Stealing bases comes and goes as a strategy, usually in lockstep tandem with the difficulty in scoring runs. This chart shows runs per team from 1901-2011, as well as steal attempts (both steals and caught stealing):
Ignore the spike seen from 1912-1916--prior to that, caught stealing was not a statistical category, and from 1917-1919 was also not tabulated. The data from 1920 on reflects full data on steals and caught stealing.
The left axis depicts steal attempts per game and the right is runs per game. It's not a perfect correlation (nothing in baseball, let alone life, ever is), but there are clear demarcations in runs per game. Here's how I see it:
Dead Ball Era-1903-1919
Lively Ball Era-1920-1940 (roughly)
Mini-Dead Ball Era-1961-1968
"Enhanced" Offense Era-1995-2007
With this backdrop, I'll investigate the effectiveness of base stealing by reviewing play-by-play data available at baseball-reference.com (B-R). I was looking for several things:
1. Is base stealing more effective today than in the past? We can measure that fairly easily.
2. Is base stealing more prevalent today than in the past? Again, as the chart shows, we can measure that easily.
3. Is base stealing leading directly to runs scored? For that, we have to use play-by-play data, since looking at aggregate data tells us nothing in this regard. For example, Rickey Henderson stole 130 bases and was caught 42 times in 1982. How many times did those 130 steals lead directly to a run? How many of those steals were CRUCIAL in scoring that run? How many of those 42 caught-stealing cost the A's a run? These are the questions I set out to answer.
Buried in B-R's splits data is base running information that can show how the incidence of base stealing has changed over the years. The data goes back to 1948, and the data from 1948-1950 is not complete, but it still presents a pretty good overview of how teams approached steals. This first chart shows how often teams attempted to steal in that time span, defined as (SB+CS/SB Opportunities):
This is taken from a different data set, but it shows the same trend as the first chart--attempts to steal bases were creeping up throughout the 50's but had a definite increase around 1973, began a fairly steady decline around 1995 and has begun to trend back up from around 2006. This next chart will show stolen base percentages by base:
Stolen base percentage has risen steadily. The decrease in success of stealing home is based on very small numbers--for example, in 2011, runners attempted to steal 2nd 3,940 times, 3rd 551 times and home 49 times. All this data can be replicated with some know-how in about an hour. What follows is what requires play-by-play data, which allows us to see how necessary these stolen bases were to score runs.
To begin, some explanation is in order:
1. A player's stolen base is considered to have been necessary to have scored a run only if he scored the last run in the inning. For example, if a player steals 2B, is driven home by a single, and then the next player hits a home run, the stolen base is NOT considered necessary--the home run would have driven him in. It's total after-the-fact analysis, and I freely admit that.
2. A player's stolen base is considered necessary only if the plays after him would not have scored him. If a player steals second and the next player hits a triple, that stolen base is NOT considered necessary--the vast majority (we're talking over 90%) of players will score from first on a triple. This isn't fair to the slower players who aren't as adept at taking extra bases on hits, but they steal less anyway.
3. To recap the second point, here's how it works for each base
Stealing 2B--nothing beyond a single will allow for the steal to be considered necessary for the run to score.
Stealing 3B--only ground outs and sacrifice flies or bunts will allow the steal to be considered necessary.
When you look at it this way, it appears the cards are stacked against ANY steal being considered necessary, and the whole exercise could be written off as foolish. However, consider this fact--in 1950, around 30% of hits were extra-base hits, and for the typical player, ANY hit, let alone an extra-base one, will drive him home from second. That number had nudged up to 32% for 2011. What this exercise does more than anything else is illustrate how often a stolen base can create a run when there isn't a hit after the steal. With all that being said, here's how the chart breaks down for 1950, 2010 and 2011:
We can look at this two ways, and I'll describe the 1950 line to show how. 845 attempts to steal second base were made (477+368), and of the 477 successful steals, a run was scored 168 times, or about 35.2% of the time. However, after parsing play-by-play data, I determined that in 104 of those cases, the player who stole second would have scored ANYWAY due to subsequent events in the inning. I'll reiterate that I fully understand that this is total ex post facto reasoning and that the game isn't played in hindsight, but it goes to the whole thinking behind base stealing. Is the risk of making an out around 30% of the time when attempting to steal second counterbalanced by the fact that in 13.4% of the cases (64/477), a run WOULD NOT SCORE without that steal? That's the decision every manager has to make, and of course it will depend on game situation and who the runner is. In many ways, it also depends on the relative ease of scoring runs--in the mid-90's, stolen bases were emphasized less because the offensive surge gave more opportunities to drive the player in. If the game is beginning a new period of reduced runs (and it's still way too early to know), then base stealing will take on a greater emphasis.
As is apparent in the chart, stealing third is a different matter altogether. I'll admit to being puzzled by the data shown, because other data I haven't shown depicts that third is often stolen with 0 or 1 out, as would be completely expected. The part that surprises me is that third isn't stolen more often in those cases in order to set up a sacrifice, but that's a discussion for a different post.