Fangraphs win expectancies: ALDS game 2

That was absolutely one of the most brutal innings I’ve ever witnessed. Joe Mauer is robbed of a double by a terrible call. He gets a single out of the at-bat anyway–but the next batter, Jason Kubel, hits a single that probably would have scored him if he had been on second. This turns out to be incredibly important because what ultimately happens in the inning is that the Twins load the bases with no outs, and then proceed to make three outs on five pitches without scoring a run. Perennial goat Delmon Young’s line-out was actually the most respectable of the three outs–he hit it right at Mark Texeira, but it was a well-hit ball that mostly turned into an out because of bad luck.

The same can’t be said of the weak efforts from Carlos Gomez and Brendan Harris. All told, the cumulative effect of those three at-bats was to drop the Twins chance of winning a mind-boggling 47.8%, from 83.5% to 35.7%.  After that, it seemed inevitable that Mark Teixeira would start off the bottom of the 11th with a walk-off homer to cap off the collapse.

If that all wasn’t painful enough for you, imagine watching it the way I did: In a New York City bar, surrounded by screaming Yankee fans who were always there to chime in with the exact reverse of my emotions. I keep telling myself that we all wrote this team off a month ago anyway, and it lessens the pain a bit.

Small comfort that it is, I can also say that I had one positive baseball experience today, when I discovered postseason.tv, the new project from Major League Baseball’s Internet department. I’ve been a customer of mlb.tv for about five years, because it allows me to watch Twins games even though I don’t have cable or live in Minnesota.  The service has had its ups and downs, but I do think it’s gotten much better since its frustrating early years. Now that it’s integrated into Boxee, it’s easy for me to stream baseball from my computer onto my TV, essentially replicating the cable experience without the expense.

With postseason.tv, MLB has implemented something I’ve wanted to see for a long time: access to all of the raw camera feeds that go into a finished broadcast. You can flip between, say, the center field camera, the dugout camera, and the home plate camera, or put up to four different cameras on screen at once. The experience takes a while to get used to, but once you get the hang of it it adds a new layer of richness to the game.

Reviews of postseason.tv haven’t been all positive. But it seems to me that $9.95 is actually a pretty decent deal for what you get, and it’s a mistake to expect the service to be a full replacement for the finished TV broadcast.  I do agree that it has some problems, many of them technical: it’s generally impossible to get all four cameras exactly in sync, for example, which makes it a bit hard to move back and forth between watching different camera angles (much less sync them up with the TV broadcast, which is supposedly what you’re intended to do with the service.)  But the underlying concept is a very good one, and I’m optimistic that it will eventually develop into something more powerful and more usable.

Ultimately, the availability of alternative video sources could be combined with another innovation I’ve long hoped for, DIY play-by-play, to create a thinking person’s alternative to the lowest-common-denominator antics of the mainstream sports broadcasts. Of course, MLB and the other sports would fight this development, just as the record companies fight against sampling and remixing. But it still looks to me like the wave of the future.

None of that, obviously, makes up for tonight’s game but there nothing to be done for that except look ahead, and hope for another miracle in the Metrodome.

Twins Geek remarks on Randy Marsh’s strike zone last night, and says that “it seems like a shame that an umpire’s very bad day could affect such a critical game”. I certainly had the same impression–although the one indisputably bad call, missing Inge getting hit by a pitch, had nothing to do with strike calling. At one point I texted my friend: “mad inconsistent strike zone tonight”. But while I was sleeping off that epic win, my pitchf/x database got updated with last night’s game, so I figured I should take a look at the data:

Randy Marsh's strike zone, October 6th 2009

These are only the balls that Marsh actually called–i.e., the ones that weren’t swung at. Green means it was called a strike, and black means it was called a ball.

We can certainly see here that Marsh’s strike zone was small, which according to the game announcers is the norm for Marsh. But I’m not sure it was actually all that inconsistent–except maybe in the 11th, when it looks like he started calling more strikes.

Here’s another way of looking at it: a graph that plots all of the called pitches from last night on one graph:

Randy Marsh's calls, October 6th 2009

The gray box is the “rulebook” strike zone (with the top and bottom based on averaging over different players). The dark lines are my estimates of the “true” strike zone. I won’t go into the details, but basically I use a regression model to figure out the point where a pitch has exactly a 50-50 chance of being a ball or a strike. Above, I show both an estimate for a normal rectangular strike zone and for an elliptical zone. My research suggests that real strike zones aren’t actually rectangles, so the ellipse may give a better sense of how pitches are actually called.

Finally, let’s look at all of Randy Marsh’s calls this year and last year (the period for which I have pitch data). Was last night’s strike zone unusual?

Randy Marsh's calls in 2008-9

Doesn’t look like it. You can barely see the rulebook zone on this graph, but Marsh just always seems to call a smaller, higher strike zone than the rulebook, and that’s what he did last night, too. And it turned out to be all the strike zone the Twins needed.

I excel, I think, at reducing large amounts of quantitative information into digestable and understandable forms. I’m not so good at conveying the emotional intensity of baseball at its best. So what can I say about Twins game 163, which is one of the very best baseball games I’ve ever witnessed?

Well, the last few days have clearly established that two of the greatest trades in Twins history were the ones that brought Delmon Young and Orlando Cabrera to Minnesota.

Of course, I don’t actually believe that. I still think those trades were bad for all the reasons that have been said before. And the actual best trade ever is obviously the AJ trade. But it’s certainly strange to come to the point where the Twins are in the post-season in no small part because of a late hot streak by Young and a timely home run by Cabrera. Of course, even those heroics don’t make up for the fact that Young and Cabrera were awful for most of their at-bats with the Twins this year. And even their awesome late outbursts didn’t contribute nearly as much to the team as Joe Mauer did by just being the best hitter in the game, month after month. But there’s objective metrics, and then there is, in Bill James’s wonderful phrase, “the politics of glory.” So I stick by a text I sent to my long-distance Twins buddy CK, after Delmon came up with another huge hit:

Dear Delmon Young: all is forgiven

Literally (like, literally literally) during tonight’s game, I was jumping in the air, shouting and pumping my fist multiple times during that game, and just as often I was on the verge of collapsing on the floor. My emotional state, recorded in a series of text messages with CK, ranged from

O-Cab: best trade in Twins history, after Delmon Young

To:

Dear god no not Bobby Keppel

And back to total elation in the end. My sunny mood is probably short-lived, because unless you believe in “momentum” (which I don’t), you have to think that the Twins are now utterly exhausted and ready to be torn apart by the Yankees tomorrow. But I can honestly say that this was a great Twins season even if it goes no farther than this.

After two straight years of this, I’m beginning to forget that baseball’s regular season is only supposed to last for 162 games.

My prediction that Joe Mauer would almost certainly not hit .400 has been borne out, and yet something just about as improbable has happened instead. Suffice it to say that at the outset of the season, nobody was predicting that things would end with the Twins making up a seven game deficit in a few weeks and becoming the first team ever to make up a three-game deficit in the final four games, all with Justin Morneau on the shelf.

And now we get something uniquely random and unpredictable in baseball: a one-game playoff. The playoffs are, in general, a crapshoot, as Michael Lewis wrote in Moneyball, due to the importance of luck in a short series. A one-game playoff is an order of magnitude moreso.

It’s hard to even know how to assess the probability of a Twins win, because one-game playoffs are so uncommon and unusual. Going by their records, of course, these two teams are evenly matched. Going by the five previous play-in games in the wild-card era, you’d say that the home team is a heavy favorite, since only 1 of 5 road teams has won one of these things. And according to both the Vegas odds and a simulation-based prediction posted over at Twinkie Town, the Twins have about a 60% shot at winning.

And of course, none of this takes into account the unmeasurable implications of a star Tiger player getting into a drunken fight with his wife on the eve of a critical game.

But as we saw last year, it can all come down to one pitch. So there’s nothing to but sit back, cross one’s fingers, and hope to squeeze a little more magic out of the old Metrodome.

Below is an overview of our two starters this afternoon, who turn out to be fairly similar pitchers.

First, Scott Baker:

Scott Baker

Next, Rick Porcello:

Rick Porcello

Some unrelated thoughts I had while watching last night’s game.

1. Daniel Hudson and chasing the high hard one

During the game, it seemed to me like Daniel Hudson was getting an awful lot of swings and misses at pitches that were way up in the zone. Looking at the pitchf/x data suggests I was wrong about that, though: although the Twins were definitely fouling off a lot of those pitches, they only swung and missed a few times:

Two of those whiffs at high pitches were by Jason Kubel, which prompted me to look up all of the strikes he’s received this year:

That’s a lot of swings and misses at high pitches! Kubel seems to be particularly susceptible to this–if you look at the graphs for, say, Justin Morneau or Michael Cuddyer, they don’t look like this.

It’s also interesting how Kubel’s “whiff zone” seems to start almost exactly at the top of the strike zone: if it’s high but in the zone, he can generally get a bat on it. If he learns to recognize those high pitches a little better, he could become an even better hitter.

2. What’s in an Error?

Tom Tango had an interesting post about how we statistically categorize the situation where a batter reaches on an error. Traditionally, it’s treated the same as making an out. But this only makes sense if you assume that the batter doesn’t have anything to do with “forcing” the fielder’s error. Tango seems to think this is wrong, but in the comments his co-author Mitchel Lichtman defends the practice.

Just after I read that exchange, I saw Orlando Cabrera reach base on an error where he bunted the ball back to the pitcher, who proceeded to throw the ball away. Now, should Cabrera get some credit for reaching base in that situation? On the one hand, it was arguably a bad situation to lay down a bunt, and less arguably, it was a bad bunt, which would normally have resulted in an out. On the other hand, maybe the fact that Cabrera chose a surprise bunt in an unexpected situation, against a rookie pitcher,  was precisely what created the situation that led to an error. It seems like there should be a way of figuring out empirically how to assign the credit/blame in this situation, but there are so many interacting factors that I can’t really wrap my mind around it at the moment.

3. You don’t see that every day

Recently, I’ve been working on a baseball simulator. The essence of the program is based on the fact that for any state of the game–that is, any combination of a certain number of outs and a certain configuration of baserunners–you can model the probability of transitioning to various subsequent states. For most events, it’s pretty easy to determine the transition probabilities empirically, because we have lots and lots of examples of different kinds of events occurring in real games. So, for example, it’s easy to figure out the probability that, if there’s a runner on second with one out, the outcome of the play will be a single and a run scored–because that situation has happened so many times in baseball history.

But as I start to refine the simulator, I have to deal with certain kinds of rare events, which only happen in certain unusual situations. One of these happened in last night’s game: a double steal.

In the data covered by Retrosheet–going back to 1911–there are less than 4000 attempted double steals. In recent years, there have been about 100 per year. By itself, this level of rarity would make it hard to accurately determine when double steals happen, but not has hard as some other, even rarer events–catcher’s interference, for example.

The bigger problem, however, is how many details of the game situation you need to look at to really know whether a double steal is likely. Consider the situation in last night’s game. You have combination of, at a minimum, the following: fast runners on both first and second (Carlos Gomez and Nick Punto); a pitcher who is slow to deliver the ball to be plate; and a catcher who isn’t that good at throwing out runners. If any of those elements had been missing, the double steal probably wouldn’t have been called. In addition, the score has to play a role: you don’t generally go for a double steal in a blowout. In this latter regard, the situation last night was kind of unusual: the Twins were ahead by four runs, and only 14% of the double steal attempts in retrosheet came when the run differential was that high.

In general, I’m pretty tolerant of the repetitive and cliche quality of Dick Bremer and Bert Blyleven’s TV commentary.  I shrug at the dopey homerism; I put up with “circle me Bert”; I don’t even mind the way Bert thinks everything is either an “inside-out swing” or an “uppercut swing”. Despite it all, I find them sort of blandly reassuring, so I tune in to them even though my mlb.tv subscription gives me the option of watching the opposing team’s broadcast.

But I really, really hate the idea of “productive outs”. In tonight’s game, with Denard Span at second, Dick and Bert kept yammering on about how Orlando Cabrera just needed to hit the ball to the right side to advance Span to third. When he did exactly that, they congratulated him on a “productive out”.

Now I suppose there are different ways one could define “productive”. But according to the Fangraphs win expectancies, Cabrera’s out increased the Blue Jays’ chance of winning, from 44.2% to 45.7%. It’s hard for me to envision a definition of “productive” that’s consistent with “making your team more likely to lose”.

Of course, Orlando Cabrera sucks, so maybe advancing the runner is the most we can reasonably expect from him in that situation. But that brings up the larger question of why the @#$% is Ron Gardenhire batting him second again!? Gleeman just made that point today, so I won’t belabor it, but sheesh.

This graph approximates the feeling in my stomach at the end of today’s game:

As the cliche has it: every time you watch a baseball game, you can see something you’ve never seen before. In this case, it was Joe Nathan fielding a ground ball and screwing up a routine throw to second, turning a possible double play into a gut-wrenching bases loaded situation.  When the play went down, my super was setting out the building trash for tonight’s pickup, and I think he probably heard me cry out in anguish.

All’s well that end’s well, and Nathan managed to twitch his way out of a jam. But I got to wondering–just how often does something like that happen? I’m defining “something like that” to mean “save situation with runners on first and second and less than two outs, where an error on the pitcher loads the bases”.

The answer is, it doesn’t happen often. In the retrosheet era (meaning since 1956), I only found 13 games that fit the bill. (I’m excluding save situations that came in extra innings; including them would add a few extra games. Also, if this has happened already this year, I wouldn’t know about it, because retrosheet only updates their data after the season is over.) The last one was more than eleven years ago, when the Brewers’ Bob Wickman loaded ‘em up with Phillies on a Scott Rolen ground ball that looks like it was probably similar to the one Nathan fielded.  There was one out at the time, just like in today’s game. And like Nathan, Wickman got out of it, striking out the next two batters.

Oddly enough, a similar situation had arisen earlier in 1998. Even better, this one involved the Twins! Only that time, it was the Twins who loaded the bases, and other team’s closer who committed the error. In this game, trailing by a run in the 9th against what was then known as the Tampa Bay Devil Rays, current Twins TV announcer Ron Coomer (wonder if he remembers this game?) and Terry Steinbach hit back-to-back singles, bringing up perennial ’90’s utility man and career .251/.310/.344 hitter Denny Hocking. Hocking tried to bunt the runners over, and Hernandez fielded the ball and committed an error to load the bases. There were no outs in this case, and Hernandez wasn’t as lucky as Nathan or Wickman: after getting a force at home on the next batter, good old Matt Lawton hit a two-run single to put the twins up by one.

And that would have been that, except that Rick Aguilera came out in the bottom of the 9th and proceeded to give up an RBI double to a washed-up former rookie of the year named Jerome Walton. The game would ultimately last until the bottom of the 14th, when Mike Trombley gave up a walk-off homer to Bobby Smith, who doesn’t appear noteworthy in any way, except that he happens to have the same birthday as me.

I didn’t watch this game when it happened–it came in the depths of the “dark ages” of Twins suckitude between the ‘91 World Series and the resurgence of 2001, a period when I stopped paying attention to the team.  One day, we’ll probably have access to a complete archive of televised baseball games, allowing us to go back and watch any play we want. For the time being, though, I just had to look at the play-by-play record and recreate the game in my head. It was a nice little stroll through the archives–so thanks, Joe Nathan, for giving me a nice opportunity to dig into a rare event and discover a forgotten moment in Twins history.

I think I agree with Gleeman that the Twins this year are, if not dead, then at least on life support. And that was before Jason Kubel got banged on the knee and left tonight’s game.

But right now I’m watching the team play the Rangers, and the Twins have just come back from a 5-0 deficit to tie the game in the sixth. They’ll probably lose, especially considering that they already went to the bullpen after four innings. But one thing I can say about this team, that I couldn’t say about any of the other incarnations of the post-1990’s, resurgent Twins: they can really hit. After watching the good pitching-bad hitters-weak division combination for so many years, I’d forgotten what it’s like to watch a team that’s capable of hitting its way back into a game at any time.

Now if only someone could hold on to that lead for a few innings…

[This one's written from my hotel room in San Francisco. ]

I ditched a day of the conference I’m attending to go see a game, and I’m glad I did. I’ve only been to five other major league parks–the old New York parks, both Chicago parks, and the Metrodome–but I’d say the Giants’ park is superior to any of them.  (Wrigley Field would be in the running, except that it’s always full of Cubs fans.)

It’s a beautiful park, with a view of the bay. It’s well-designed; even the cheap seats feel close to the field.  The architectural details are charming without being corny. In a nice, if ludicrous, California touch, there’s a guy walking around in the stands selling Frappuccinos.  And unlike Hennepin County, San Francisco had the good sense not to put their stadium directly in the middle of downtown, choosing instead a fairly remote part of the waterfront.

This is the only place on Earth that accepts the truth about Barry Bonds: he's awesome.

The first time I went to this park was two years ago, and that game turned out to be kind of an archetype of the Giants as they were at that time. Barry Zito pitched a shutout, and the only scoring came from a Barry Bonds home run.

Today’s game wasn’t quite so elegant. The Giants lost 5-2, even though they really deserved the victory.  For one thing, the Reds put out one of the worst offensive lineups I’ve ever seen outside of September. Yes, Brandon Phillips has been good, and Joey Votto is genuinely excellent, but the rest of these guys were pathetic scrubs at best. When your 1-2 hitters are Willy Taveras (.236/.272/.283) and Alex Gonzalez (.198/.250/.286), something has gone terribly wrong with your team.

Despite that, Matt Cain was bad enough to dig himself into a 5-1 hole. And then, in the bottom of the 7th, the Reds loaded the bases with two outs, and the pitcher’s spot came up. Aaron Harang had already thrown nearly 100 pitches. A base hit would have more or less put the game away. So naturally, Dusty Baker let Harang hit, and he predictably made an out to end the inning. Harang then proceeded to come out and load the bases with nobody out, whereupon he had to be immediately yanked for a reliever anyway.

Baseball is a game of probabilities. Over the long run, good decisions will lead to good results the majority of the time, while bad decisions will fail a majority of the time. But in any given game, we only see one outcome out of the whole distribution of possible outcomes. Many times, bad decisions are punished, but many other times, they are not–and a fool lives on to mismanage another day.

Dusty Baker oughtta buy this guy a drink.

In this case, Dusty Baker was bailed out by a Pablo Sandoval double play ball and a Bengie Molina fly out. But that doesn’t make his decision any less stupid. Had he pinch hit for Harang he might have put the game away. Instead, he let the Giants back into the game, and as a result Francisco Cordero came in for an easy and pointless save. But his team won. So because fate, or the laws of probability, were with Baker that day, no-one will remember how stupid his decision-making was. And he’ll go on making the same mistakes.

In the comments to one of my recent posts,  a commenter was confused about my post on Joe DiMaggio’s hitting streak:

Are you saying that there is a 0.003% chance that a .350 hitter would have this streak on any random 56 game stretch? Or in their career? I’d like to know what the chances that someone would have done it in the history of baseball.

This is a good question, and something I should have been more clear about. It was the former probability I was talking about: the chance of sustaining a hitting streak over any particular stretch of 56 games. But the commenter is right, the second question is the more interesting.

And it’s an answerable question, it just requires figuring out how many chances–how many runs of 56 games–there have been in baseball history. Luckily, before I started to contemplate figuring this out myself, a friend reminded me that someone had actually addressed this question recently (thanks, Andy G!)

Unfortunately, the results of that research kind of undermine the point I was making before. When the mathematician Steven Strogatz ran an extensive series of simulations of the history of baseball, he found that a 56-game streak actually wasn’t so unusual after all. You can see the popular version here, the detailed and geeky version here, and listen to an interesting podcast  discussion of all this from the show Radiolab here.

Strogatz’s conclusion was certainly a little disappointing to me, even though he tried to sugar coat the bad news by noting that even if the existence of a 56-game streak wasn’t that surprising, the existence of such a streak in 1941 was, given the levels of offense that year. But more interesting than that was something else Strogatz pointed out, a little article in the Baseball Research Journal by one Trent McCotter, who did another study of the frequency of hitting streaks.

You can read that article here, and a summary here. It’s an attempt to assess one of the assumptions in all of these hitting streak discussions, namely that a player’s probability of getting a hit in a game does not depend on whether they are on a hitting streak. In other words, there is really no such thing as being “hot” or “cold”, and these apparent streaks are really just arising due to random chance, like a coin coming up heads ten times in a row.