THE BOOK--Playing The Percentages In Baseball

OK, so they are slightly over-performing their pythag record…

Mike Scoscia angered the Gods when he decided Jeff Mathis was better than Mike Napoli.

I blame poor clubhouse chemistry.

They’ve outperformed Pythagenpat by 1.5 games through yesterday’s results.

scored 561
allowed 590

Pythagenpat record 60.5-66.5
Actual record 62-65

That would make this year fall in with the 2004-2005 range for them rather than the crazy results of 2006-2009.

One thing that I think “Angels > Pythagorean” backers are usually glib about is whether they are claiming that

a) actual performance > projections
b) actual RS/RA > component RS/RA
c) actual WinPct > component WinPct (Pythaogrean)

or some conflation of the three.  Usually it is whatever has the greatest contribution to the anomalous result that is touted as being the Angels secret sauce.

#4, right.  The thing that most people talk about is the great managing, the great coaching, the great base running, the great chemistry, etc., that allows the Angels to vastly outperform their pre-season team w/l projections as determined by the (credible/sabermetric) projections of the individual players.

What happened to that?

I am being facetious of course, as almost anything can happen in 125 games by chance alone.

But, as usual, if anyone wants to make any claims at all going forward, like a certain team or player is likely to (substantially) outperform its X (fill in the blank), where X is determined by a credible analytical process, you have my email address if you want to put your money where your mouth is…

Well, if you’d have been given 2:1 odds they’d underperform their pythag you’d have lost. ‘slightly’ is still >0.
7 straight years of over-performance seems statistically significant to me.

Mettle/7, I believe MGL is referring to articles like this one in addition to simply what the sign is on the difference between their actual record and their pythag record:
http://baseballanalysts.com/archives/2010/02/pecota_and_hist.php

Still not sure if I get the point.
The past 6 years of data show that the Angels outperform PECOTA or pythag by an average of 5-10 wins. This year they’ll outperform it by ‘only’ a win or two. Even if they underperformed by a win, as should be acknowledged, that’s not “evidence” they don’t over-perform. It’s just one data point, so the gloating is at the least ironic.

Furthermore, depending on how you frame your hypothesis, these results are also statistically significant, even if you include this year, in term of rejecting the null hypothesis of pythag or PECOTA. If you asked the question specifically about the Angels overperforming (i.e. one-tail, not correcting for multiple comparisons) your p=.017. So I think the gloating is also wrong, or at least on shaky ground.

I agree with mettle.  This is like Tango saying that Strasburg’s performance proved his point about projections.  The thesis and the data point don’t fit.

MGL- Do you think that the differences between actual record and Pyth record is just luck? I don’t know the answer. But should be interesting to know which factors can explain those differences.

The first thing that comes to my mind is that a team with a good bullpen should win more close games than a team with a bad bullpen.

I don’t know, maybe I am wrong but would like to make a study about which factors explain those differences.

Dayan/11
The data can answer your first question, but it depends how you frame it.
If you ask, will there be some team over the history of baseball (i.e. 100’s of samples) that outperforms PECOTA by as much as the Angels have? Then no, you can’t rule out luck.
But if you say, will this specific Angels team in the recent period where we have these data outperform their projection by this much, then the answer is it’s unlikely (<2% chance) that it’s luck.

I feel like we’re missing something...obviously I don’t know what it is

It’s a generalization, but I always felt the Angels beating their Pythags was in part due to their over-aggressiveness on the bases. They often run on contact, take extra bases and force throws, and Dino Ebel likes to impersonate Pete Townsend in the third base coach’s box. They’re just a slower outfit than they’ve been in recent seasons. This year they have 78 steals, and it might be the first time since 2000 they’ve finished with less than 100. It’s also the first season since 2000 they’ve attempted fewer than a steal per game, and they’re working on their lowest success rate of the decade.

How does baserunning translate into beating Pythag?  Good baserunning translates into scoring more runs.  Which means Pythag knows about those runs.

I don’t think I understand the point.

1) The Angels beat their Pythag every year
2) The Angels are beating their Pythag this year
3) So this is a great opportunity to suggest that there is no skill involved in beating one’s pythag?

I don’t get it.

The Angels’ Pythagorean gap is largely the result of 2008, when they beat it by 11.9 games.  For 2002-09:

-3, -3.3, .9, 1.3, 4.5, 3.9, 11.9, 4.3

The standard error is around 3.95, so only two seasons outside of 2008 are even a little bit eyebrow-raising in isolation.

I do find it amusing that the Angels team that had the most on-field success is one that bucked the “Angels way” and underperformed their Pyth.

John/16: I think he’s trying to say that the Angels *only* beat their pythag by 2 this year, so all those crazies who said that they beat it by 12 every year are wrong, wrong, wrong.
e.g. It’s only 73F today and not 90F, so global warming = not true.

The “beat the projections magic” is currently on the DL, actually.

A bit more seriously, the “ability” to beat Pythag is often tied to (1) winning close games and (2) losing blowouts.  The former depends to some degree on having good relievers that are well-leveraged, and the latter depends to some degree on putting the poor relievers into games with low leverage.

Historically, the Scioscia Angels have had a good back of the bullpen and he has leveraged them well.  However, this year (and last year, too) the bullpen has not been very good on the whole, and the “good” relievers aren’t as good as they were in days of yore.

Also, Scioscia does tend to throw his lousiest relievers out to the wolves when a game gets out of reach, and that hasn’t changed, so that factor is still in play.

Even IF a team had an ability to out-win Pythag, we would expect it to regress, so that (even if we suppose, for the purpose of the conversation, that some teams have this “ability") along with the factors above would explain why the Pythag is being exceeded by only 1.5 wins so far this year.

As far as the run-scoring outperforming run estimators—I don’t know if it has for the Angels, historically, but if it has it’s because of a combination of clutch hitting and productive baserunning (NOT SB/CS, but advancing on hits and outs), which in combination means that Angel hits may tend to be more valuable than the average hit.  As pointed out above, the Angels are a lot slower this year (the SB/CS is evidence for this), and their RISP performance is roughly the same as their overall performance.  So we would expect run estimators to be more in line with their actual run scoring for these reasons, regardless of the regression that we would expect anyway.

In terms of the projections, the DL line is legit.  Kendry Morales was a big loss to the team, and though Joel Pineiro is not a star, losing him was a death knell for the season.  Scott Kazmir lost time to the DL (and pitches like he’s been pitching hurt even when “healthy"), also.  If you told me before the season that the vast majority of Morales’ PA would go to a combination of Jeff Mathis, Bobby Wilson, and Maicer Izturis, I’d have said that the team would struggle.

Sky, I don’t know. I know Pythag is just a formula that translates RS and RA into a predictive winning percentage, and I know the Angels have beaten this formula for most of the decade. Like most, I’m curious why. I’m just throwing out ideas. Something distorts it…

I think my original comment got eaten.

The Angels as Pyth outliers phenomenon is largely due to 08, when they exceeded their Pyth by 11.9 games.  The rest of their 2002-09 results are far from mindblowing:

-3.0, -3.3, .9, 1.3, 4.5, 3.9, 11.9, 4.3

Keeping in mind that the standard error is around 3.95, there aren’t many individual results that would raise your eyebrows in isolation.  Taken together, it’s a little more impressive, but a lot of people like to play multiple endpoints with it (and thus excluding 2002 and 2003).  And (apropos of nothing but amusing nonetheless) is that it was 2002 in which they actually won the WS, despite not displaying the pluck or whatever it is that allows one to outperform Pyth.

The issue with every explanation being offered here and elsewhere is that none are quantifiable or testable.  They have good baserunning or clutch hitting - great!  How good?  To what degree does that regress?  What is the mechanism by which that translates to beating a projectiong/scoring more runs than component/wining more games than component?

And, most importantly, find the comparable teams doing comparable things and show me that they are also beating projections/scoring more runs than compnent/winning more games than component.

Yes, it’s the bullpen--then it’s all luck.  I have a theory that pythagorean variances have gotten more extreme in recent years, due to an increased use of the bullpen, but I actually haven’t found that to be true when examining the numbers.  I’m still clinging to the theory, however.

It’s also interesting Pythag seems to be in line for the most part this season across MLB. There are really only four significant variances: Houston (+7), Pittsburgh (+6), Kansas City (+5) and St. Louis (-6).

To tell you the truth, I’m not sure how much use the Pythagorean has on the team level.  To describe what actually happened, we have actual wins and losses.  “True talent” only really applies for projection.  And, sure, you can say a team’s W% in year N+1 will regress whatever it is toward their Pythagorean estimate for year N, but even that assumes that (1) they have substantially the same personnel, (2) that that personnel is not centered on a part of the aging curve that would cause us to project abrupt improvement or decline the following year, and (3) that the personnel on the team, in season N, as a group performed at their true talent level.

Now, if you really want to create a team projection, you’ll of course be doing individual player predictions and summing them, and THEN using the Pythagorean to estimate that team’s wins.  If you don’t have a huge projection database at hand (like most of us), you can maybe use season N’s Pythag as a proxy or starting place ... but it’s probably not going to be a final answer.

So I wonder if the real answer to “Do teams have a true ability to surpass their Pythag?” isn’t “Does it matter?”

If you want to beat pythag, then make sure you don’t score unnecessary runs.  In other words, don’t hit a grand slam to win a game when all you need is a sacrifice fly.  If only Kendry Morales had paid attention.

BWV, there’s been a lot of interesting work done on whether a team’s true wins and losses are a better predictor of their record during the rest of the season, or if their Pythagorean record is.  Clay Davenport did some interesting work on this.  I believe Pythagorean was until about 2/3 of the way through the season, when actual wins and losses became the better predictor.

Plus, you can learn a lot about baseball and teams by examining their pythagorean variances.  It may not be “useful,” but sabermetrics doesn’t have to be just useful.

Well, Mettle (and anyone else), if you want to give me 3-2 odds that the Angels (or any other team) will beat their pythag and/or a consensus (from several sabermetric sources) of their w/l projections, either from now to the end of the season or for next season, just let me know how much.  As usual, put your money where your mouth is.  That is the true test of conviction…

Nobody’s going to take that bet.  Doesn’t mean nobody thinks the Angels will or won’t beat pythag over whatever period specified.  It is possible to think they are slightly more likely to do so but not so certain that they’d give you 3-2 odds.

----"As usual, put your money where your mouth is.  That is the true test of conviction…”

No, to me it’s more a test of how much disposable income that a given poster has…

I’m not suggesting that the Angels DO have some inherent Pythag ability. So no, I won’t bet you on that.

It’s just that this post seems to suggest that they don’t have the ability, and then uses the fact that they’re beating their pythag record to support that claim.  It’s confusing, which is what I wrote.

It is not unlike Tango saying that it was silly for people to project Strasburg to do so well, and then using the fact that Strasburg performed very well to support his point.

These arguments (if that’s what they are) seem sort of...weird.

As per #31, I don’t have the cash handy for a full on vegas style bet, but I’m happy to bet a token amount and bragging rights on the Angels outperforming their PECOTA in 2011 on the one condition they’re above 500. Perhaps a front-page contrition would do.

The bigger point Jon/32 and I made stands, though.
It’s very odd/irresponsible/ironic/silly to highlight the one data point you did as proof of anything at all.

I SAID after the initial post that I realize that they were (marginally) beating their pythag record and my major point was all the hoopla about them beating their pre-season projections!

Do you guys not read English?  Read my #1 and #6 posts!

Can I collect on this, btw?
http://www.baseball-reference.com/teams/LAA/2011.shtml

+1 is still > 0.

"As per #31, I don’t have the cash handy for a full on vegas style bet, but I’m happy to bet a token amount and bragging rights on the Angels outperforming their PECOTA in 2011 on the one condition they’re above 500. Perhaps a front-page contrition would do.”

I don’t really remember the details of this thread, but two things:

One, I don’t see a bet in there from you, nor do I see an acceptance from anyone, so what exactly do you want to collect?

Two, and more importantly, if you only include > .500 records OF COURSE you have an edge in that bet!  You want to eliminate all the “unders” when the team is below .500 and you want to take all the overs?  Who are you trying to scam?

We can roll it over to 2012:

A locally brewed 6-pack (Anchor Steam) on whether Scioscia’s Angels beat their pythag.

In?

Sure. As much as I hate betting when I don’t have an edge.

+1 in each of ‘10 and ‘11.  Think they are doing the Jack Morris playing to the (pythag) score thing?