The Changing Caught-Stealing Calculus
Leafing through an old Sports Illustrated, I recently happened upon this stellar article by Mr. Albert Chen entitled “Revenge Of The Base Stealers,” in which Chen analyzed the league’s continued shift towards base-pilfering over base-trotting.
With the whimper-death of the Steroid Era, league strategies have swung towards old-school baseball. Most winning teams now employ some combination of great defense, strong base-runningitudes, and notable pitching-miraculosities. As such, wise teams have found employs for otherwise marginalized speedsters.
The net result has been an uptick in the value of a stolen base, according to linear weights:
This chart shows how the cost of a caught stealing (the red line) is trending towards zero (meaning a caught stealing is costing less — in fact, much than its .400 runs high point in 2000) while the gains from a stolen base (0.161 runs in 2012) have remained strong.
Whither belongs the blame for this change? Simply: Home runs. And where do we wander from here? In short: Deep into the heart of Speedster Kingdom.
In 2000, during the apex of the stolen base depression, teams needed to maintain a 69.7% success rate on the base paths just to simply break even on the run value game (granted: game context — such as the pitchers, stadium, weather, and score — allows for varying levels of acceptable SB-rates). But by 2012, the break even point dipped to 66.6%.
Let us put that in real terms: A 2012 speedster can succeed with, out of 50 SB attempts, about 33 SB and 17 CS. Mr. 2000 speedster needed a balance of 35 SB and 15 CS. These differences seem small on the individual scale, but for a manager setting strategy across an entire roster this difference can have a considerable impact.
Before we expand that concept, we must understand why the run values for CS and SB have changed. In the height of the Steroid Era, home runs came in discount baskets. In the aforementioned 2000 seasons, the HR/PA rate nearly hit 3%. Compared to the 2.68% of 2012. That comes to a difference of almost a whole home run per five or six games.
The higher home run rate meant getting on base ended more often with a tater trot than in present seasons. Now, more teams are sliding across home plate as singles, doubles, and the odd triple must replace the extra weekly dinger.
The relationship between home run rate and the value of a stolen base are indeed quite related. Let’s examine at the break-even value for SB success, defined as the absolute run value of a CS divided by the absolute run value of a CS plus the value of a SB.
Unsurprisingly, the relationship between homers and the run values of steals are considerably and inversely related: The break-even rate for SB success shares a .806 .685 R-squared with HR rates, when looking at MLB seasons from 1950 through 2012.
Moreover, we find this:
A simple equation of:
Break Even Rate = 0.590 + 3.33 x (HR/PA)
This phat function allows us to predict the optimal break even point not just for a league, not just for a team, but for a lineup.
Take, for example, the San Francisco Giants. They do not hit homers. The led the league in not hitting homers. They are so superb at not hitting homers, they almost did not crack 100 dingers in 2012. They were the only team under 2.00% HR/PA. They posted a 1.72% HR/PA — excluding pitchers. With pitchers, they hit only 1.66% HR/PA — that’s a mere 3 to 3.5 total homers per week.
According to our handy Break Even Rate regression, we can predict the Giants break even SB rate in 2012 was 64.53% — well beneath the league average. Meanwhile, the homertastic Yankees (3.93% HR/PA) had a break even rate near 72.09%. So are teams being too aggressive or passive?
2012 Teams
| Team | PA | HR | SB | CS | HR/PA | actual SB success% | break even SB success | net |
|---|---|---|---|---|---|---|---|---|
| Pirates | 6014 | 170 | 73 | 52 | 2.83% | 58.40% | 68.41% | -10.01% |
| Orioles | 6160 | 214 | 58 | 29 | 3.47% | 66.67% | 70.57% | -3.90% |
| Diamondbacks | 6150 | 165 | 93 | 51 | 2.68% | 64.58% | 67.93% | -3.35% |
| Rangers | 6216 | 200 | 91 | 44 | 3.22% | 67.41% | 69.71% | -2.31% |
| Mets | 6091 | 139 | 79 | 38 | 2.28% | 67.52% | 66.60% | 0.92% |
| Cubs | 5967 | 137 | 94 | 45 | 2.30% | 67.63% | 66.65% | 0.98% |
| White Sox | 6111 | 211 | 109 | 43 | 3.45% | 71.71% | 70.50% | 1.21% |
| Astros | 6014 | 146 | 105 | 46 | 2.43% | 69.54% | 67.08% | 2.45% |
| Rockies | 6183 | 166 | 100 | 40 | 2.68% | 71.43% | 67.94% | 3.49% |
| Cardinals | 6326 | 159 | 91 | 37 | 2.51% | 71.09% | 67.37% | 3.72% |
| Tigers | 6119 | 163 | 59 | 23 | 2.66% | 71.95% | 67.87% | 4.08% |
| Dodgers | 6091 | 116 | 104 | 44 | 1.90% | 70.27% | 65.34% | 4.93% |
| Indians | 6195 | 136 | 110 | 44 | 2.20% | 71.43% | 66.31% | 5.12% |
| Blue Jays | 6094 | 198 | 123 | 41 | 3.25% | 75.00% | 69.82% | 5.18% |
| Yankees | 6231 | 245 | 93 | 27 | 3.93% | 77.50% | 72.09% | 5.41% |
| Nationals | 6221 | 194 | 105 | 35 | 3.12% | 75.00% | 69.38% | 5.62% |
| Rays | 6105 | 175 | 134 | 44 | 2.87% | 75.28% | 68.55% | 6.74% |
| Mariners | 6057 | 149 | 104 | 35 | 2.46% | 74.82% | 67.19% | 7.63% |
| Red Sox | 6166 | 165 | 97 | 31 | 2.68% | 75.78% | 67.91% | 7.87% |
| Reds | 6115 | 172 | 87 | 27 | 2.81% | 76.32% | 68.37% | 7.95% |
| Braves | 6126 | 149 | 101 | 32 | 2.43% | 75.94% | 67.10% | 8.84% |
| Athletics | 6183 | 195 | 122 | 32 | 3.15% | 79.22% | 69.50% | 9.72% |
| Brewers | 6224 | 202 | 158 | 39 | 3.25% | 80.20% | 69.81% | 10.40% |
| Giants | 6200 | 103 | 118 | 39 | 1.66% | 75.16% | 64.53% | 10.63% |
| Angels | 6121 | 187 | 134 | 33 | 3.06% | 80.24% | 69.17% | 11.07% |
| Padres | 6112 | 121 | 155 | 46 | 1.98% | 77.11% | 65.59% | 11.52% |
| Royals | 6149 | 131 | 132 | 38 | 2.13% | 77.65% | 66.09% | 11.55% |
| Marlins | 6057 | 137 | 149 | 41 | 2.26% | 78.42% | 66.53% | 11.89% |
| Twins | 6209 | 131 | 135 | 37 | 2.11% | 78.49% | 66.03% | 12.46% |
| Phillies | 6172 | 158 | 116 | 23 | 2.56% | 83.45% | 67.52% | 15.93% |
This data suggests the Pirates — blasphemous on the base paths by league standards — may not have been just as apocalyptically bad as they may have otherwise seemed. Their paucity of home runs allowed for a lower success rate. But their ineptitude still went too far.
Meanwhile, the homer-happy Orioles — perhaps more homer-happy in hindsight than in projection systems — proved likewise overly cavalier on the paths despite a solid 66.67% success rate. They clapped homers with the consistency of club that needs a 70% success rate.
The Phillies and Twins, both teams who appeared judicious on the paths, were in fact cowardly, gutless, un-American — North or South — base runners. They left far more productivity rotting on the table than the Pirates threw into the festering waste bin that was their 2012 season.
Of course, just because teams should be willing to have lower SB totals does not mean the Giants should start sending Pablo Sandoval (Note, however, the relatively efficient and base-stealing fanatical Rays did get 3 SB and just 1 CS out of Jose Molina in 2012). Every manager needs to play with the intention of successful steals, but in general, teams need to accelerate the running game.
The acceptable SB success rate of the last three years (about 66%) is on par with the late 1980s. But in the early 1980s, teams were attempting steals in about 3% of all PA — compared to the 2.5% of modernity. That’s around 1 steal per week, though probably more, that teams should be attempting, just to match the 1980s’ figures.
The Twins led the American League with 135 steals, but they also appeared to miss out on potentially more runs by not sending the likes of Ben Revere, Alexi Casilla, Darin Mastroianni, and even Denard Span more often.
Here is a look at the stealing tendencies across the league:
The horizontal axis shows the steals per opportunity. The vertical axis shows the team’s steal success rate minus their break-even rate. The bubble sizes indicate the team’s total steals on the season.
The Padres led the league in attempt rates in 2012 (which is roughly SB+CS divided by SB opportunities), but since they still hovered 12% above their break-even success rate, they could have easily afforded an even more aggressive base running style. Instead of running on 9% of opportunities, they could have gone for 10% or 12% — running until opposing teams throw more pickoff moves than strikes.
Obviously, this analysis of the running game is crude. But the ideas welling underneath it point to a concentric theme: Managers need to buckle their courage pants and start beaming the green light again. Home runs may never come back in quite the same fashion, so stealing deserves its renascence.
Chen’s SI article ends where the modern discussion of speed oft ends: Billy Hamilton. The 22-year-old shortstop has golden legs and may be approaching the major leagues at just the right time to challenge Ricky Henderson’s filthy, filthy records. Young players like Hamilton have a chance to begin a new era of baseball, to replace the wood-cracking echos of the 1990s with the sounds of fabric once again sliding on clay, to rebirth the small ball era and make running cool again.
But they will only do such things, of course, if the managers do like my charts’ color schemes, and go green.
Great analysis. Are you strictly talking about swiping 2nd base? How does the calculus change when going from 2nd to 3rd, or 3rd to home?
This! And really, base-out states ought to considered too. I’ve never seen a full table of seasonal RE24 (or WPA) by team on base-steals.
Not to mention the inconsistent treatment of pickoffs, which ought to be part of these sorts of equations but end up being at the scorer’s discretion as to whether or not they are CS. And then there’s hit-and-runs, and missed squeezes, and so on.
Mets and Cubs: models of efficiency
Excellent article. I think the next step would be to rank break-even rate by position in the batting order. Since your highest HR rates are typically going to occupy the 3-5 spots in the order, I would expect that the break even rate for guys in the 1-2 spots are lower, but they would start rising as you go from 4-8.
I guess I’ve never really thought of it, but if you have a guy that has a high OBP, a high HR rate, and good SB speed (say, Ryan Braun)–instead of batting him 2nd or 3rd where you would traditionally place him, put him behind your slower boppers at #5 so that you can utilize his speed more in front of your singles/doubles guys at the bottom of the order. Would the RE advantages gained from extra bases outweigh the handful of extra plate appearances from the higher batting slot over the course of a season?
If you’re saying drop Braun in the batting order because he can steal bases, I disagree completely, only because you don’t want to take plate appearances away from your best hitter. However, the idea that the break even threshold will be lower for players batting AFTER the big HR hitters in the lineup makes sense. This makes it seem like speed guys should be batting lower in the lineup, or at least that the lower spots in the lineup should be attempting more steals. This doesn’t mean you have to drop a speedster (e.g. Trout) just because he’s fast, just that a batter at the top of the order might need a higher success rate to break even than if he were batting, say, 6th or 7th.
Well that is what I’m contemplating. Conventional wisdom is that you don’t sacrifice the additional plate appearances. But all late game plate appearances (the ones you would be adding) aren’t necessarily high leverage. The idea of moving him down to take better advantage of his speed is to make better use of every time he reaches first base. Including the fringe benefits, like the lesser hitters in the 6/7 spots getting more fastballs.
Two spots is a ways to move down though. That’s giving up roughly 30-40 PA’s per season.
Ian, please stop contemplating. Braun’s stolen bases should have no impact on moving him lower in the order. No impact at all.
However, moderate/below average on base guys with speed but no power (Juan Pierre back in the day, Michael Bourn…and eventually Billy Hamilton) *should* hit 6th/7th to minimize the risk taken on when he does run.
Bham running in front of Votto/Bruce? Makes me cringe. Bham running in front of Mesoraco/Cozart? Go for it, bro.
To this point, the table should reflect the success rate at which you stop attempting steals AT THE MARGIN. If Rollins is standing at first and he has >68% chance of success, he should steal; if not, he should stay put.
A team that is operating perfectly in terms of decision-making will undoubtedly see its success rate exceed the figures in this table. The Mets, for example, were definitely too aggressive, since they wound up with 0 value from SBs.
If the Mets wound up with 0 value from SBs, doesn’t that mean they were essentially the right amount of agressiveness? If they had negative value, they were too aggressive. If they had positive value, they should’ve run more often. With 0 value, it’s neither good nor bad.
We aren’t trying for indifference here, but runs. Teams should be maximizing (actual success rate – breakeven rate)*total attempts. Hitting the breakeven point is quite bad. Not Pirates-bad, but still not good.
well… we ARE talking about the Mets – I wouldn’t rule out their desire for indifference. (but your point is absolutely correct – on a discrete basis, it might be okay to say that we are solving for indifference on a play-by-play basis, but certainly those decisions should result in a cumulative positive result)
I think if you had a power/speed/low OBP guy it would make sense. But a high OBP is going to be more important in front of the other sluggers than the speed behind them.
I guess the guy they should be batting fifth is Carlos Gomez.
I’m a bit confused about one point. What inherently makes what the Phillies and Twins did “too passive?” By this logic, it feels like you’re implying that it’s best when the actual success rate equals the break even rate. If a team such as the Padres stole more frequently as suggested, there’s the faulty assumption that they’d steal bases at the same rate. Drastically increasing SB attempts just because you have “banked” surplus value to work with doesn’t seem wise. However, with this in mind, there is the idea of using your advantage. If SB success is a strength, then what is the optimal SB success rate above break-even (i.e., the rate at which you’re maximizing your attempts and your success)?
Maybe this is just my risk averse side speaking, but while I like parts of this article a lot, the implications about what to do with a break-even rate are confusing.
I have a similar question about how to reason with the break-even rate.
You pretty much answered your own question… the optimal success rate is the whole point of the article. He’s not arguing that the PHI or MIN should maintain their SB rate no matter how many attempts. That is the whole point. The could steal more and more until their rate of success decreased to their optimal breakeven rate. By having such a high rate of success, it suggests they are being too selective in their attempts, and not stealing enough.
No one is saying it is good to get CS more, but that even as their success rate declines with more attempts, they are still getting a net value from those additional bases.
If a team is stealing bases at their break even rate, they are getting exactly 0 value from stealing bases. If a team attempts 0 stolen bases, they have are getting 0 value. But somewhere in between, attempting with enough restraint to maintain a good SB%, they will add positive value. So the question is, where is the optimal point between never attempting to steal and attempting so often that your success rate reaches your break even rate?
This confused me a bit too. Maybe the thinking is that since their success rate was so high, they must have had the personell in place to steal bases successfully, and even if their success dropped a bit in the extra attempts, their break even point was so low that they would likely still accumulate value through their extra attempts.
That’s what I ultimately think, but this work does not include how to know if they’re getting the extra value because of those attempts, or when a team should stop doing that extra stealing.
So it’s like a suplly demand curve?
In the case of the Phillies, if they were stealing that succesfully, the point is that they should steal more. Then they’ve stolen tons more bases and presumably scored more runs by having more guys in scoring position, etc.
In theory, once they’ve ramped up the steal attempts, at that point they re-assess their SB success rate. If theyre still stealing that successfully, or even more successfully, with more attempts, than they’re golden. Maintaining that 16% differential is incredible if they’re stealing as much as possible. However, if they start stealing more and the success rate declines, that’s OK as long as the rate doesn’t get below the break even point.
That’s at least my understanding. Having a 16% differential isn’t inherently bad. But it means you should be stealing as much as possible because you’re so successful at it.
“They left far more productivity rotting on the table than the Pirates threw into the festering waste bin that was their 2012 season.”
I have trouble understanding how this follows from the statistical analysis. And neither of your responses justifies it. Maybe a little productivity was left on the table because we can safely assume that had they attempted more steals at least some of them would have been successful. But we can’t be sure how to project their success rates as they take more risks. And to turn the rates into productivity we need to know how runs correlate to number of attempts. In other words, we need an optimization formula before we can infer that the teams “left far more productivity rotting on the table than the Pirates threw into the festering waste bin.”
@LTG- Exactly!
This is exactly the idea here. I did not have enough time or space to fully flesh out this idea, but since even the most prolific base-stealing teams are running on only 10% of chances, there is reason to believe, teams with relatively effective running — like the Phillies — should be running even more.
That’s not to say a 0% net is the goal. The goal is perfect base running, obviously. But we have seen an uptick in base running effectiveness. Now it’s time to see an uptick in base running quantities.
Quoth LTG:
That is precisely the next step, in my opinion. Granted, I am looking at the SB/CS balance with a wide lens, and getting more granular (examining 2B and 3B steals; looking at running going, foul ball frequencies, etc.) will have its benefits, but first priority, I think, is to find the more literal, less hyperbolic “productivity rotting on the table.”
Thanks for your responses, but they skirt around the point. It seems to me you just should not have written that sentence about the Phillies and Twins productivity because it isn’t justified by anything in the article, yet appears to be justified by the break-even rate because you draw an analogy with the Pirates. The sentence is just confusing and should be stricken, even if, when the final analysis is complete, you are correct.
Otherwise, it’s a very interesting article and I look forward to the follow-ups.
I think it is fair to say that the Phillies should have stolen more – they are sub-league average in attempt rates, well-above league average in success rates and have a large surplus in success above the b/e rate. I agree with the consensus that work needs to be done on an optimization algorithm, but the Phils are just too much of an outlier….
Steven,
1) Check out Travis’s post below for reasons why it is not fair to say the Phillies should have attempted more SBs.
2) You still miss the point. The evidence in the article doesn’t support the conclusion about the Phillies. Even if the conclusion is true, it shouldn’t be there because the article misrepresents the inferential relations between the evidence and the conclusion.
Well the author is wrong about this. If you studied economics, you would know that they’re only leaving production on the table if the marginal value of an additional steal is positive.
An analogy would be you would want to find the profit maximizing rate, not the rate when your company breaks even.
If the actual success rate is only slightly higher than the break even success rate, you’re probably stealing too much.
Exactly….
The other problem is that stolen base capability is not uniformly dispersed among a team. If 2 or 3 good basestealers are nearly maxed out on attempts (from an “efficiency” perspective), it hardly makes sense for slower runners to run more just so the overall success rate drops to equal things out. You may pick up a few extra stolen bases but at that point it may have negative marginal value.
In short not all stolen base opportunities are created equal, yet they are being treated as one large predictable aggregate.
What would be the most “profitable” (optimal) stolen base phiolosophy then?
Bang! Bisu and Hank have nailed it. The important thing to know is the marginal value of each additional steal attempt, which will vary depending on the base stealer, pitcher, catcher, which base is being stolen, and the chance that someone behind him in the lineup will homer.
These are all known variables in most cases, so I think a formula could be worked out that takes all that into account and projects the marginal value for each steal attempt, which would tell you whether it makes sense to run or not. That would be a pretty cool tool.
The break even point is the point at which Stolen Bases stop producing positive run value. So if your break even point is 65% and you steal at an 80% success rate then you’re producing positive value. If you attempted another 100 SB and were successful 66% of the time then you were still producing positive run value. You’d have produced more run value than if you had stoppe at your previous 80% success rate. As long as your success rate is above the break even point you are producing positive run value. If you stop with your success rate far above your break even poin then you are producing less total value than if you had stolen more bases at a success rate marginally above the break even point.
What program did you use for your graphics here?
Google Drive. Would you believe it?
I was hoping to make them interactive, but that effort proved rich with problems. Hopefully Google will sort out some of their more complex embeddable problems.
Third to last paragraph, last word, should be renaissance?
Good article!
I’m not sure if it’s British spelling or what, but that’s what my spellchecker suggested.
http://www.thefreedictionary.com/renascence
Steal a bag for Lennay!
With regards to the Twins and their rates…..are these figures regressed? As in….should one have expected those rates to stay the same if the Twins had attempted 1.5x as many thefts? 2.0x?
Is regressing the right strategy? It should be done, yes, but to me there is a far more glaring factor when discussing whether teams should steal more.
Presumably, teams are far more likely to steal when given a good opportunity to succeed. This means they probably are doing a lot of their running on favorable counts, against pitchers who are slow to the plate, against catchers with bad arms, etc. Undoubtedly if they are to steal more, that would involve running in less favorable situations. This would undoubtedly lower success rate even if a team’s true talent success rate at their number of attempts is exactly what you see above.
Also, the more a team runs, the more the other team prepares for the run. Players who rarely run sometimes catch opposing teams with their pants down this way. See Jeff’s humorous article about pitcher steals.
What this article is missing, is some estimate of how many runs were are talking about, by being more aggressive. http://sportsillustrated.cnn.com/vault/article/magazine/MAG1125906/index.htm
Very interesting analysis.
However, giving major league managers a bit of credit, you’d have to think that – perhaps – the increased aggressiveness would come at a poor success rate. If the Royals’ high success rate came from aggressive but judicious baserunning then adding another 10 SB might take 20 extra attempts. The numbers are arbitrary but the point remains – I don’t think there’s added value in making more attempts just because you’ve got ‘outs to burn.’
That being said, it is interesting to note the correlation between HR rates and the value of a SB.
You beat me to it. Yeah, it seems likely to me that increased aggression would come at the cost of lower success rates, due to [probably] more attempts during more unfavorable opportunities to steal.
By how much? Who knows. Seems that figuring that out would take something like timing pitchers to home when there are runners on and estimating the likelihood of the particular runners stealing the bases, based on their past times as well as the catchers’. THAT would actually tell you how good the opportunities were, and how likely each was to have been a wasted one (of course, you’d also want to take lineup effects and outs into consideration, in determining the break-even).
I love the analysis, though, Bradley.
This is a fabulous and fascinating article. Even though I don’t understand the math, this is cool stuff and you have to wonder if MLB teams are doing analysis at this level.
What software did you use for the analysis? I like the output scheme.
All of it was done on Google Drive’s spreedsheet interface. Real basic stuff.
Here’s a little guide for how to look at the ideal attempt rate.
Say a team’s break even rate is 65%. Their success rate is 75%, or 150 steals and 50 CS. This team might be advised to steal more. Let’s say next year they attempt 20% more, so 40 more total attempts. Let’s say their success rate goes down to 70%, or 168 steals and 72 CS. Since they’re still stealing at a better rate than their break even rate, they’re still adding positive value. What is not clear on its face is whether they were better off with fewer attempts at a better rate. Obviously stealing more is good if you maintain a better-than-break-even rate. The question is what is the trade-off in value between more attempts and less success?
We do have an easy way to measure this though. With 200 attempts at 75% success, they’ve added a certain value, X. Those hypothetical 40 attempts that don’t happen add 0 value so their total value is X. Let’s say they try 40 more times. Those 40 SB attempts are independent from the original 200 let’s say, which may be effectively true. Attempting a steal doesn’t affect the value gained or not gained from the other 200 attempts. So we can keep that number X and simply add the value of the additional 40. Of those 40 steals 18 were successful and 22 were not. That’s a 45% success rate. Those attempts, also being independent, can be assigned a value Y, which is clearly negative. So the total value of the 240 attempts is X + Y, which is less than X. They were better off not trying those extra times.
So the point here is that an attempt rate is optimal when additional attempts will come at the break even rate. This is based on the assumption that the derivative of success rate with respect to SB attempts is uniformly negative, something I think is safe to assume, once you have a significant number of attempts.
It seems to be that the optimal number of steal attempts is where you attempt to steal any time doing so has positive marginal value — i.e., when the chance of stealing a base is greater than the break-even CS% for the particular situation. That approach would maximize the profitability of stealing, since you would only steal when doing so is profitable.
If you do this stealing analysis by taking a team average approach, the result is you get a suggestion that a team should steal more or less, but it does not tell you the who and when. The real answer is likely that they should steal more in certain situations, and less in others.
You make a good point in your first paragraph. I’m sort of assuming that teams do what you’re saying to some extent when I hypothesize that success rate will decline with number of attempts, and a large reason is that teams generally steal when they feel they have a good chance of being successful, and stealing more means stealing in situations where they have a smaller chance.
But depending on how precisely we can measure the chance of success, that would be the way to go. We wouldn’t even be able to measure the value of such a strategy the way I’m doing it though, because I’m assuming a constant break-even percentage. What we could do is average the break-even percentage for each state where a steal is attempted. It’s irrelevant to include data from PA’s where no steal is attempted, since non-attempts are value-neutral by designation. So what we’d have is a lower break even percentage, and since success rate is also maximized, we’d also have a higher success rate.
I got into baseball in the early 80’s. I grew up learning the game while guys like Henderson, Raines, & Coleman were running the deals. I LOVE SMALL BALL and STOLEN BASES! Dude! If we’re heading into an era where that’s more available than the 90’s slugfest (which I never really liked), than I am all for it. Gimme a modern pair of young 100-base stealers! Yes! Maybe I’m being too hopeful…. but we’re waaayyy overdue for someone to reach the 80 steal mark. I think the last time was in ’88. I’m still having withdrawals from the pre-Bash brothers baseball.
Bradley, I liked the concept of the different breakeven points by teams. I am sure that this could even can be applied by hitter. But my question is, How does the number of Outs change the Breakeven rate? I am sure it has too. While the Giants, as a team, need only a 64.53% BE point, I am certaing that has to vary based on who is at the plate, and the number of outs? Has this research been done? For example, whats the BE success rate when there is a runner on 1st and the choices are, Barry Zito is batting with 0 out, vs Buster posey batting with 2 out. Also the run differential could be a factor as well. I noticed a previous reply where you mentioned looking at this problem through a wide lens and feel you did a great job of it, is there any anlysis being done to look at in a more specific setting?
I think the ‘runs left on the table’ argument is a little overstated. Echoing earlier comments, it occurs to me that the breakeven rate will be highest for the first few batters in the order, since the batters while they’re on base will be most likely to hit a home run. However, most teams place their speedsters in the 1 and 2 spots in the batting order, at least to a significant extent. That means that the runners who take the most chances on the base paths are the very runners who need a higher breakeven point to gain value, and therefore the team’s stealing success rate, aggregated, should rightly sit higher than the lineup’s overall breakeven point calculated above.
This isn’t to say that there isn’t value to be gained by stealing more. But it’s a reason that the numbers probably aren’t as high as the table suggests. There’s also the angle of re-ordering the lineup to account for stolen base runs, but that requires other considerations as well. Game theory plays into it as well, as increasing the running game will result in tighter control of the running game, and previously high-percentage steal attempts will become less successful.
Great read for sure; thank you, Bradley. I enjoyed learning how the value of a SB is quite team dependent.
I’d love to see some breakdown of how often each team stole 2nd and stole 3rd, and with how many outs, and how that further impacts each team’s NET % based on the new run expectancy. Seems to me some of the numbers may change based on this data, no?
Good analysis but I just can’t get the image of “more teams … sliding across home plate as singles, doubles, and the odd triple” out of my head. What would that even look like?
I looked a little more at the Phillies stats since they are who I root for. I’d be more concerned if they were say last in SBs, they were 11th. So it’s not as if they were trying less than other teams. They had 4 principal base stealers: Juan Pierre, Jimmy Rollins, Shane Victorino, and Chase Utley. Only Jimmy Rollins was a starter for the whole year. Shane was traded at the deadline, Chase was injured for the first half, and Pierre didn’t start every game. Shane was 7th[w/ LAD too], Pierre 10th, Rollins 18th, so lets say they were mostly maxed out. Hardly anyone else stole any bases[Pete Orr has a relatively high steal attempt per time on base rate, but he wasn’t on base much as he didn’t play much]. So a couple of theories:
a) Rookie types either aren’t given the green light, or are afraid to make a mistake. It seems like players like Freddy Galvis, Dominic Brown should have tried to steal at some point, and yet they didn’t. Hell, Cole Hamels and Cliff Lee both have an attempted steal.
b) Alternatively, players that weren’t around when Davey Lopes was in charge of the running game are less apt to steal.
c) Players at the bottom of order are afraid to steal. On the one hand it would seem this is when there is the most upside to stealing. On the other hand, especially a rookie type might not be allowed to steal with 2 outs and the pitcher batting for instance.
Looking forward to 2013, 2 of their base stealers are gone, but one has been replaced with Revere who should have a green light and steal and decent amount. A full season of Chase Utley healthy might get them another 10 steals or so. But Michael Young and Ryan Howard are never going to steal a base, and I’m mildly surprised that Ruiz had 4[some of those must have been the pitcher not paying attention]. Ruff isn’t going to have any steals either if he earns significant playing time. So really that leaves them with Brown/Mayberry attempting some steals if they don’t want a large drop in their stolen base numbers[though just Revere/Rollins/Utley will get them 80-90 in all likelihood].
In my opinion the Phillies are probably right where you want to be. My early impression that it is better to have a great success rate in fewer attempts. Read my post above. I show that stealing at a rate 10% above your break even point is much better than attempting 20% more at 5% above. The Brewers stole 42 more bags, but I’m inclined to think the Phillies still got more overall values from their attempts.
Doesn’t this assume homogeneity of HR rates within each teams lineup? If a batter gets on first before the 7-8-9 batters come up, the HR expectations have to be much lower than before the 3-4-5 batters…
It assumes a lot of things. It’s the first in step in what can potentially be a very deep and complex analysis.
Great analysis. You could see the game changing in this direction so it’s no surprise that the value of good base stealers is on the rise. I always thought the good base stealer was improperly undervalued even during the steroid era. The statistics that devalued steals during the steroid era looked like they really focused on the + value of a steal vs. the – value of the caught stealing. I don’t think those statistics captured the + value of just having a feared base stealer standing on first(not sure where the “fear” line is drawn).
When certain base stealers get on 1b in a close game we have always seen the following 2 or 3 things consistently happen:
1.) One of the middle infielders pinches in toward 2nd base which is a different spot than they would normally play that batter.
2.) The pitcher often changes the mix and location of pitches normally thrown to the next batter to better hold the stealer at 1b and/or to give the catcher a better chance to throw him out.
3.) Some pitchers actually go to seldom used mechanics for getting the ball to the plate (the slide step or a quickened pace).
So to me in close games a feared base stealer has always changed the way the opposing team and pitcher want to do things. I’m not sure how you capture this effect or even if you have to. I think it goes without saying if you can cause a defense or pitcher to change what they would prefer to do, you have gained an advantage.
Click my name for a Google spreadsheet with net value and a plot of efficiency vs. value. The outlier on the right is the Phillies.
Net value is in units of SB, so one thing that’s clear is how little this means. I don’t have the precise lwts numbers and Bradley’s link is not helpful, so I’ll just leave it there, but the Pirates lost somewhere in the neighborhood of two runs and the Padres gained somewhere in the neighborhood of four.
Excellent. Thank you. If you don’t mind, I will probably use this as a model for my followup analysis.
No problem.
Wonderfully insightful and rigorous article, a good read.
Very interesting stuff. Where did you get the break-even rates for all the seasons from 1950-2012 to plot those points?