with just one peak (uhm...did I say the same when he was DFA’d?).

This is all a bit arbitrary of course, because we haven’t decided together on what a reliable reliever is. To start the conversation on that, I have given you my view on what a reliable reliever is and I hope you will share your thoughts on what yours is!