Thanks Dan, nice analysis.  This should be required reading for anyone who opines about why the White Sox were a successful team this year.

BTW, in the other thread you said that the White Sox were more dependant on the home run than any other team in 2005.  Is this because they had the highest % of their runs scored on a home run?  Do you have a list of all teams or at least their team %?

BTW, in the other thread you said that the White Sox were more dependant on the home run than any other team in 2005. Is this because they had the highest % of their runs scored on a home run? Do you have a list of all teams or at least their team %?

I’m actually wrong - they’re only 2nd.  Looking at team XR and team XR without any plate appearance that was a home run, here’s the sorted list of teams and what percentage of their run production was from the homer.

Texas - 56%
White Sox - 49%
Cincinnati - 49%
Yankees - 48%
Cubs - 46%
Cleveland - 46%
Baltimore - 45%
Arizona - 45%
Mets - 44%
Atlanta - 43%
Milwaukee - 43%
Detroit - 42%
Houston - 41%
Boston - 41%
St. Louis - 40%
Dodgers - 39%
Oakland - 39%
Tampa Bay - 38%
Philadelphia - 38%
Colorado - 37%
Angels - 37%
Pittsburgh - 36%
Minnesota - 36%
Seattle - 35%
San Francisco - 35%
Kansas City - 34%
Toronto - 34%
San Diego - 33%
Washington - 32%
Florida - 31%

Interesting. Very incomplete of course. The running game has effects that are extraordinarily hard to evaluate. If a team did not run at all, how that would change defensive positioning, pitching patterns, etc., is very hard to know.

The running game has effects that are extraordinarily hard to evaluate.

Right.  I always find stuff like this very unsatisfying.  Having a stolen base threat on base changes the way the pitcher pitches and the way the defense defends, and we do not have a way to measure what effect those changes have.  So to say we understand the net effect of the stolen base is just wrong.

Doesn’t stop people from doing it though.

One way to try to account for those side-effects is by multiple regression. The regression in the book Curve Ball gives .19 runs for a SB and -.12 runs for a CS. The .19 is around the actual value of a SB, but the -.12 is different than the actual CS value of -.29 or so.

And one of the old Scoreboards had data on how hitters fared in at-bats with a stolen base attempt rather than otherwise.  Hitters hit quite a bit worse than usual when a stolen base is attempted.  Hitters also hit worse the more pickoff attempts are made.

If there are real, positive effects of stolen bases other than what is given by various formulae have to manifest themselves in runs at some point if they have any value.  And if they manifest themselves in runs, we have a plethora of mathematical tools to find them.

The White Sox did not score more runs than one would expect from the sabermetric stats - in fact, they scored less.  This isn’t a case of the 1985 Cardinals exceeding runs created by an unusually impressive amount and we have to find other reasons why.

Every shred of available objective evidence suggests that the White Sox, while a tremendous pitching and defensive team, were an inept offensive team that didn’t do much other than hit homers in a very homer-friendly park and that includes stealing bases.  Luckily for the pitchers, hitting home runs is the best thing for offenses to do.  Otherwise, the 2005 White Sox wouldn’t be curse-ending champions but the 1984 Pirates.

I agree with SF that this can’t be considered complete.  For one, it also matters quite a bit who’s pitching and who’s hitting. 

Honestly, from the White Sox games I watched, I find it unlikely the times were chosen all that carefully.  Podsednik in particular seemed to pay no attention whatsoever to game context.  I thought he was just an idiot with a green light.

The only thing that leaves is some halo effect of basestealing - defense moving around, pitcher distracted, things like that.  I’m guessing it’s unlikely those things are particularly important, but ya can’t prove a negative, and maybe the White Sox maximized these effects.

The only thing that leaves is some halo effect of basestealing - defense moving around, pitcher distracted, things like that.

Of course, there’s no reason to assume the halo effect is negative only for the defense.

Hitters hit quite a bit worse than usual when a stolen base is attempted. Hitters also hit worse the more pickoff attempts are made.

That is no way to evaluate what the threat of base stealing is worth. The hypotetical positive you get from the threat of stealing should be a pretty constant factor over the whole season.

Maybe statistically it is best to pitch in the lower outside corner, that doesn’t mean it is a good idea to do it all the time. If predictability hurts you, it is good to mix in some “inferior” strategies. And being aggresive with basestealing hasn’t actually cost the Sox anything according to the table.

And as long as you can’t prove that mixing strategies hurts you it’s a good default choice. Because it’s the correct choice most of the time.

Andere -

Sure.  I’m saying we don’t know.

It’s also possible that the overall halo effect or contextual effect or whatever is negative on average across baseball, but it was nonetheless highly positive for the White Sox.

I don’t think it’s the most likely scenario.  As I said, from watching the White Sox, I think they ran pretty much willy-nilly.  But that’s not something we can prove without much finer tools.

The White Sox are a Moneyball team, duh.

Hitters hit quite a bit worse than usual when a stolen base is attempted.

From observation, it seems that this was true for Tadahito Iguchi. He often took a strike, and sometimes two strikes, to allow Podsednik to attempt a steal. It would be interesting to see if the batters behind a stolen-base threat hit in more unfavorable counts than the rest of the batters hitting primarily in that same spot in the lineup.

Hitters hit quite a bit worse than usual when a stolen base is attempted.

From observation, it seems that this was true for Tadahito Iguchi. He often took a strike, and sometimes two strikes, to allow Podsednik to attempt a steal. It would be interesting to see if the batters behind a stolen-base threat hit in more unfavorable counts than the rest of the batters hitting primarily in that same spot in the lineup.

I don’t buy this for a second and never have. Even if you were to find that the aggregate batting average after a SB attempt in MLB was .220, this tells you nothing about whether or not the SB attempt helps the hitter or hurts him.

A. Are you more likely to attempt a steal when it’s 3-1 or 0-2?
B. Are you more likely to steal when Carlos Delgado is the batter, or when it’s Jeff Conine?
C. What types of pitchers give up more stolen base attempts, Randy Johnson/Roger Clemens or Andy Pettitte/Terry Mulholland?

Looking at his splits, it’s frankly preposterous to suggest that taking pitches for Pod hurt Iguchi. Iguchi hit .326, slugging .511, with a runner on first, and .302 with a runner on second only.

Furthermore, why is it always assumed that taking pitches hurts the hitter????? Is it only strikes that they’re taking? I suppose the lesson to be learned here is that hitters should stop taking so many ####### pitches, whether there’s a basestealer on or not.

B. Are you more likely to steal when Carlos Delgado is the batter, or when it’s Jeff Conine?

Baudib, I believe the studies that have looked at it have controlled for the batter.  IOW, it doesn’t look at the aggregate MLB performance in PAs with a stolen base attempt; it compares how Iguchi does with and without a stolen base attempt, and how Conine does with and without, and how Delgado does with and without, etc.

A. Are you more likely to attempt a steal when it’s 3-1 or 0-2?

3-1.  Which seems like it would inflate the numbers for when a steal is attempted.  Of course, it’s far more likely that a steal will be attempted on the first pitch or 2-1, and the batter will take the pitch, often putting himself in a bad count.  IMO, this is the biggest reason hitters hit worse in PA with a steal attempt.

That said, the proper thing to look at isn’t how batters do when a steal is attempted; it’s to look at how they do when a base-stealing threat is on base.  Iguchi takes strikes for Pods in situations when he doesn’t go, too.  And controlling for the pitcher (including platoon effects) seems necessary as well.

It’s also possible that the overall halo effect or contextual effect or whatever is negative on average across baseball, but it was nonetheless highly positive for the White Sox.

And you run into the “Where’s the Beef” problem?  The White Sox weren’t good at much else than home runs and did not score more runs than you would expect from their aggregate numbers. 

If the halo effect was real, our models would underperform reality for those teams.  Or the White Sox would be doing something unique.  They weren’t - they were a team that everyone knew would run itself into outs which they did early and often.  The Mets and Angels ran just as often as the White Sox - if anyone could get a positive halo from disturbing the defense, you’d think it’d be the teams that were actually successful at stealing bases.

A. Are you more likely to attempt a steal when it’s 3-1 or 0-2?

3-1. Which seems like it would inflate the numbers for when a steal is attempted. Of course, it’s far more likely that a steal will be attempted on the first pitch or 2-1, and the batter will take the pitch, often putting himself in a bad count. IMO, this is the biggest reason hitters hit worse in PA with a steal attempt.

I think it also depends on the number of outs in the inning. Maybe my perception is faulty, but it seems that a lot of baserunners will take off on an 0-2 count with two outs, reasoning that there’s little to lose in that situation, especially if the batter isn’t very good. I don’t know if there are run expectancy tables that also take the count into consideration, but I would be interested in seeing what the break-even point is in this situation.

I don’t buy this for a second and never have.

Well, I watched probably 140 regular-season White Sox games this year, plus the playoffs. Again, it’s entirely possible that my perception is faulty, because maybe I remember the times Iguchi took pitches practically down the middle to afford Pods the chance to steal because this ticked me off. But sure seemed to happen an awful lot. I don’t have the pitch-by-pitch data to check it out (although I really want to), so I offered my observations. Perhaps I should have included how often I saw the team play in my earlier post.

Looking at his splits, it’s frankly preposterous to suggest that taking pitches for Pod hurt Iguchi. Iguchi hit .326, slugging .511, with a runner on first, and .302 with a runner on second only.

Those figures include all of Iguchi’s at-bats with a man on first, not just the ones where Pods was the baserunner. I’m sure Pods was the baserunner in those situations more than any of his teammates, given their lineup positions, but he doesn’t account for all of them.

“...it seems that a lot of baserunners will take off on an 0-2 count with two outs, reasoning that there’s little to lose in that situation, especially if the batter isn’t very good.”

Aren’t most teams less likely to run in this situation?  If the runner is thrown out, then you’re left with a “not very good” batter to lead off the next inning.  Frequently, commentators will point out the likelihood of a steal attempt in the 0-2, 2 out context, if it’s a decent hitter up, but never if it’s (for instance) the pitcher, or somebody who hits like one.

Doing regressions and stuff the biggest side-effect I’ve seen come up from lots of basestealing is that the base stealing teams have a tendency to have good defenses. Which of course makes perfect sense: being able to run well affects defense and then of course there’s probably a significant correlation between a players footspeed and his overall agility.

I think in a sense a team that goes out and get a bunch of basestealers to try and have an offense that is built on the stolen base, can often succeed by the indirect effects of improving the overall team defense. That appears to have exactly happened with the White Sox, except its unclear to what extent the effect was intentional or not in their case. It seems a great deal of it was intentional in the White Sox case, but it’s tough to know for sure how much they thought it would help.

None of that means the stolen base has extra value, the player’s footspeed (and possibly his agility) adds that value, not his willingness to attempt to steal.

That said, the proper thing to look at isn’t how batters do when a steal is attempted; it’s to look at how they do when a base-stealing threat is on base.

In an unpublished study I did, I looked at the batting line for players at bat with a base-stealing threat at first (and no runner on first or third) and without a base-stealing threat on first (and no steasl attempted).  I used matched pairs to control for the batter (for example, if Delgado were at bat 5 times with a base-stealing threat on first and 10 times with a non-basestealing threat, his batting line with the threat and without the threat would both be weighted by 5 (the lesser of the two, 5 and 10).

The biggest flaw in previous studies about how batters do with and without base-stealing threats on first is that they did not control for the batters.  Obviously the pool of batters is different depending on what kind of player is on first.

Anyway, here are the results:

Non threats on first (per 500 PA, sac bunts removed)

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS

6872 85.9 23.7 2.4 16.3 35.8 3.7 81.4 5.5 .279 .336 .446 .782

Threats on first (per 500 PA, sac bunts removed)

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS

6872 88.0 24.2 2.7 15.4 33.9 4.6 71.4 5.9 .282 .338 .447 .785

As you can see, the OPS’s are almost exactly the same, suggesting that hitters do around the same whether there is a threat on first or not.  As you can also see, the individual components differ markedly.  The suggestion is that while the pitcher and the defense may be distracted and the pitcher may throw more fastballs/strikes, this effect is almost completely negated by the batter taking more strikes and/or being distracted himself.

Actually, the real test as to whether there is an overall difference in hitting is to compute the linear weights for each line and to use the proper linear weights values (with a runner on first).  Doing this, we get:

Non threats on first

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS

6872 85.9 23.7 2.4 16.3 35.8 3.7 81.4 5.5 .279 .336 .446 .782

Lwts = -5.50

Threats on first

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS

6872 88.0 24.2 2.7 15.4 33.9 4.6 71.4 5.9 .282 .338 .447 .785

lwts = -5.23

So even using custom lwts (lwts values of each of the events with a runner on first), they are about dead even, with an insignifcant edge to the batter with a threat on base…

I’m just reformatting parts of MGL’s post so I can see them better.

Non threats on first (per 500 PA, sac bunts removed)

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS
6872 85.9 23.7 2.4 16.3 35.8 3.7 81.4 5.5 .279 .336 .446 .782
</pre>
Threats on first (per 500 PA, sac bunts removed)

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS
6872 88.0 24.2 2.7 15.4 33.9 4.6 71.4 5.9 .282 .338 .447 .785
</pre>
Second set of figures:

Non threats on first

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS
6872 85.9 23.7 2.4 16.3 35.8 3.7 81.4 5.5 .279 .336 .446 .782
</pre>
Lwts = -5.50

Threats on first

PA’s sngl dbl trpl hr bb hbp so roe BA OBP SA OPS
6872 88.0 24.2 2.7 15.4 33.9 4.6 71.4 5.9 .282 .338 .447 .785
</pre>
lwts = -5.23

What should I make of the fact that the White Sox had a higher percentage of their runs come from the HR than all but one team and at the same time gave up more outs to CS (baserunners getting CS were not hitting the HR most of the time) than any other team?

It would be interesting to see the distribution of results on outs with a threat on first vs. having a non-threat on first. With a threat on first, it’s clear that the ball is being put into play more often, and there are more outs-in-play, so I’d be curious where the runners are ending up. Not that I buy into the idea behind “productive outs”, mind you, but I’m just wondering if there’s a net advantage to the batting team when a hitter makes an out with a threat on first.

-- MWE

What are you implying, Mike?  On a gorund out, I’m sure you will have more advances to second and fewer DP’s, as the “threat” is a faster runner.  As well, some of those ground outs will occur while the “threat” is stealing (which the data does not indicate).  A few tag ups on fly balls too. There will be a few more DP’s on line outs when the runner is stealing.  So I would guess that the “net result” of an out is a little higher with a “threat” on first.  Most of that is speed however and has nothing to do with the “threat” itself.  Some of that is on an “undocumented” steal which is a “hidden” advantage of the stolen base.  IOW, a stolen base is worth more than we think because some of those stolen base attempts go unrecorded when the ball is put into play and the runner advances an extra base more often on a single or double, stays out of the GDP more often and advances to second on a ground and sometimes fly out more often....

What are you implying, Mike?

Nothing in particular - mostly along the lines of your last sentence, I guess. I do think there are hidden benefits from the running game in the ways that you outline there, although I don’t have any idea if they are significant (mostly likely not terribly significant in the aggregate, I wouldn’t think).

-- MWE

You might also need to account for quality of pitcher.  I think most managers - especially those in the Guillen mold - are more likely to try and make something happen against the better pitchers, hence the poorer batting numbers when stealing, etc.

Remember that we (I am at least) are looking only at a “threat on first” versus a “non-threat” on first.  The quality of the pitcher will be not be biased in either case.  Trying to look at what happens at bat when the runner is in fact running is problematic in the first place.

The perception touted by commentators is that the value of a basestealer is more than his SB/CS numbers, that he “disrupts” the pitcher and the defense, presumably such that the batter does better than he would have had there not been a base stealing threat on first.  As is often the case, the commentators may be right, but they are talking out of their a**es, as they have no idea whether in fact batters do better or worse (or the same) than they would with a non-threat at first.  They just assume in their little minds that that must be the case, because it seems like it should, according to their perception, and because it gives them something intersting to say to the viewing audience, and gives them an opportunity to look smart.  As James used to say, the first thing a sabermetrician does is to ask questions like, “Is that true,” and “Why would that be true,” and then tries to find out, using the scientific method, whether it is or is isn’t, rather than just say the first thing that pops into their head, or parrot what someone else said, etc.  So as it turns out, we have no evidence that a threat on first causes the batter to hit better than with a non-threat on first.  Woops, there goes another baseball truism.  Hard to believe, isn’t it…