Over the past three years, the Detroit Tigers have attempted to steal a base 206 times, and were successful on 143 of these attempts. Both of these figures put the Tigers last in baseball, and it’s not even close. Over that time frame, the average major league team has stolen 307 bases in 420 attempts. In 2013 alone, two teams (the Royals and the Rangers) surpassed the Tigers’ stolen base total from the past three years combined, with a third team (the Brewers) falling one steal shy. I could go on, but the point is pretty clear: the Detroit Tigers have preferred not to run.
And who could blame them? On a team where Miguel Cabrera has been just the third-slowest regular starter, it might not be the best idea to put the pedal to the metal. And while the team has had the lowest Ultimate Baserunning (-25.9 runs) and second-worst Bill James Speed Score over the past three years, the Tigers managed to finish fourth-to-last in runs gained from stolen bases (-14.3 runs). Two of the teams behind them, the Diamondbacks and Pirates, have attempted more than twice as many stolen bases as the Tigers in that span, but their low success rate has cost them precious runs. Given the team the Tigers have put on the field the past few years, most would agree that limiting the stolen base attempts was probably beneficial to the team.
However, the Tigers’ running game could be receiving a boost this season, in the form of both a coaching and personnel change. New skipper Brad Ausmus has been very open about his desire to make the running game a bigger part of the team, going as far as saying that “everyone has the green light.” Of course, that’s a lot easier when your team trades away Prince Fielder for Ian Kinsler and adds Rajai Davis, one of just two players to average over 40 stolen bases per year over the last five years (along with Michael Bourn).
With spring training in the books, it appears that Ausmus plans on keeping his word, as the Tigers led the league in stolen bases and attempts. While this could be a sign of things to come, it’s also easy to dismiss. After all, most of us are conditioned to believe that spring training numbers aren’t predictive (other than pitcher velocity, perhaps). The samples are tiny and the level of competition varies wildly, so there’s nothing we can gain by looking at spring training numbers.
However, unlike most of the other statistics, stolen base attempts are something that remain almost completely under the control of the manager and/or player performing the action. No need to worry about quality of opposition or BABIP. There’s a runner on with a base open. Do you send him or not? This simple yes or no decision is the reason why if there’s any statistic that could have some meaning in spring training, it might be team stolen base attempts.
A bit of background before I get into the numbers. To adjust for the changes in baserunning trends across baseball, I took all the metrics I used and calculated a z-score. This means that the metrics are measured relative to the league average and standard deviation for a given season (or spring training season). With that being said, let’s look at how various stolen base metrics correlate year-to-year over the past five years.
|Year to Year Stolen Base Correlations, 2009-2013|
|Metric||Y2Y Regular Season Correlation||ST and Regular Season||ST and Previous Regular Season|
- SB = Number of successful stolen bases
- Attempt = Total number of stolen base attempts (SB+CS)
- Attempt% = Attempts divided by 1B + BB
- Attempt/G = Attempts divided by games played (which varies in spring training)
- Success% = SB / Attempt
*Attempt% uses 1B+BB as a surrogate for “stolen base opportunities”, a number which is not tracked during spring training. However, over the five-year sample used here, there was a 0.89 correlation between 1B+BB and stolen base opportunities (SBO from baseball-reference).
Looking at this chart, a couple things stand out. First, success rate is the least predictable, with a year-to-year correlation of .317. On the other hand, all the stolen base metrics have a fairly high year-to-year correlation, with Attempt% being the highest. The next column shows that while steals and attempts in spring training are somewhat predictive of the upcoming season, we’re slightly better off looking at the numbers from the previous season.
This might be where our study ends if we don’t look at the final column, which tells us that the correlation between spring training numbers and the previous regular season is actually quite low. So, while last year may be a better predictor of the new season on its own, there might also be some useful information in spring training.
While the Attempt and Attempt/G metrics were slightly better for spring training, they are both influenced significantly by team offense and number of opportunities, so moving forward I will be focusing on the Attempt% and total SB, along with Success% (as a sort of negative control).
If the spring training numbers are meaningful, then we should be able to include them in a linear regression to improve our ability to predict the 2014 season. As shown in the table above, on their own, the previous year’s Attempt% had a correlation coefficient of .512, while spring training came in at .374. What happens when we combine the two?
When accounting for both the previous year and spring training, the correlation jumps from 0.512 to 0.593, which may not seem like a huge improvement but is fairly significant. Notably, while the model places more weight on the previous regular season (as expected given the higher correlation), spring training still factors in significantly (roughly 40 percent). Similarly, factoring in spring training improves our prediction of total stolen bases (r = .523, up from .459), but not success rate (r = .320). Also of note, Attempt% can be used to predict future total stolen bases just as well as the steals themselves (r = .537).
So, it looks like we may have found some useful spring training statistics after all. While the previous year’s numbers are more valuable, looking at whether teams are giving the green light in spring training can help us improve our ability to predict the upcoming season.
But what about the Tigers? Is there a way to predict teams that are poised for a break-out year on the basepaths?
To answer this question, I calculated the change in z-score between the previous regular season and spring training, then looked at whether those changes signaled an impending regular-season change. I focused on Attempt% over the past five years, breaking teams into three groups: teams with an increase of at least one standard deviation, teams with a decrease of at least one standard deviation, and then all the teams in the middle with little to no change.
|Changes in stolen base metrics|
|Spring Training||Regular season|
|< -1 StDev||33||-1.66||-0.73||-0.73||-0.48|
|> +1 StDev||32||1.69||0.83||0.83||0.50|
While the teams that increased their stolen base attempts from the previous year in spring training didn’t hold onto all of their gains, they still showed a significant increase during the regular season and increased their total number of attempts by 22 percent. They also showed a slight increase in success rate, which could be an indicator of improved talent (which I’ll discuss shortly).
This group of stolen base surgers includes 34 teams over the past five seasons, and they have some shared characteristics. As you may have guessed, these teams didn’t steal a lot of bases the previous season, with an average zAttempt% of -0.723. These teams also had below-average success rates, which could have influenced the low attempt rate.
Of course, it makes sense that the teams with the lowest stolen base attempts have the most room to improve. Sure enough, there’s a significant negative correlation (-0.450) between the previous year’s Attempt% and the increase in Attempt% seen the following season. For teams that are already running a fair amount, it should be more difficult to find additional opportunities without sacrificing success rate. So, what if we break down this group a bit further into teams that were close to average in Attempt% (within 0.5 standard deviations of average) and teams that simply didn’t run very much (at least 0.5 standard deviations below average).
|Changes in stolen base metrics, by previous season Attempt%|
|Spring Training||Regular season|
|Previous Season zAttempt%||Teams||ΔzAttempt%||ΔzAttempt%||ΔzSB||ΔzSuccess%|
|Didn’t Run (zAttempt% < -0.5)||21||1.77||1.12||1.02||0.56|
|Average (zAttempt% > -0.5)||11||1.53||0.27||0.46||0.42|
So, here we see that teams that already had an average number of steal attempts didn’t attempt many more steals in the following season, even if they were running wild in the spring. They did, however, see a slightly increase in success rate.
On the other hand, the teams that didn’t attempt many steals the previous year but started running more in the spring were likely to carry much of that increase into the regular season. On average, this group’s total stolen base attempts jumped from 101 up to 135, a 34 percent increase.
This is all very interesting, but we still haven’t addressed the issue of changing personnel, which is clearly a driving factor for many of these teams. In 2011, rookie Ben Revere amassed 34 stolen bases in 43 attempts, accounting for 37 percent of the Twins’ steals in one third of the team’s attempts. Rajai Davis, one of the new members of the Tigers, chipped in 34 stolen bases in 35 attempts for the Toronto Blue Jays in 2011 after being traded from the Athletics in the offseason.
While there are certainly cases where one player can skew a team’s baserunning statistics, how often was that the case for the above group of teams? Eight of these 21 teams had a new player join the team and attempt at least 25 steals (including Brett Gardner returning from injury for the 2013 Yankees).
|Changes in stolen base metrics, by personnel changes|
|Spring Training||Regular season|
|New player w/ 25+ Attempts||8||1.72||1.69||0.64||1.59|
|No new player w/ 25+ Attempts||13||1.80||0.77||0.9||0.06|
It shouldn’t be surprising that teams that added a big-time base stealer saw the biggest increase in Attempt% the following season, along with a big boost in success rate. If you take these teams out of the group, the remaining teams still averaged an increase in Attempt% by 0.77 standard deviations ( 28 attempts). And while there may be some personnel changes, a lot of the increase is with the pieces that were already in place from the year before. It should be noted that four of the 13 teams in the second group had managers who were either brand new or starting their first full season as skipper (the Cardinals, Cubs and Red Sox in 2012, and the Nationals in 2010).
Also importantly, the increase in stolen base numbers is not simply regression to the mean. Teams that didn’t run a lot (zAttempt% under -0.5) and didn’t increase their attempt rate during spring training by at least one standard deviation barely increased their Attempt% during the regular season (+0.124 standard deviations).
Looking at the 2014 spring training numbers, which teams might be looking for a stolen-base breakout this season? Four teams match the criteria described above (2013 zAttempt% < -0.5, and an increase in spring training by at least one standard deviation), and the Tigers top the list. While much of this change is personnel-driven, Ausmus hasn’t been shy about sending players who barely ran last year, like Torii Hunter (five attempts in 2013, six attempts this spring) and Nick Castellanos (five attempts in Triple-A in 2013, four attempts this spring).
Interestingly, two of the other three teams (the Seattle Mariners and Cincinnati Reds) also have new managers this season, although they haven’t been quite as vocal as Ausmus about their willingness to give the green light. While the Reds’ success was driven by a single player (Billy Hamilton, with nine steals), the Mariners’ attempts were spread around more evenly, with no player having more than four. The St. Louis Cardinals round out the quartet, as Mike Matheny might try to replicate the success his team saw when he took over in 2012. (The Chicago Cubs and Washington Nationals, the other two teams with new managers, both ranked in the top eight in increase in zAttempt%, but missed the first cutoff as their teams were average runners in 2013.)
So, what does this mean for the Detroit Tigers and other teams that haven’t run much in the past but might be giving the green light this spring? First, if the team significantly increases its attempt rate during spring training, it’s likely to carry some of that increase into the regular season. Of course, the increase is more likely to be maintained in the regular season if it is driven in part by an influx of baserunning talent.
While sending Rajai Davis may be a no-brainer, do teams really benefit by giving the green light more frequently to players who haven’t run as much in the past? The 13 teams above that increased their attempt rate without adding a big base stealer didn’t hurt their success rate, but it also didn’t really improve and remained below average. On average, these teams attempted 20 extra steals the following season at a 71 percent success rate.
In terms of linear weights, these teams increased their wSB by 1.22 runs. Given what we know about the cost of a win on the free agent market, this translates into a value of $900,000, which isn’t a very large number in baseball terms, but it’s also not nothing.
More importantly, these numbers don’t take any of the other aspects of the game into account. Sending runners more frequently keeps the pitcher on his toes, and can influence defensive positioning and BABIP. In a low-scoring game with an ace on the mound, a single stolen base can mean quite a bit more.
All in all, giving runners the green light won’t have a very big impact on team performance, unless you’re sending new guys who are burners. When a team acquires a new player who is a talented baserunner, his manager would be wise to give him the green light, even if he hasn’t done so very frequently in the past.
Even without significant personnel changes, teams that haven’t run very much in the past might be able to pick up one or two runs by taking calculated risks. However, it’s important to remember how important success rate is to determining value from stolen bases. While some teams, such as the 2012 Cardinals, were able to run more with most of the same players they had in 2011 and pick up six extra runs on the bases, other teams that attempted more steals also saw a decrease in success rate and lost several runs (including the 2013 Tigers).
In stolen base metrics, we have a rare set of numbers that can provide meaning from spring training. The research suggests that the Tigers are ready to turn up the heat on the basepaths in 2014, despite their abysmal numbers over the past few years. And while they are poised to have some significant gains in their baserunning, the numbers suggest that most of this value will come from their new players, not their new manager.