26 April 2017

Strikeout Rate Stabilization and Time

Thanks to the wonderful work of Baseball Prospectus, we understand that strikeout rate - and other offensive categories - stabilize at a relatively small number of plate attempts. While statistic stabilization is commonly referenced, it's worth exploring what this means - and how it can be taken advantage of.

Rate stabilization gives us a sense of how small of a sample size is needed before a player's average talent can be estimated. Baseball Prospectus determines this by finding the sample size at which any randomly selected sample has a high correlation (over 0.70, to be exact) with any other randomly selected sample of the same size. For example, Manny Machado's last 60 plate attempts should feature a strikeout rate that is highly correlated with his strikeout rate over 60 plate attempts in 2015. Further, Manny Machado's strikeout rate over any 60 plate attempts, even non-continuous ones, is highly correlated with his strikeout rate over any 60 other plate attempts. The concept of course makes perfect sense: a player is who he is, and has a true level of ability. The important part is the number of plate attempts it take to estimate with a degree of certainty a player's true talent level.

The concept of rate stabilization is frequently used to make judgments on a player's ability early in the season, or sometimes early in his career - and they should be. The concept of rate stabilization is proven over the entire population of Major League Baseball and holds true overall. However, in the interest of testing heuristics, I sought to understand how much variation there is in a player's 60 plate attempt strikeout rate over time. The addition of time is, I think, very relevant to understanding a player's true talent level, and something that I might argue is missing from the common knowledge on any rate stabilization.

I want to bring time into the conversation of rate stabilization because it allows for players to grow, adapt to the speed of the Majors, change their approach, fix their swing,... Basically for players, especially young players, to mature into the professional ballplayer that they will be for the majority of their careers. It wouldn't be fair to assume Albert Pujols' 13.8% strikeout rate in his first season was his true level of talent and to be repeated for the next 17 years of his career. In fact, Pujols recorded sub-10% strikeout rates in nine of the next eleven seasons.

Was Pujols simply lucky in the first 10 years of his career to consistently record strikeout rates below that of his first year? It's far more likely that his true, natural strikeout rate decreased (read: his true talent level increased) as he grew to know the speed of MLB pitching and learned tendencies and pitch sequences and studied film and practiced. Pujols probably practiced a lot. How many other players can we identify as elevating (or lowering) their level of true talent rapidly - before seasons end or before they become a coveted free agent?

I believe we can, and I set out to do so among Orioles players. Consider this a proof of concept. By examining the rolling 60 plate attempt strikeout rate of Orioles players, we can observe instances in which a batter is striking out more or less often (a LOWESS line makes this trend easier to pick out). We expect some variation, obviously, since performance over 60 plate attempts is simply correlated with performance over 60 other plate attempts. However, a player that is consistently lowering his 60-plate-attempt strikeout rate may have actually improved his ability to not strike out (be it through better contact rates, zone awareness, or pitch recognition). In an attempt to parse out which changes are worth taking note of, I also plotted the number of standard deviations from the mean strikeout rate that the 60-plate-appearance strikeout rate falls. Sustained periods of time that are greater than 1 standard deviation are worth considering as repeatable, new normals. Doing so would require digging into data and talking with the player to understand why his strikeout rate is so different than normal, and if it's a new repeatable skill.

Take, for example, Machado. His 60-plate-attempt strikeout rate has increased over the last few hundred rolling instances of 60 continuous plate attempts. Although Manny's rolling 60-plate-attempt strikeout rate is still less than one standard deviation from his mean, we might begin to wonder whether his true strikeout rate is a bit higher than it was when he entered the league. This is not necessarily a bad thing. Although, yes, all else equal, a lower strikeout rate is better, all is rarely ever equal. Strikeouts are usually the price of hitting for more power. This is not an exercise to say that Machado should change his approach at the plate lest his value drop because of his strikeout rate, but instead to recognize that in hitting for more power, Manny's strikeouts might increase.

Consider also Mark Trumbo, whose extraordinary 2016 was his third out of the last four years to feature a strikeout rate of less than 25%. Trumbo has demonstrated a strikeout rate about a half of a standard deviation below his career average for long enough that we might begin to question whether his recent performance is a better indicator of his talent than his career average:

Ryan Flaherty's 60-plate-attempt strikeout rate has consistently increased since 2013, ranging from a full standard deviation below his career average to nearly a standard deviation above average. Perhaps it's no surprise that offensive opportunities have dried up for Flaherty:

And finally, Joey Rickard has only six plate attempts this season but rolling plate attempts over seasons shows a pretty dramatic decrease in 60-plate-attempt strikeout rate. With some practice and some exposure to Major League pitching, Rickard may have learned to handle the strike zone a little better than when he first entered the league. This is an instance where a player's improvement would merit some further investigation; Rickard has never been much for on base percentage, but slight changes in approach might have made his offense slightly more valuable than expected. Obvious follow up questions that seek to determine whether Rickard has actually gained more control of the strike zone include whether Rickard is swinging less often, swinging at pitches out of the zone less often, whiffing less often, and walking more (walk rate stabilizes over 120 plate attempts).

Rickard is also a fine example of an instance in which this method might be used to judge a young player's performance with more information than his career averages or any randomly selected 60 plate attempts. We should expect Rickard to have grown and adapted to Major League pitching; to expect his potential and performance at ages 25-30 to match identically his performance at age 24, which he was getting his very first exposure to the Majors, would be rather shortsighted. Rickard may not ever be a stud (he won't be), but he might scratch the replacement level mark this season as a batter with a 15% strikeout rate instead of the 20% strikeout rate he's projected by ZiPS and Steamer to exhibit.

A prospect with more promise or better peripherals might show a similarly snaking rolling 60-plate-attempt strikeout rate over his first season or two. This might prompt a team to focus on or reinforce plate discipline, contact, or control over the zone depending on the drivers of the change in strikeout rate. Similarly, if an established player exhibits a strikeout rate consistently below his career strikeout rate, a team might consider his career average to underestimate his ability to reduce strikeouts.


Roger said...

The kid's gotta stay healthy enough to measure......

Jacob Smith said...

Another possibility to consider with Rickard may be that he was used more and more as a platoon-only bat as the season went along. That could make his talent level appear to grow as a result of being utilized proportionally more from his stronger side. I don't know this to be true, but it seems like his season went that way, and his platoon splits were massive.

Patrick Dougherty said...

Jacob - that is certainly a possibility! In this instance, it takes a look at Rickard's usage and platoon splits. I did not, but that kind of follow up is exactly what this indicator should prompt. The same kind of analysis could be done across handedness splits as well, questioning whether Rickard's ability to avoid strikeouts improved or remained consistent throughout last season and the beginning of this one.

Anonymous said...

Chris Davis will be in the top 100 strikeout list in MLB history in a couple of months, already! Move over, Reggie!