On average, how many games do you need to attend before you see a stolen base? It seems like Billy Hamilton steals a few every night, so for Reds fans, the stolen base is about as sure of a bet as Pete Rose. Yet if you look closely, the stolen base is slowly fading from the game: this year there will be the fewest stolen bases per game since 2005. In 2015, if you randomly attended a game, there is about a one in twenty five chance you will see a successful stolen base. What is driving this decline in thievery? Are players getting slower? A newfound morality regarding the sanctity of the base progression? Tighter pants?

From fangraphs, it is fairly easy to show the decline in stolen bases per game. Here are the number for the last fifteen years:

Year Games SB SB/gm
2000 62068 2924 0.047
2001 61325 3103 0.051
2002 61602 2750 0.045
2003 61148 2573 0.042
2004 61111 2589 0.042
2005 59935 2565 0.043
2006 60495 2767 0.046
2007 60785 2918 0.048
2008 60759 2799 0.046
2009 59989 2970 0.050
2010 59813 2959 0.049
2011 58099 3279 0.056
2012 59122 3229 0.055
2013 58596 2693 0.046
2014 58819 2764 0.047
2015 55913 2403 0.043

Graphically, these numbers are a bit easier to digest:


What is interesting is that teams began stealing more after the steroid era than they did before. If this trend was rationally driven it could be explained that teams did not want to risk one out for one base when they had sluggers up and down the lineup.

Yet the number of stolen bases per game peaks in 2011 and quickly collapses. Part of this may due to the decline in the number of base runners. As we have extensively written about in the past, from 2010 onward, baseball could be called “revenge of the pitcher”. A slight push back against this is that the relationship between OBP and number of stolen bases is zero (r^2=0.001).

One common explanation for the decline in stolen bases is that analytically oriented teams don’t try to steal many bases. This argument states that prior to SABRmetric thinking, teams were both overvaluing the benefit of moving a runner one base while also downplaying the risk of that runner getting thrown out.

The run expectancy charts are brutal on stolen bases: with zero out, if a runner moves from first to second a team’s run expectancy only increases by about 0.20 runs. Yet if that team loses both a runner and creates an out, the drawback is 0.60 runs (from 0.85 to 0.25). Or put another way, teams would be no better off stealing three bases if they are thrown out once.

If a team can diminish the risk of getting thrown out, then the benefit of moving up a base becomes more reasonable. Due to the small benefit of moving up but deep drop off should the runner get caught stealing most methods predict the “break even” point to be about a 75% success rate. And that’s just to get back to a point where you would would be no better off than before.

If you are a normal reader of Redleg Nation, none of sounds especially surprising. The next thing you would expect is a distribution of stolen base percentages over the last 5 years that shows teams are increasingly adopting this “break even point” and that’s explaining the drop off in stolen bases. This was supposed to be a nice, easy 500 word post for the last Friday of the year. I downloaded the data from Fangraphs. And that’s when this column started getting weird.

Here is the stolen base percentage (stolen bases/[stolen bases+caught stealing]) for all teams in MLB over the past five years (grouped by year):


Green diamonds are 95% confidence intervals

Not only has the league never stolen at a 75% success rate, it doesn’t look like teams are making a discernible effort to become more successful than in the past. For all the talk about MLB front offices turning into computer labs, 2015 is the first time in the past five years that the stolen base percentage will fall below 70%. While for statistical reasons we cannot reject the idea that 2015 is any different than the other years, we also can’t say there is a trend in these data, either.

Yet maybe SABR team’s are not trying to steal that many bases. Stolen base % is declining, so these analytically-oriented front offices are keeping runners parked at first; its a perfectly consistent argument. So here is the same data by team:



SBA is stolen base attempts, same calculation as above. Green triangles are the 95% confidence interval around the mean (green horizontal line). Five years of data 2010-2015.

That’s a lot of data there to digest, but the top five teams that attempt to steal the most bases are: the Rays, Astros, Rangers, Padres and Royals. These five teams were labeled either “All in” or “Believers” when ESPN ranked the most analytically oriented baseball teams. The bottom five: the Braves, Cubs, Cardinals, O’s, and Tigers. Strangely,this is a more mixed bag than the top of the charts due to the Cubs and Cardinals being “All-in” on analytics while the Braves and the Tigers are “skeptics”.

Keep in mind the limitations of these data: it could very well be that regardless of front office philosophy, some teams just have better base stealers than others. This is certainly true in the short run, but over a five year span, player selection and development becomes more reflective of front office ideology than it does when talent is a fixed aspect of a club. Five years might not be enough for talent to be that fluid, so this is something to keep an eye on over the next few years. Another possibility could be that some teams are just better at getting on base and therefore create more opportunities for stolen bases. I tried to control for this by adding in team OBP over this span, but this is an imperfect control because high OBP players might be terrible base stealers (see: Dunn, Adam).

These numbers raise more questions than answers. On the one hand, it seems that analytically sophisticated teams do not steal more or less frequently than the “traditional” baseball clubs. Yet on the other hand, we see the Astros, Rays, and Padres all grouped together at the top of the chart. If anything, there are more analytically oriented teams in the top half of the distribution than there are in the bottom half. This poses an alternative narrative: traditional front offices are driving the decline in stolen bases. Why? Because traditional front offices have their “stolen base guys” and their “sluggers”. As the league OBP declines, “stolen base guys” get fewer opportunities to steal bases. Yet analytically oriented teams might make decisions about when the steal based off of other factors, such as if a pitcher-catcher combo has a low caught stealing percentage or the club’s ability to predict when pitchers are most likely to throw a breaking pitch.

I have no data to support this theory, but its becoming increasingly unlikely we can say that the decline in stolen bases is due to ‘SABR’ front offices dropping anchor at first. The overall trend is there: stolen bases are declining and we need to find a more nuanced explanation for this trend.

Rest assured Nation, we know that the Reds analytically-savvy front office will soon have an answer. That is, as soon as Walt gets off the phone so they can dial into their American Online account.

8 Responses

  1. gaffer

    Clearly you can gain value by baserunning (both real and perceived) but SB numbers are not a good way to measure that. Some guys get cheap SB in blow outs or just when there is a situation with no real value added. SB% is OK, but there are clearly runners who do not steal much but are excellent at advancing and getting that extra base and not baserunning gaffs. Heck, the coaching is a part too. I am certain the Reds gave up 3-5 games getting thrown out at home by Steve Smith!

  2. cfd3000

    Mike – I agree that there are at least as many questions as answers in this data but I couldn’t get past some obvious errors here. Does it really seem right to you that you’d need to watch 25 games to see a successful stolen base? It’s not! Think about it – Hamilton stole one on average every three games this year. One guy. Where did the data come from? 62,000 games a year? How about 2,480 (give or take)? 30 teams, 81 home games a year. So that’s about 1.0 steal per game, right? I’m fascinated by this question of whether attempted steals add value but this data looks really off to me and as a result not very informative. Am I missing something? Email me at Cfdeblois on my yahoo account if you want to communicate directly. And either way thanks for delving into this interesting question.

  3. George Mirones

    Let me offer this thought, The stolen base becomes a weapon when the opposing catcher doesn’t make an accurate throw. Knowing the number of times a runner actually advances past second base to third due to an errant throw, that ends up in centerfield, by the catcher might change the calculation and result. Just a thought.

  4. WVRedlegs

    That is a stunning dropoff from 2011. Over a 25% reduction in just a few short years.
    The stolen base can still be an effective weapon. It seems now that teams have to be more choosy on when to send runners. If a pitcher is bad at holding runners, then run all day. Like Chapman not even looking at a runner on 2nd base. The same with bad arm catchers, run like gazelles on them. But with the likes of Molina and the Russell Martins, or pitchers with good moves, just put those SB’s in your pocket and save them for another day. Pick and choose wisely when employing the SB.

  5. WVRedlegs

    Maybe WJ is a genius to go with America Online. It is so antiquated that the Cardinals, the Russians and the Chinese cannot hack into their system.

  6. Jeremy Conley

    These are all interesting questions, but I really don’t see any trends in the data provided. It looks like stolen bases went up, then down, then up, then down again, which seems a lot like random variation. The point where we’re at now isn’t even the lowest it’s been in the last 15 years, so even the premise that stolen bases are going down seems suspect, considering they’ve gone up since 2004.

    Some of the variation in SB/game could be caused by the league having more and less baserunners, as pointed out in the article, which could make the variation presented look even less like a trend.

    But to address the idea of “SABR” teams, I think basically the idea that they would steal less came from Moneyball, which depicted Billy Beane as being staunchly anti-steal. But really, any analyst worth their pay would tell you that if you can steal above the break-even line (which is actually lower than 75% in most situations http://www.fangraphs.com/blogs/breaking-down-stolen-base-break-even-points/), then you should steal as much as possible.

    So what I would like to see is the break down of “SABR” teams not by stolen base attempts, but by stolen base percentage. What I would expect is that teams run by old-school front offices and managed by guys like Dusty Baker would pay less attention to their success rate than more analytical teams, and likely have a lower stolen base percentage because of it.

  7. Craig

    It seems like base stealing would be one of the easiest areas to improve upon using data and analytics. If you know the amount of time it takes a guy to get from 1st to 2nd base and you know the amount of time on average it takes a given pitcher catcher combo (from the beginning of the wind up to when the catcher’s throw reaches 2nd base) then you should be able to predict that a guy’s success rate in stealing. There would also be a sliding scale of success based on the type of lead the runner is able to establish and the probability of an off-speed pitch, low pitch, inside, etc. So data, telling Brandon Phillips that with a 6 ft lead will be able to steal 2nd base 90% of the time on an off-speed pitch against today’s pitcher or whatever….it just seems like saying that instead of using sabremetrics to ask whether or not to steal, clubs should be asking how they can use sabremetrics to decide when to steal.

  8. JRAL

    I think it primarily boils down to roster composition. If you have fast players you are more successful at stealing bases and therefore you steal more. If you have Billy Hamilton, you steal more. I think the post by CRAIG is spot on as to how sabremetrics should be used.