Stealing Bases Isn’t the Uphill Battle It Used To Be. Can Defenses Maintain the High Ground?

Brett Davis-USA TODAY Sports

As we know, baseball is a bit of an oddball relative to other ball-centric sports for several reasons. Prominent among them, the defense controls the ball at the start of each play, whereas in basketball, football, soccer, and hockey, to be on offense is to be the team with the ball. There exists a mindset difference between playing offense and playing defense, or rather between controlling the ball versus not controlling the ball. One is proactive, the other reactive. As players develop they, whether consciously or not, sort themselves into positional groups partially based on their preferred mindset (alongside their natural skills and physical attributes). Some need the comfort of control, while others thrive on guessing their opponents’ next move.

Pitchers and catchers fall in the proactive category, selecting pitch types and locations to best baffle hitters. Position players react both at the plate and in the field. On the basepaths, the roles reverse. Runners make the active decision to advance, leaving pitchers and catchers to react. It’s an abnormal experience for everyone involved.

Season four of Stranger Things hit Netflix on May 27, 2022; around Opening Day of the 2023 major league season, you finally got “Running Up That Hill” by Kate Bush out of your head. (If you don’t watch Stranger Things, just know that the song features prominently throughout the show’s most recent season.) And as the new season dawned, baserunners went wild on the basepaths and all the chatter about running wormed “Running Up That Hill” right back into your brain. Much in the way the show revived a song from the 1980s, changes to MLB’s rules regulating base sizes and pitcher disengagements revived ‘80s-esque stolen base rates.

Running up a hill is hard. Most of us have the luxury of simply not doing it if we don’t want to. When position players find themselves in the atypical role of proactively initiating a play, they too can simply choose not to. Whether they resent the added pressure of making a decision, carry an aversion to the injury risk associated with basestealing, or have seen the break-even probabilities and perceive themselves as substandard baserunners, they can decide ahead of time to stay put. Pitchers and catchers don’t have that luxury. They find themselves needing to guard against a sneak attack. They’re the ones donning headphones and pressing play on a comfort song in an attempt to maintain a safe, controlled environment, just like Max avoiding Vecna.

On a per-team-game basis, stolen base attempts are up 33% so far this season when compared to 2022, which amounts to about 36 additional attempts per team over the course of the season and, at the current success rate, about 29 extra bags. The hill that is stealing bases was once steep enough to discourage runners. Now it has leveled out. But those numbers are all averages. Several teams and players seem to be coasting downhill on a skateboard. The teams listed below are on pace to roughly double their stolen base attempts while also improving their success rate relative to their own 2022 baseline. For a few teams, the increase is partially attributable to roster turnover, but such steep increases signal a strategic shift as well:

Largest YOY Increase in SB Attempts
Team 2022 SB Attempts 2023 SB Attempt Pace SB Attempt % Change 2022 SB 2023 SB Pace SB % Change
Reds 91 224 146% 58 178 208%
Rays 132 250 89% 95 198 108%
Padres 71 168 137% 49 137 180%
Pirates 121 215 77% 89 162 82%
A’s 101 192 90% 78 158 103%
Stats current through 6/26

To ease comparisons on the player level, the counting stats listed below have been adjusted to a 162-game pace. The players shown here rank in the top 10 with respect to the year-over-year increase in stolen-base attempts per game. For most of them that translates to a hefty percentage increase in attempts (less so if they’re ratcheting up an already healthy volume of running), and with the exception of Jeremy Peña, who is getting thrown out a 40% clip, a comparable or better percentage increase in successes. So again, more aggression, with no harm done to the success rates:

Largest YOY Increase in SB Attempts
Player 2022 SB Attempt 2023 SB Attempt SB Attempt % Change 2022 SB 2023 SB SB % Change
Ronald Acuña Jr. 54 85 56% 39 73 84%
Cody Bellinger 19 43 126% 16 36 129%
Bobby Witt Jr. 40 61 53% 32 48 49%
Ha-Seong Kim 15 35 132% 13 28 120%
Thairo Estrada 31 51 62% 24 43 77%
Nico Hoerner 26 45 71% 24 40 69%
Christian Yelich 23 41 77% 20 37 84%
Starling Marte 37 54 46% 25 47 91%
Jeremy Peña 15 31 100% 13 18 35%
Trent Grisham 9 21 157% 7 15 105%
Stats current through 6/26

But stolen bases represent just one baserunning threat. The rule tweaks that empower runners to swipe bases with greater audacity apply to other acts of aggression on the basepaths, such as using greater primary and secondary leads to take the extra base on balls in play.

When tracking the rate of a runner moving from first to third or second to home on a single (with no other runners on to act as an impediment), we see only a very small uptick. The probability of a runner attempting to go first to third is up just under 1% from last season, while the success rate on attempts is up just over 1%. Overall, it appears there’s not much to see here, but as above, averages sometimes obscure more extreme data points. Drilling down situationally shows the small increase in likelihood of an attempt happening with two outs. With fewer opportunities for a teammate to hit them in, runners decide to proactively create their own opportunities:

First-to-Third Attempts Split by Outs
Outs 2022 Attempt Rate 2023 Attempt Rate 2022 Success Rate 2023 Success Rate
0 26% 27% 98% 99%
1 28% 29% 96% 98%
2 34% 36% 96% 97%
Stats current through 6/26

Comparing 2022 to 2023 and assessing first-to-third probability relative to the situation’s leverage index (which abbreviates to LI, so I like to call it the Lemonbooty Index) reveals that the importance of the moment doesn’t seem to sway the decision much:

First-to-Third Attempts Split by LI
LI 2022 Attempt Rate 2023 Attempt Rate 2022 Success Rate 2023 Success Rate
0 to 0.5 26% 27% 97% 99%
0.5 to 1 34% 35% 97% 98%
1+ 29% 30% 97% 97%
Stats current through 6/26

Meanwhile, second to home attempts on a single have seen a slightly larger jump in likelihood, going from just under 62% in 2022 to 66% in 2023 with a success rate holding steady at 95%. When splitting on outs in the inning, the data shows that runners are least likely to attempt going second to home with no outs, but the increase in attempts from last season to this season came almost entirely in this situation, with attempts rising from a probability of 36% in 2022 to 44% in 2023. Runners’ likelihood of going home with two outs has hardly changed from last year to this year, but there was also less room for an increase since runners were already going for it 83% of the time. And all of this comes with no meaningful change to the upper-90s success rate:

Second-to-Home Attempts Split by Outs
Outs 2022 Attempt Rate 2023 Attempt Rate 2022 Success Rate 2023 Success Rate
0 36% 44% 95% 98%
1 55% 58% 97% 99%
2 83% 85% 95% 94%
Stats current through 6/26

Looking at LI, runners’ newfound boldness actually manifests more in higher leverage situations. This year, runners attempt to go home in the highest leverage situations at the same rate attempted in medium-leverage situations in 2022. With a greater reward attached to touching home than touching third, runners tolerate greater risk when making the decision to go from second to home, as opposed to going from first to third:

Second-to-Home Attempts Split by LI
LI 2022 Attempt Rate 2023 Attempt Rate 2022 Success Rate 2023 Success Rate
0 to 0.5 66% 68% 97% 95%
0.5 to 1 64% 70% 97% 98%
1+ 59% 64% 94% 96%
Stats current through 6/26

The chorus of “Running Up That Hill” begins,

And if I only could
I’d make a deal with God
And I’d get Him to swap our places

In a radio interview, Bush explained that the swap refers to a man and woman, symbols representing a dichotomy, where conflict often stems from a lack of shared perspective that she feels can only be remedied by a role reversal, “And if we could actually swap each other’s roles, if we could actually be in each other’s place for a while, I think we’d both be very surprised! And I think it would lead to a greater understanding.”

The baserunning dynamic forces the role reversal yearned for in Bush’s song. The proactive become reactive, and vice versa. While I doubt anyone in this scenario was yearning to be Freaky-Fridayed into their opponent’s situation, trying out new roles does spark growth, and watching elite athletes adapt and respond to challenges is a big part of what we’re here for.

As demonstrated, runners are embracing their new role and behaving more proactively on the basepaths, albeit more so with respect to steals than taking an extra base. For pitchers and catchers to settle into their new role, they need to acquire skills and weapons for reacting to and defending against the baserunning monster before them.

The gate that opened to unleash active basestealers this year is the new rule limiting pitcher disengagements, which for the purposes of this conversation translates to limiting a pitcher’s primary means of reacting to an increased running threat: the pickoff throw. With pitchers only able to throw over twice before an unsuccessful third attempt results in a balk, the baserunner’s mental risk assessment becomes much simpler.

Though the pickoff throw has lost some effectiveness as a run defense tactic, it’s not completely useless. It’s the difference between Steve Harrington defending himself against Demodogs with a standard wooden baseball bat and one with nails protruding from it. In the past, throwing over was low-risk, high-reward. Pitchers risked a throwing or fielding error that allowed the runner to advance, but that doesn’t happen too often at the professional level. The reward is either an out or a warning to the runner about the dangers of straying too far from the bag. Now, both the pitcher and runner know that throws over are a finite resource, and as quantities dwindle, usage becomes stingier. With each attempt, the runner knows exactly how much ammunition the pitcher has left and that the threat against him is shrinking. Rather than a pickoff saying, “Hey, I’ve got my eye on you,” it says, “I’m running out of ways to stop you.”

So how does the probability of a stolen base attempt change as pickoff throws increase? Were pickoffs actually discouraging runners under the old rules? How much has the impact of a pickoff decreased under the new system? The answers to such questions depend heavily on the runners and game states involved, so it’s necessary to compare runners only to themselves and at a minimum keep the situations competitive. This requires chopping the data into pretty small pieces, which limits the types of conclusions we can draw.

Looking at players who were on first with no other runners on base in a reasonably close game (i.e. neither team’s win expectancy was greater than 65%), and who experienced at least five plate appearances with a throw over, I compared the likelihood of taking off in the PAs with a throw over versus those without. Overall, the difference in attempt rate was negligible. There were certainly individual players who were both far more and far less likely to break for second, but at the individual player level, the samples are so small and the game situations so specific that I hesitate to draw sweeping conclusions about the effectiveness of the pickoff in discouraging runners from attempting to steal.

But while the pickoff throw doesn’t definitively dissuade runners from going, it does seem to impact their effectiveness in doing so. The success rate on stolen base attempts dropped 14 percentage points in 2022 for attempts following a throw over when compared to attempts that didn’t follow a pickoff throw. Thus far in 2023, we still see a drop in success rate, but a less dramatic one of only nine percentage points.

When comparing multiple throw overs to a single check-in, the stolen base success rate dropped 11% after one throw, and dipped 38% after two or more pickoff attempts in 2022. This year, the diminished number of pickoffs means we don’t have a strong enough contingent of runners who have experienced multiple throws to discern a defined rhythm in the noise, but the 2022 numbers suggest additional throws to first do not obey the law of diminishing returns.

So runners gonna run, but checking in with them does limit their impact. However, that’s less true now that runners know pitchers are restricted in their ability to throw over. Pickoff throws present a strategy where the threat is more important than the act itself. Runner and pitcher both know the throw is unlikely to result in an out, but since the out is so debilitating to the offense (costing around half a run of scoring potential depending on the situation), runners must remain alert. This season, runners have an easier hill to climb.

Now defenses need to find a way to steepen the grade or create a rougher terrain for runners. Pitchers may not be able to throw over more than twice without potential consequences, but no such restriction exists for catchers. The benefits of a catcher back-pick aren’t a one-to-one match, but they’re comparable. Depending on the catcher and handedness of the batter, there may be a higher risk of error. The act also disrupts the runner at a different point in the “to run or not to run?” calculus, so it may take some time to establish the threat of a catcher back-pick such that runners consider it as they mull their options.

Again, we’ll look at situations with a runner on first and the other bases empty, as that is the most common situation for a steal, and here we’ll examine the outcomes following pickoff attempts and back-picks. Since the start of 2022, the probability of an out is about 1.5% on a pitcher throw over and 3.3% on a catcher back-pick. The probability of an error that allows the runner to advance is less than 1% for pitchers and around 4% for catchers. As it stands, catchers are more selective with their attempts, so an increase in throws behind the runner may dampen the out probability, but the increase in reps may also lower the likelihood of an error. And in most baserunning situations, an error hurts the defense less than an out hurts the offense. For example, using a run expectancy table calibrated to this year’s run environment, with a runner on first and one out, the runner advancing to second on an error only provides a 0.119 bump in expected runs, while picking off the runner slashes 0.359 runs from the expected total.

To make an even more informed throwing decision, the break-even method, which is often applied to the basestealing decision, can be used to determine the minimum probability of an out and the maximum probability of an error necessary to mitigate risk. That is, we can estimate the throwing proficiency required of pitchers and catchers for the strategy to pay off.

Break-even formulas evaluate a decision by finding the change in run expectancy associated with each outcome of the decision and then computing the sum of the change in expected runs for all possible outcomes, weighted by the likelihood of each outcome. Offenses, attempting to score runs, aim for this total to be zero or higher. Defenses, attempting to suppress run-scoring, want their decision to yield a total change in expected runs at or below zero.

While the calculation shown here only explicitly considers three outcomes (the runner is out, the runner is safe, the runner advances on an error), as discussed above, the pickoff throw also serves to mitigate the runner’s ability to take an extra base on a ball put in play, which is an outcome also worth considering. But since the run expectancy table takes into account the current scoring environment (which is up about a quarter of a run over last year), any additional scoring attributable to more aggressive baserunning via both potential steals and extra bases is inherently folded into the estimate of expected runs. Thus, the increase in run expectancy associated with an error naturally accounts for this year’s slightly increased likelihood of a second-to-home scenario.

Now to the actual calculation. Consider the break-even calculus on the decision to throw for both the pitcher and catcher in a one out, runner-on-first situation. Going into this scenario, the expected runs scored sits at 0.521. In the event of an out, that number drops by 0.417 runs, and in the event of an error with the runner advancing to second, that number goes up by 0.157 runs. If the runner retreats safely, the run expectancy doesn’t change, zeroing out the impact of that outcome in the equation. Therefore, from a defensive perspective we get the following formula:

Pout * (-0.417) + Perror * (0.157) ≤ 0

For the numbers to balance, pitchers need to get an out more than 0.28% of the time, assuming the historical error rate, while catchers only need to convert outs greater than 1.57% of the time under the same assumption — both very reasonable goals. While assuming historical out conversion rates, pitcher pickoffs need to result in errors less than 3.85% of the time, while catchers need to stay under the 8.85% mark. Plenty of wiggle room in both cases.

While pitchers must preserve their warning throws, the numbers suggest catchers could be throwing behind runners significantly more in the name of keeping runners on edge.The probability of a runner attempting and successfully stealing a base following a pickoff throw sits at 16% this season, but those odds drop to 5% following a back-pick. There’s certainly some selection bias exaggerating the difference (i.e. which catchers are throwing behind which runners), but it’s a large enough difference to suggest a reasonable complement to the nail-studded bat.

As presented here, the back-pick strategy lacks the specificity required to make it truly actionable. We’ve basically only learned that catchers should consider throwing behind the runner more. But a team wouldn’t need to mess around with league average pickoff numbers, or limit themselves to one or two game states. Instead they might calibrate the break-even probabilities to their rostered pitchers and catchers, or home brew their own, more granular run expectancies to consider the baserunning tendencies of upcoming opponents, or derive the run values associated with the throw-over events themselves to get a more precise measure of their impact.

Then, after carefully curating the data, the coaching staff can craft a list of specific recommendations for throwing behind the runner, a custom playlist of comfort songs for pitchers and catchers to defend themselves. Because the new rules have flung open the gates and hitters are running on through.





Kiri lives in the PNW while contributing part-time to FanGraphs and working full-time as a data scientist. She spent 5 years working as an analyst for multiple MLB organizations. You can find her on Twitter @technical_K0.

19 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
wagnerhofmember
9 months ago

This was fantastic. Terrific insights and wisdom that go beyond baseball. As a Stranger Things fan, it was great to learn more of the meaning behind the Kate Bush song as well, and loved the Stranger Things references throughout. Thank you for this.