Baseball is a game of moments. The parts of the game that some non-fans believe to be slow and boring are really just build-up to split second plays that have the potential to determine the outcome of entire games. Stolen bases are fun because they are a microcosm of this exact aspect that makes baseball great. As fans, we watch over the pitcher’s shoulder toward first base, calculating in our minds when the runner needs start towards second to get a good jump and predicting whether he will land in scoring position safely or be sent walking, with shame, back to the dugout. As the runner’s legs pump, propelling him towards his destination, so do our hearts. The play converges at one endpoint, culminating in an emphatic ruling from the umpire on the runner’s fate — and in the post-2014 era, an ensuing video review that halts play for about two minutes. But stolen bases seem to be vanishing as of late. With all of the rule changes and trends that are forcing a transformation in baseball as we know it, there can be a ton of different reasons for the growing abandonment of the stolen base. Catchers with better arms to gun runners down? A league-wide move towards powerful, homerun-hitting players and away from elusive speedsters? It’s a phenomenon that’s affecting the league whether we like it or not, so I’m going to attempt to use some basic statistical analysis and barebones visuals to get to the bottom of the decline in SBs.
Are stolen bases actually on the decline?
Before we start digging into why stolen bases are declining, we should probably do some basic research into if the number of stolen bases are, in fact, declining at all (Spoiler: If the answer was no, I probably wouldn’t have written this article).
Here’s a graph of the number of stolen bases per game across the MLB in each season since 1970. I use per game stats in a lot of these visuals because it allows me to include seasons that were shortened by things like player strikes (1981,1994,1995).
The answer is yes, stolen bases are steadily decreasing, down from their heyday between the late-70s and mid-90s. In 2018, there were 2474 total stolen bases in the MLB. This is the fewest in a non-shortened season since 1973, when there were 6 fewer teams in the league. As I mentioned, stolen bases hit a steady groove around 40 years ago that it maintained through the heart of the 80s. As a result, moving forward, we’ll use the 80s as a benchmark to compare to present day.
But wait, are ATTEMPTS actually on the decline?
That graph is nice and stolen bases are definitely lower than they used to be, but how do we know that runners aren’t still attempting to steal the same number of bases (with less success)? If attempts are the same, then that would mean that we’re still getting the same number of stolen base events, with all the same excitement. Well, it’s as simple as graphing Stolen Base Attempts (SBA) over time.
Yes, MLB teams are attempting to steal bases at the lowest rates since stolen bases were in their prime. So we’re not crazy. But what’s behind these drops in SB and SBA?
Are teams getting fewer opportunities to steal?
Here’s where things start to get interesting. Baseball has transformed in more ways than one since the 80s. One highly documented trend is the heavy increase in two highlight-worthy events, one on each side of the ball. Offensively, teams have been loading up on players that can hit the ball out of the park. The last 3 seasons all reside in the top 4 all-time for most HRs league-wide. Pitching-wise, strikeouts have become a bigger part of the game than ever before. In fact, each of the last 11 seasons set the record for most league-wide strikeouts in MLB history (only to be broken the next year, of course). Both of these plays, while fun to watch, keep runners off the basepaths. Which begs the question: Have stolen base opportunities decreased?
…and the answer is… kind of? The number of stolen base opportunities the past few years is slightly below that of the 1980s, but there was also a significant peak from the mid 90s to the mid 00s, which is around the same time there was a big crash in stolen base attempts (refer to the last graph). However, if we combine the last two graphs that we saw and put them together, we get a better idea of whether stolen base opportunities are the culprit for the decline in stolen bases.
The graph here is of a stat that I’ll call Stolen Base Attempted% (SBA%). Simply put, it is the percentage of stolen base opportunities in which a stolen base is actually attempted. Mathematically, this is simply SBA/SBO. The value in analyzing the trend of this stat is that it measures the number of stolen base attempts while accounting for any fluctuation in stolen base opportunities. What we find is that there is an obvious decline in SBA%, which we can interpret as evidence that any decrease in SBO is not the reason for the decline in SBA.
Stolen Base Success Rate?
To recap, stolen bases are definitely on the decline, and so are stolen base attempts, but neither are a result of a decrease in opportunities. However, even with both SB and SBA down, it is still possible that the stolen base success rate (SB%) is going down, as long as SB is decreasing at a faster rate than is SBA. If this is the case, then we could probably point to an increase in difficulty of stealing bases, whether it be because players are getting slower, catchers’ throws to second are getting faster, or something else, like a league-wide change in pitchers’ wind-ups.
Nope. Still haven’t found our answer. Not only has SB% not declined over the years, it has actually pretty steadily gone up! But wait, if teams are getting more successful at stealing bases, why are they stealing fewer? One potential reason is that teams are getting more selective about who tries to steal for their team. If teams are placing a sort-of “ban” on certain players to stop them from trying stealing bases and only giving a small handful the green light, that could both decrease the number of stolen base attempts while driving up SB%, since only the fastest players would be stealing. If SBA drops enough, even with an increase in SB%, total SB would still decrease.
For me, this brings up another important question, though. Why would teams suddenly start to shrink the number of players that they allow to steal bases? The answer that jumps out to me is that it has to do with the value of a stolen base. For one reason or another, it seems that teams now see the value of a stolen base as lower than they used to, when stolen bases were more commonplace.
The Value of a Stolen Base
The value of a stolen base can be directly quanitified using a run expectancy matrix known colloquially as RE24. The RE24 matrix is generally used to measure “the change in run expectancy from the beginning of a player’s plate appearance to the end of it”. Essentially, the matrix gives the run expectancy, or “the average number of runs and average team would be expected to score during the remainder of the inning”, for each “base-out state”, where a base-out state is the current state of baserunners and outs. For example, in the generic RE24 matrix below, in the state where there are runners on first and third and 1 out, the average team would be expected to score 1.14 runs, on average, over the course of the remainder of the inning.
Going back to the context of our question, the RE24 matrix is helpful because we can use it to calculate the change in run expectancy that results from a stolen base (or a failed stolen base). In practice, we have to get the run expectancy of a starting base-out state (e.g. runner on first, 1 out) and calculate the difference between that and the run expectancy of the ending base-out state (e.g. runner on second, 1 out). To make things simpler, we will only take into account stolen base situations that begin with a sole runner on first that attempts to steal second (rather than include other situations, such as a runner on second that attempts to steal third). We still have to account for each of the situations in which a runner is on first and can steal second (0, 1, and 2 outs). We can then apply weights to each of those three situations based on the number of occurrences that each situation occurs and take the average. This would give us the average estimated value, in expected runs, of a single stolen base. Since we are checking whether the value of a stolen base has decreased, we should calculate the value for a reference year and then compare it to the value of a stolen base in 2018. As I mentioned before, the 1980s featured a pretty steady number of stolen base attempts, so we will use 1985 as the reference year.
Using our method, we can calculate the average value of a stolen base (from first to second with no other runners on) in 1985 using the above matrix. We get an estimated value of .164 runs. It would be helpful to also calculate the value of a failed attempt at a stolen base (which would be negative) and use it to determine the specific stolen base success rate needed to make even attempting a stolen base worth it in the long run. To avoid boring people with the basic algebra needed to calculate this, I will do it behind the scenes and cut straight to the answer.
In 1985, I estimate that a stolen base success rate of 69.83% would, in the long run, break even in value for a team. This means that the sum of the values of all the attempted stolen bases would equal 0. Of course this estimation is very rough, and the real value is probably slightly different, but it is helpful in comparing the value to that of present day.
Using the 2018 RE24 matrix above, I calculate that a stolen base success rate of 70.08% would break even. This means, according to my estimations, the value of a stolen base is, in effect, not different than it was in 1985.
Although these calculations didn’t tell us that the value of a stolen base has changed, it has highlighted something important. In the years before 2000, when SB% was routinely below 70% (refer back to Figure 5), teams were stealing at a rate that would have actively hurt them in the long run. It’s possible that teams in the heyday of stolen bases simply had no way of tracking how valuable a stolen base was and, as a result, attempted to steal with much less caution than they do in the present day.
Could the archetype of the average MLBer play a role?
The final thing that I want to question is if a movement towards teams using bigger and more powerful players has any role in the decline of stolen bases. Players have certainly gotten bigger, and ISO, a stat that attempts to measure power using extra base hits, was the highest in history in 2017 (though this could have been affected by things other than size). It would seem that the bigger players get, the slower they are. It is hard to substantiate this claim with the data that I have available to me, however it is worth posing the question for future research, especially since SLG and ISO made a significant jump in the late 90s, around the same time that SBA and SB began to dip.
It’s hard to make a conclusion on something that probably has several different factors play into it. I don’t have the ability to ask teams why they’ve changed their strategies or what their analytics teams are telling them. However, I can make an educated assertion as to what has happened. My hypothesis is that in the 80s and most of the 90s, and before that as well, teams were mostly clueless about the potential value of stealing. Advanced stats didn’t make an entrance into the MLB until the early 00s. Before that, teams were likely very inefficient in many aspects of the game simply because they had no way to measure how efficient they were. As analytics ramped up and teams began to track every part of their strategy as well as the expected value of every move, they realized that stealing isn’t as lucrative as they once thought. It is also possible that teams have moved away from prioritizing speed on their rosters, shortening the list of players that can steal with a beneficial success rate. Certain teams have more speed than others, which is why certain teams steal more often, and it’s obvious that teams are assessing how good they are at stealing bases given that the teams that are attempting more steals are the ones with the highest success rates.