THE BOOK--Playing The Percentages In Baseball

Oh my.  I just started reading the ESPN comments.  There’s 1600 so far!  Plenty of snark, but also a fair amount of reasonable skepticism.  Quite the fire the ESPN authors have set…

”...if you were to do all 30 teams, I’m sure you’ll find a couple of other teams with a bigger advantage at home than the Jays hitters.”

Milwaukee seems the much better suspect statistically with regard to sign-stealing or other funny business. They hit below league average on the road, but their home OPS is over 100 points higher than the overall league average at Miller Field. Visiting NL teams don’t get a bump in that park--their cumulative OPS is below league average in Miller.

Nice first-punch with Milwaukee (2011 I suppose?).  Time to continue.

The basic process has been laid out.  It’s time for the young whipper-snapper saberists among you to make their mark.  Jeremy, Eric, Matt, Max, et al: make us proud.  The dance floor has been cleared for you.

Tom, yes, 2011.  I’m no sabermagician, nor do I have the statistical or database chops to determine the kind of meaningful analysis of you and others, but at a casual glance, Milwaukee really sticks out.

Through 8/8/11:
NL overall: OPS = 709
All NL teams in Miller: OPS = 753
Only MIL in Miller: OPS = 817
All non-MIL NL hitters in Miller: OPS = 685
MIL away from Miller: OPS = 675

I’m not casting stones, and wouldn’t make a direct accusation of quote-unquote cheating. I’m not a fan of the Brewers or any other NL Central team, so I have no dog in the fight. But I do find the splits highly questionable. I’d really love an advanced statistical explanation of how home cooking is making a team that is 34 points lower than league average on the road 108 points higher at home, in a park where other teams experience a drop in their pop.  I think it’s also highly suspect that the Brewers walk 8.56% of the time at home, but only 6.68% on the road, as if Miller suddenly makes them more patient and selective at the plate.

Sure, the Brewers have the 4th best pitching fWAR, which could help suppress what visiting teams do in Miller. I can’t see that bridging the gap, though, in conjunction with ‘home cooking’, between what Milwaukee and their opponents have done outside and inside Miller Field.

According to the ESPN article, all that is being tipped off is whether the pitch is off speed or a fastball.  The batter knows nothing about location.  This would affect walks how?  Doesn’t this throw most of the OPS analysis off?

Are you suggesting that if the batter knew 100% if the pitch was a fastball or breaking pitch that his walk rate would not go up?

Not a player, just a watcher, so I don’t want to suggest it wouldn’t be some help, but I don’t think it would help nearly as much as it would augment a player’s power numbers. 

I agree that if pitches are being tipped, all types of hits should increase—not just homers, but its not obvious to me that walks would increase the same way.  If anything, knowing the speed of the pitch might lead to extra agressive cuts, leading to more strikeouts and fewer walks. 

Anyway, I’m looking forward to reading the full ESPN article.  The universal reaction has been to slag it, but for me there is enough there to make me want to read more.

@Beyonder/7

I would assume walks would dramatically increase if the hitter knew what was coming. If it looks like it’s starting out in the zone, and you know it’s a breaking pitch, it’s pretty easy to assume it’s going to dive out of the zone and that you can safely take the pitch.

The point is that all events are in play.  Therefore, there’s no need to just ignore certain events.

The only reason to ignore walks or strikeouts or other things, is if you think the changes there is noise.

Before you do that, you have to prove that they would be noise.

Otherwise, you have to keep everything in.

In any case, it doesn’t take away from the fact that the true outlier is not 2009-2011, but 2005-2008.

Tangotiger @9

The major point that gave me pause from the article was that the Jays were only a slightly above average home run hitting team on the road in 2010 compared to the rest of the AL—and in contrast, a very good home run hitting team at the Rogers Centre compared to visitors at the RC.  According to the article, the “home court advantage” enjoyed by the Jays was the 3rd greatest in the last 60 years. 

In challenging that theory, you broadened the analysis to OPS, reasoning that the “tips” should affect everything (walks, singles, doubles, triples, and strikeouts), the same way they allegedly impacted home run totals.  My point is that while the tips may well affect all these things, they may affect them in different ways, and to different extents.  If you are going to refute a narrower theory by introducing a broader base of data, I would have thought the onus was on you to explain why it should be taken into account.

I take your point at #10 though.

#11) The problem with the article is that the writer(s) knew there was a charge of stealing signs (from unnamed players), and rather than look at everything and give that evidence, they only gave the data for home runs. Why should they assume it would only affect home run rate. It’s misleading.

My point was just to open the dance floor, and let the other dancers do their thing. 

Ideally, you would do a component breakdown, and even look at things like “swing and miss” by fastball/breaking, etc.

Talk of just HR is just too myopic.

#12

That is called journalism.

The ESPN article suggests that any shenanigans were most likely taking place in late 2009 and early 2010.  Eventually, it looks like at least a few opponents in 2010 took counter-measures. 

Would it be helpful to examine smaller chunks of 2009 and 2010 (at the risk of reducing the sample sizes?)

Just eyeballing the month-to-month RS numbers, Sep 2009 and May 2010 were pretty big offensive months for the Jays.  April 2010 was nothing special on the other hand. 

[It’s tempting to consider the possibility that either Anaheim (Apr 16-18 2010) or Boston (Apr 26-28) was the anonymous opposing team in the opening paragraphs of the ESPN paragraph, and that the final 2 games of (at least) one of those series were played with a temporary loss of sign stealing].

Hi Tom,
I did a quick look at the swing and miss rates in and out of the zone and absolutely no smoking gun whatsoever.

The overview is here: http://freekraut.com/?p=4278

I might add some more data later

The Rockies experienced a similar effect last year.  They hit 66 percent more home runs (108 to 65) at Coors, while their opponents hit just 32 percent more (79 to 60).  This trend holds up (though to a lesser extent) over the past three years. 

I believe, however, that this has to do with the ground ball and fly ball tendencies of the pitching staff and the lineup.  The Rockies have become more and more reliant on ground balls.  Ground balls don’t typically go for home runs, no matter what stadium you’re playing in.  As a result, Rockies pitchers have been far less effected by the Coors altitude than opposing pitching staffs.  Additionally, the Colorado lineup hits a lot of fly balls. 

Similar statements can be made about the Blue Jays.  They have a ground ball-favoring staff, but hit fly balls at the highest rate in baseball last year.  Therefore, the Blue Jays would likely be effected more than opposing lineups. 

If the Blue Jays played at PETCO Park, they would still be more effected by the ballpark, but in the opposite direction.

Why should they assume it would only affect home run rate. It’s misleading.

Who says they assumed that?  The focus on home runs is clearly a tie-in to the shouted words from the visitors bullpen aimed at Jose Bautista which were quoted in the original article.

If someone had instead screamed at Bautista “It’s not too [f------] easy to lower your swing & miss rate by 1.73 standard deviations when you don’t know what’s coming!”, then maybe the authors would have addressed that point.

They can’t include everything.  This was an ESPN the Mag article, and print articles, even more so than online articles, have strict space limitations.

Really, the fact that no former players have come out and discussed this is pretty damning evidence against the “stealing signs” crap isn’t it?

What allegiance does John Buck have to Toronto now? Wouldn’t it be in his team’s best interest to mention something publicly? Or hell, you would expect at least one of the many hitters that have played for the Jays over this time period to be uncomfortable with this system and leak the info to the press.

It just seems like a run of the mill conspiracy theory. Hunt anomalies and use anecdotal quotes as evidence for your point. Next week on ESPN: “What Really Happened on September 11? Evidence suggests it could have been A-Rod!”.

Why is sign stealing illegal at all?

You don’t exactly need to work for the NSA to figure out how to keep the other team from knowing what’s coming.

The 2010 Jays batters are notorious for being A) extreme pull hitters and B) unusually right-handed (75% of PAs were as RHB, AL average around 55%).

Extremely roughly, at the Rogers Center the Jays hit 3/4 of their homeruns to left, their opponents hit 2/3 to left. Across the AL, the LF/RF split would be around 55%. I don’t know about the left/right split of their opponents but lets assume it’s roughly average. The skydome walls are symmetrical, but this suggests its easier to hit a homerun to left than right.

Wouldn’t you expect the 2010 Jays to hit a bunch more homeruns at home?

@21

Isn’t there a chicken/egg problem here?  Are they notorious pull hitters because they’re getting tips and recognizing pitches that can be pulled to the short porch?

#19/Mr. Cthulu:
A former player would probably warn his new teammates.  But what’s his motivation to discuss it publicly and attach his name to it?

Also, how likely is it that four (albeit anonymous) sources would conspire to completely fabricate a story like this out of thin air and independently corroborate the same observations to a reporter?

Some entirely observational notes from a Jays fan.

The Jays are a gang of notorious uppercutters. To some extent this seems to be an effect of their successive hitting coaches.

The Skydome (no new name for me thanks) plays as an extreme HR park when the roof is closed. When the roof is open, prevailing winds come off the lake (as anyone who ever went to the former Exhibition Stadium will confirm and hit the “scoop” in which the roof sections fold up. This pushes the wind down - right on top of flyballs. So the HR effect is much smaller (this is anecdotal, although I remember looking at this via the numbers once).

I thought the effect of a very high HR park factor on contact, for home players only, might be related to this - Jays players uppercutting into a very friendly park environment when the Dome is closed (i.e. especially in April, May and September). But this, from a quick review, does NOT appear to be the case with the 2010 data.

Chris St. John at Steal of Home identified the White Sox in May of last year, as the team that had the kerfuffle with Jose Bautista. Bautista subsequently confirmed this to the press today.

"Are you suggesting that if the batter knew 100% if the pitch was a fastball or breaking pitch that his walk rate would not go up?"|

If I could sit dead red with confidence, my walk rate would go way down.

@23 I’m not suggesting that the four players are fabricating the story. I’m saying that their perception of events is not convincing evidence. Just because they remember that they could accurately predict fastballs based on the hand gestures, doesn’t mean that actually happened. They easily could have simply remembered the hits and forgot the misses once they decided he was stealing signs (how else would psychics stay in business if not for this human quirk).

As for former players not going to the media you could be right. But let’s say every former player warns his teammates, teammates like to gossip with other ball players and soon pretty much everyone in baseball would know the story. Someone would say something, or at least a complaint would be filed.

19/26 Mr Cthulu:

There’s rarely a benefit to being a whisteblower in an industry that guards itself as closely as MLB. If, to use your example, Buck were to go to the media, he would most likely be ostracized within the industry. It would effect his future employment/salary. Sign-stealing is probably not a big enough issue to risk your multi-million dollar per year job living the dream of playing in the major leagues.

No active players piped up to the media about PEDs. No Wall Street guys or gals left one bank for another and aired out the shady stuff happening at their former place of employment that risked crashing the economy and screwing up millions of peoples lives. No Hilton manager moves to Hyatt and announces to the media how infrequently the duvets are laundered at Hilton.  All that stuff would bring as much heat on your industry in general as it would your past employer. People simply don’t find it worth being a pariah in an industry in which they’re still employed, right thing to do or not.

"No Wall Street guys or gals left one bank for another and aired out the shady stuff happening at their former place of employment that risked crashing the economy and screwing up millions of peoples lives.”

I’d just like to point out that this is not true. Lots of people inside the financial industry, and lots who left, said over and over again that there were serious problems with all sorts of things that were happening, from structural problems with leveraged CDOs, to rating agency problems, to problems with wildly overaggresive mortgage origination. To name just three things that I myself would happily tell all and sundry about in roughly 2002 when I started to become aware of the issues.

But bearers of bad news and ill tidings were ignored by everyone. There was a boom on, it was raining soup, and everyone was too busy grabbing buckets to listen to a lecture.

bojan: great stuff!

Tango, even though it is likely that 1 or 2 teams would be 2 SD from a mean in whatever category, if the claims of sign stealing are made independent of of any knowledge or suspicion of statistical anomalies, then ANY difference from the mean (in the right direction of course) is evidence in favor of that claim.  Obviously the larger the number of SD the stronger the evidence, but in this case we are not looking at all teams H/R splits and then suggesting that the ones with large splits may be stealing signs…

What if they are making those claims because they are paranoid because they see Bautista hit so many HR? 

“Look, there’s Glenn Close and her white hat!  And look, there’s Robert Redford hitting a HR!”

@30/MGL: In the case of Milwaukee, I believe the only evidence of sign stealing is the unusual splits. I maintain that without outside evidence, unusual splits are not evidence of cheating at all. Unusual events happen. The suggestion is that the explanation is sign stealing, but without independent evidence that this is happening, we cannot make that charge with any justice. There could well be some other explanation, or it could just be normal random variation in performance.

Whether sign stealing actually occurred and whether it had much net impact are two separate questions.

We know the ‘51 Giants stole signs at the Polo Grounds, even though no one (AFAIK) has found anything in their home/road splits that year to suggest it helped (see http://espn.go.com/classic/s/neyer_on%20_shot.html)

It’s possible that it doesn’t help as much as we would expect or that it’s very hard to pull off with the kind of frequency and reliability to produce net results that show an obvious benefit.

#32, right.  If you start with the observation of extreme splits, then you can NOT conclude that sign stealing is likely, since a small percentage of teams will likely have unusual splits by chance, and if you apply a Bayesian analysis, you will still come up with a very small likelihood of sign stealing for a team with large splits.

Tango, sure that would suggest some SMALL selective sampling, but hardly enough to be equal to the MIL situation.

Apparently, there have been several accusations of sign stealing based on things that have nothing to do with splits.  Now, it is certainly possible that some people were aware of the splits and THEN created the story of the “guy in the white shirt,” etc....

MGL or Alex. 

Would you mind explaining the point to me more simply?  I think I get where you are coming from, but would like to get a better handle on it.

It reminds me of those discussions to the effect that it is hardly a rare coincidence if two people in the same classroom of say 50 people have the same birthday.  It is, however, much more of a coincidence if someone in that class shares your particular birthday. 

Apologies if this is too basic a question for an advanced thread like this (or if I am totally off base), but I would like to understand what you are saying.

It’s a bit odd that the exchange between Bautista and the ChiSox happened the way it was reported.  Through that point of the 2010 season, Jose Bautista was anything but a slugger.  By the end of the game in question, he was hitting .161 with a single homer.  Other than a big September in ‘09, there was no indication of what he was about to do.

Why would someone call out “it’s not so easy” to a guy hitting under his weight?

The only way it makes sense at all is if the comments weren’t actually directed at Bautista, who finished 0-3 with a walk.  The Jays had 3 hits and two doubles in the fifth, then went down in order with 2 Ks in the sixth.  It had to be a generic comment about the Jays rather than anything specific about Bautista.  Jose was just the fielder closest the the Chicago bullpen.

A compilation of articles on the subject:

http://blogs.thescore.com/mlb/2011/08/11/reactive-reactions-spy-dome/

Beyonder, here is an analogy that might help explain the “Bayesian math”:

Say I flip a coin I find in my pocket 10 times and come up with 9 heads and 1 tail.  The chances of that happening by chance is very small (9 out of 10 or 10 out of 10 only happens around 1 out of 100 times or something like that).

Do I conclude that I have a biased or “trick” coin?  No, because the chances of a random coin from my pocket that looks and feels normal being a trick coin that lands on heads 90% of the time or so is probably less than 1 in a million.

So in order to figure the chances of this event being random versus having a weighted coin, we take the ratio of the two, which is, say a million to a hundred, or 10,000 to 1.  So it is 10,000 times more likely that this is a random event (than I have a trick coin).

So even if we look at only one team and find out that their splits are large, like MIL, in order to estimate the chances of them being sign stealers, we have to take the ratio of the splits happening by chance versus the chances of a team stealing signs and having those kinds of splits. 

Obviously we still don’t know what affect sign stealing would have, numbers-wise (do the math is complicated and we would have to make several assumptions and estimates), but let’s say that the chances of those splits were 1 in a 100 (a little less than 2.5 SD for a one-tailed test) by random chance and let’s say that the chances of any team being sign stealers is 1 in a 1000.

We can then conclude that there is only a 10% chance that they are stealing signs and a 90% chance that they are not and the extreme splits are due to random chance since the ratio of 1000 to 100 is 10 (to 1).

Now, let’s say that we look at all 30 teams and find out that 1 of those teams has extreme splits. Instead of the chances of those extreme splits occurring by random chance being 1 in a 100 times, it might be 25% or something like that (the chances that at least 1 team in 30 is 2.3 SD from the mean).  So now we have the ratio of 1 in a 1000 to 1 in 4, which is only a 1 in 250 chance of this team being a sign stealing team.  (Even if the chances of any random team being a habitual sign stealing team is 1 in a 100, our ratio would still be 25 to 1, or a 4% chance of this team being a sign stealing team.)

So because we looked at 30 teams and found 1 (or more) with extreme splits, our chances of finding a 2.3 SD split goes up like 25 fold.

Same with the birthdays.  The chances of any one person in a room having the same birthday as I is 1 in 365.  If there are 20 other people in the room there is a 5% chance that someone has the same birthday as I.

But the chances that at least 2 people out of 20 has the same birthday is much, much higher of course (around 40%), since there are all kinds of combinations of people that can have the same birthday. It is 5% for person #1 and then a little less than 5% for person #2, etc.

I find it interesting to note that as one’s “estimate” of the likelihood of cheating increase, so does one’s likelihood of interpreting inconclusive data as suggestive of cheating. 

That’s quite a profound illustration of how one’s preconceived notions about a topic (aka “baggage") can pollute what you might hope was an objective analysis.

MGL, how does it change if the equation has more factors than one--the gap in splits?  If we consider the chances of a certain gap in H/A splits in a vacuum, I can see how it would not tell us much. But the components of the gap in their H/A splits give us a few more coins to flip.

What are the odds....

-that Miller Field depresses all other teams to the degree it does, based on historical park factors and level of competition, ie MIL’s pitching and defensive efficiency?
-that Miller Field increases MIL’s line to the degree it does, based on historical park figures and the level of their opposition’s pitching and defensive efficiency?
-that all other parks in the NL are depressing MIL’s offense to the degree they are, based on park factors, opposition, etc?
-that the sum all other parks in the NL, for hitters facing MIL pitching and defense, are not having the same impact on them that they have when playing in Miller Field?

As I said above, I don’t have the database or statistical chops of many of the fine folks on this forum. I’m learning a lot in this thread and hope to learn a lot more. Also as I said above, I have no dog in this fight, and don’t care at all if MIL is up to something funny as much as I care about if it’s something that can have doubt either removed or cast based on statistical analysis. Thanks for your time, and feel free to be blunt if I’m way off base.

MGL. Very kind of you to take the time out of your day to explain that for me. Great explanation. Many thanks.

Beyonder, my pleasure.  Glad you were able to decipher my gibberish!

John, sure, as I said, the actual model and calculations are much more complex.  I was illustrating and explaining the basic idea.  The key factor is the “a priori” probability, which we can only guess at - that is, the chances of a team being a habitual, ling-term sign stealer.  We can also only guess at the effect (and extent) of sign stealing.  We would have to posit a number for both of these things in order to do the calculations, and estimate the chances of any given team being a sign stealer given the splits we observe…

knowing if the pitch is fastball or not will indirectly affect a hitter getting walks.  if a hitter knows the pitch is gonna be fastball, he can sit fastball, and only react to pitch location.  whereas if a hitter does not know, he’ll need to react to pitch speed and location, hence increasing the likelihood of getting strikeouts.

All I’m saying re: the walks, is that the relationship between tipping and home runs is one way.  It only improves your chances at hitting homers. 

With walks, it arguably cuts both ways.  Yes, you may have an easier time recognizing balls and making contact on strikes, but it should also lead to a greater percentage of swings (both accurate and innacuarate ones). 

So with walks, the effect of tipping is multi-directional (it giveth and taketh away), while with homers it will only make

New point. When asked, Jason Frasor didn’t deny that the tipping was going on, and only said that he wasn’t aware of it, and that “if this was bbeing done, it could only have been done by a small group of guys.”

Forget for the moment that Frasor now likely plays with some of the accusers on the White Sox.  This would make much more sense than postulating a conspiracy for the entire team from the top down.  As AA said in the press conference, this would be a very difficult secret to keep.  If we are talking about only a select few players, most of the statisical calculations done so far would be moot. 

To me, the most probative stats in the article were the stats that showed that at the Rogers Centre the Blue Jays enjoyed a “home Court advantage” in home run hitting seven times greater than the major leage average.  Is there a way to isolate whether any particular players enjoyed a massive home court advantage?.

Oh boy that “7 times” is just so useless.  It’s why I hate people manipulating numbers.

It works something like this:  suppose that the average split, in whatever, is .101 at home and .100 away (differential of 1 point).  Then someone comes along with a split of .107 at home and .100 away (differential of 7 points).

All of a sudden, someone is going to say: “Look!  His split advantage is 7 times larger at home than the average guy at home!”

That’s just such a crappy way to present the data.

Talk about standard deviations from the mean, or just don’t talk about it at all.

Umm.  It’s ESPN magazine.  You may want to talk about standard deviations from the mean, but there’s no reason to pitch something to a general audience that way.

Anyway, the reference to “7 times the advantage” is pretty clear in this case.  It’s not at all what you are talking about.  On the road in 2010, the Jays hit homers in 4% of plate appearances where thay made contact, compared to the AL average of 3.6%.  So the Jays are plus .4% on the road.  At home, their homer rate on plate appearances where they made contact is 5.4%, a difference of 1.8% over the AL average. 

This is apparently 7 times the magnitude of the normal home field advantage. 

We don’t have the raw data, but I think its clear we’re not talking about the trivial type of advantage contained in your example.

It’s the same basic premise!  It’s an incredibly crappy way to explain things.  “7 times”.  Please. 

What counts is that you take a difference of the two differences.  You do not, absolutely not, DIVIDE the two differences.

After all, why are you disputing my illustration?  It gives you “7 times” in either case.  Oh, that’s right.  We don’t care about that!  We care about the magnitude of the numbers.

It’s incredibly crappy.  No need to defend it further, and I won’t call it crappy any longer.

I suppose it is possible that knowing pitches could decrease your walks, but I think that is highly unlikely.  I am 99%+ certain that knowing your pitches will greatly increase your walk rate.  Basically you won’t swing at all those breaking pitches outside the strike zone (down, away, or both) and fastballs up in the zone when you know what’s coming.  And you will turn a lot of 0-2 and 1-2 K’s and weak hits into walks because you are not swinging at all those off-speed pitches out of the zone.  And 3-2 walks would increase tremendously. The more I write about the more certain I am that there is close to a zero chance that knowing pitches can decrease walks....