Crapshoots from a Different Angle

As longtime readers know (do I actually have any longtime readers?) I write a lot about uncertainty.  If I’m known for anything beyond a certain jovial nature when under the influence of alcohol, it’s for my oft-repeated (and I think proven) claim that the playoffs are a crapshoot.  A nice post from the pseudonymous tangotiger late last month made concrete the influence of uncertainty in another way I thought I’d share with you.

There are lots of ways to guess what some guys performance is going to be this year: ZIPS, Steamer, Pecota, etc.  Are they any good?  Tangotiger created a prediction method he called Marcel, whose object was to be what he called the replacement level of prediction:  something that was guaranteed not to be awful. 

His prediction for any quantity (hits, homers, ERA, whatevs) has three components: (a) average the last three years of data with higher weights on the most recent years; (b) add an aging effect; (c) regress the result towards the mean.  And it turns out that this method, which can be implemented in about ten minutes in a spreadsheet, is pretty good!

And while [ZIPS and Pecota and Steamer] may be better than Marcel, whatever advantage they have is going to be slight. I mean, we are talking about Marcel winning probably 48% of the individual head to head matchups. Everyone is really fighting for that 2%. If Marcel is an 81-win kind of a forecasting system, everyone else is an 82 or 83 win system. 84 if they’ve really tapped into something the others haven’t. There’s just not that much you can do, such is the power of Random Variation.

The “power of Random Variation” is just another way of saying what I’ve been saying about the playoffs.  There’s just a limit to how much any predictive model can do if the outcome isn’t determined by the past, but only influenced by the latent talents we can observe by looking at the past.

Two Out Hitting

I’ve never discussed this much, but Chip spends a lot of time discussing RBIs with two outs.  Every two out hit with runners in scoring position is treated as either the height of clutchness or the apotheosis of Why We Can’t Have Nice Things, depending on whether the hit is executed by the offense or allowed by the defense.  He has been known, for example, to describe a few Swanson two-out RBIs as INCREDIBLE!

But it’s not.  Just as we have pretty good evidence that clutch hitting isn’t a skill (conditional on having reached the major leagues) we have some pretty obvious evidence that two out RBIs aren’t that extraordinary.  Look at the following table, whose source data is the entire Retrosheet event database:

OutsRunsRuns (excluding sacrifices)
0340,626 (23%)314,922 (22%)
1597,945 (40%)535,235 (38%)
2569,411 (38%)569,411 (40%)

Once you exclude sacrifices (runs which are impossible with two outs) two out runs are actually the modal result. 

The data-savvy among you might notice a problem with this result, in that there is no control for opportunities.  But of course every at bat is an opportunity to score a run, as Chip beats us over the head with every night.  But we can take that seriously as well, with this table:

1st and 2nd0.2520.2700.269
1st and 3rd0.6210.5370.319
2nd and 3rd0.5850.5080.368

This table gives the expected runs on a batting play with the given base situation and number of outs (sacrifices have been dropped).  For bases empty, man on first, or man on second, two out hitters do slightly better than other hitters.  For men on 1st and 2nd, the one out hitters do very slightly better, but not significantly so.  Once there’s a man on third, the no out and one out hitters do much better, since they can bring home a run with a groundout, but note that with 2 outs and the bases loaded, batters bring an extra half-run over one out and men on first and third and 0.77 runs more than the one out and men on second and third. 

File this one under: Clutch Hitting Is A Myth – Exhibit 45a.


So the alltime franchise record is 19 in a row set by the 1891 Boston Beaneaters. I’m old, but not quite old enough to remember the 1891 team — plus they were in Boston. Furthermore, even Retrosheet doesn’t have game-by-game records from back then, so our knowledge of the specifics behind that streak are limited.

But a couple of interesting things. First, this streak was 19 of the last 20 games of the season. When the streak started, on September 16th, they were 5.5 games out of first. After the 14th win in a row, they were tied for first. After the next 5 in a row, they were 3 1/2 games up with one game left — a game they lost. That’s a nice stretch run.

Second, that team not only had 4 eventual HOFers (John Clarkson, Joe Kelly, King Kelly, Kid Nichols) and the alltime MLB leader in errors, Herman Long (1,096 errors in 1,882 games) but also had one of my alltime favorite names: Pretzels Getzien, although he was released on July 16th and wasn’t around for the streak. (The 2000 Braves who won 15 in a row had only 3 HOFers: Chipper, Maddux and Glavine. The jury is still out on Andruw, though. That was Smoltz’s Tommy John year.) I leave as an exercise to the reader how many HOFers we’re looking at this year.


Question from JoshKinNJ:

Do we think Dansby has taken the leadership mantle from Freddie?

Answers, from me:

I have no idea, but I am absolutely sure that the concept of “leadership” in this context is sufficiently elastic as to be meaningless, or, to state it another way, to mean whatever it is you want it to mean….Group culture is a thing, peer effects are a thing so I’m never going to say there’s nothing there. But the notion that a team neads a “leader,” or two “leaders” or seven “leaders,” some of them hortatory and some of them quiet, etc, etc, is an explanation in search of a problem, IMO. There’s still a lot we don’t know about how some teams (in general, not just sports) function better than others, but I strongly suspect that “teams need peer leaders” is the sort of thing created to give sportswriters and broadcasters something to talk about.

After some pushback, I promised I’d write about this, but ububba’s answer pretty much beat me to it:

And we were a sub-.500 team until the first week of August last year with Freddie.

I don’t know the answer. I know that some teams (the early ’70s A’s, the mid-late ’70s Yankees) fought like cats & dogs, had guys who absolutely hated each other, and they won titles. (Of course, the best player on both teams was the same hated guy.) The ’86 Mets had a guy who thought he was the leader (Gary Carter), but everyone laughed at him behind his back, and they won a title.

Each situation is different & I’m not sure the notion of needing a leader can be proven one way or another. I mean, it seems like the 2 best players on the Braves the last few years didn’t really dig each other and it wasn’t like Freddie ever really convinced Ronald to act like a pro.

But because the Internet has given me a few more column-inches to fill, let me elaborate a bit more.  There are all kinds of players cited as leaders without any actual playing exploits to match their “leadership.”  Pablo Sandoval and Steven Vogt last year are paradigms of that.  Yet somehow, when we lost all that leadership from Pablo after he was… ahem… traded, the team did a lot better.  There are all kind of reasons for this which might simply mean that the loss of leadership is offset by, say, great hitting by Eddie Rosario, but nobody wants to talk about it because the leadership talk was really just empty nice words about Pablo Sandoval, who people seemed to like having on the team.  And that’s fine:  all things equal I’d rather work with people I enjoy and people I respect than with people I hate or people I think are stupid – maybe I’d even perform better in a good environment than in a bad one.  And maybe having someone I respect tell me how to act when I’m acting inappropriately would help.  Or maybe having someone who “leads by example” to copy would help.  And maybe that job is better performed by a peer than a manager or coach in certain circumstances.  But it all sounds pretty squishy when you put it that way, right?  And it is. 

Players vary as to their needs.  They vary in age, in experience, in language skills, in temperament, in work ethic and in any number of different ways.  The notion that they need a particular peer to guide them is, at best unproven, and at worst, utter crap.  Even if very, very few players are self-reliant, what they need to rely on will vary, and the notion that there’s a guy with a mantle (Mickey?) that is required by every team or even helpful to any particular team seems really really doubtful to me.   

So what we’re left with is a good story.  Reporters and broadcasters love telling good stories, and they love talking intangibles, because they can never be wrong.  So the Braves went four months with Freddie’s leadership last year and were profoundly mediocre.  Then he stepped up his leadership (Did he?  How?) and we’re World Champs.  Now he’s gone and his leadership is gone, so unless you want to admit that that was all just hot air, you need to find a new leader.  Dansby?  Sure!  Why not?  Because it’s all just hot air.

But just because it’s hot air doesn’t mean Freddie didn’t have a big influence on Ozzie and Austin and (somewhat more controversially, unless you agree he was misquoted, Ronald).  But maybe having instilled the virtues he was capable of instilling, he isn’t needed any more.  Until the team forgets those lessons, they don’t need a leader.  Maybe.  Or not.  I don’t mean to pooh-pooh the notion that peer leadership doesn’t matter.  All I mean to say is that the notion that  team is better owing to the intangibles of any player is a hypothesis to be demonstrated, not a lesson in moral uplift.