The Dying Gasp of the Luddites

If you're wondering what kind of impact new analytics are having on the baseball landscape, I commend to your attention this dying gasp of Luddite literature from Paul Daugherty of the Cincinnati Enquirer. His half-hearted attempt to wish away knowledge is an ode to new analysis in its way. So tepid and conditional are his objections that we're left to conclude the argument may finally be over.

You can read his diatribe yourself, but I'll summarize it here:

• Baseball stats can't measure professionalism. No, they measure results.
• Some new stats, like Deserved Run Average, are complicated and I don't understand them, so they must be useless. When's the last time you measured ERA yourself?
• The number crunchers think digits can replace eyeballs. Show us the guy.
• A new measuring tool only measures what it measures, so it must be useless. There go all stats.
• If a new measuring tool tells you what you already believe, who needs it. If it contradicted what you believed you'd similarly scoff.

This is now the state of discourse from the hidebound. Notice that Daugherty admits new numbers have their place. Notice that he recognizes all front offices have had to adopt advanced metrics or preside over withering franchises.

But also notice that old false dichotomy between measuring performance and watching games. Would the seamhead who disputes the value of visiting the ballpark please identify himself, because I can't find him. This straw man argument forms the bedrock of most anti-progress rants, as if SABR rattlers are walking the countryside like Johnny Saberseed planting the notion that numbers can capture all the nuance of the game.

But numbers can tell you this: for all his professionalism, Marlon Byrd is batting .212 and playing a low-value defensive position (left field). We could watch Byrd and agree that he's a good guy, a hard worker and a positive influence on his teammates. But we'd also notice that Father Time has taken a chisel to his skills, and left the Reds with an out machine.

His objections amount to a surrender, really. It's like the last wail of an infant before they go back to sleep. The innumerate want to register their discomfort one last time before the next BABIP discussion.

Sorry, Paul, to burden you with added insight into the game, with better predictive tools, with interesting information about plays and players. You're free to ignore all that while you watch a .212 hitter hustle. Say hi to Fred and Wilma for us back in Bedrock.

Best & Worst New Metrics Part III: OPS

The days of explaining why batting average, home runs and RBIs don't adequately capture an everyday player's performance are long gone. Even the retrogrades who cling tenaciously to the old numbers recognize in their hearts why on-base percentage and slugging percentage are more comprehensive and why RBIs don't tell a useful story.

Adding on base and slugging gave us OPS, the shorthand statistic of choice to seamheads in the seamheady early days of the sabremetric movement. But analysts cannot live by OPS alone; indeed, OPS and its close relations, like OPS+, now long in the tooth, have become something of an anachronism, rarely used to make anything but the crudest point. 

The reason is simple: not all OPS is equal. Seamheads have long known that OBP is more important than slugging, about 20% more when talking about run creation. So simply adding the two numbers devalues on-base percentage.

OPS is not park-adjusted, position-adjusted or competition-adjusted. The narrative it creates doesn't describe whether a batter hits same-handed pitchers or has to be platooned. It's also largely irrelevant when discussing leadoff hitters, whose OBP is invaluable and whose slugging is not very valuable at all.

OPS was good to the analysis community in its prime, but has lost a step and is now little more than a veteran presence off the bench, backing up starters TAv, WAR/WARP and their ilk. We thank OPS for its contributions, particularly for introducing fans to better ways to measure players. and wish it the best of luck in its future endeavors.

Thankful

On Thanksgiving, it is proper to take a moment to ruminate over our many blessings, such as that we're not a turkey, or a Royals fan. Anymore.

My gratitude to others:

I give thanks for the "traditionalists" who claim to hate baseball statistics and then misuse them to justify foolish positions. They make possible the endless arguments that bring joy to the interstitial spaces between pitches, at bats and innings.

I give thanks for the Luddites who refuse to learn anything new about baseball, as if willful ignorance is a life strategy. A weak beam like me positively shimmers in contrast to dull mattes like that.

I give thanks for the innumerate journalists who assume that if analysis is beyond their understanding it can't be valuable. They allow a third-rate amateur blogger to produce more insightful opinions in his spare time than they do vocationally.

I give thanks for the slow learners who aren't prepared to rush into the embrace of new statistics developed just 33 years earlier and tested rigorously since, the kind of examination that exposes the irrelevance of RBIs and pitcher wins. They allow a new breed of open-minded fans to converse at a higher plane than they can fathom.

I give thanks for the robotic seamheads who forget that real humans play the game. They give cover to lunkheads still dog-paddling in the shallow end of the knowledge pool and spur the debates that I so enjoy.

But mostly, I give thanks for being born rich and free in America where I have the luxury to waste time on frivolous pursuits like this with a full stomach and the security that it will all be there tomorrow.

The Truth About Seamheads

A visit to New York gave me an opportunity to watch the terrific Met TV broadcast team of Gary Cohen, Keith Hernandez and Ron Darling for the first time.

(Full disclosure: Cohen was a college broadcast partner and is a longtime friend. Darling was a subject of a couple of those broadcasts. My friend Mitch has a brother named Keith.)

They talk constantly of on base percentage. Hernandez stresses plate discipline. They seem to understand the massive role of luck. And yet even they -- Hernandez particularly -- (sorry, Mitch's brother) repeated constantly some old shibboleths about new analysis.

Here are some common misconceptions, some of which gained currency on the telecast I saw and others that I hear regularly.

1. "Moneyball" represents stat guys.

Moneyball was a book written by Michael Lewis (not Billy Beane) and a movie sort-of depicting the book. It's about the GM of the small-market A's exploiting inefficiencies in the baseball labor market, particularly with respect to on base percentage. Beyond that, the book and movie (which took the kinds of liberties with the facts that get new head football coaches fired) have no relationship with new baseball analytics.  Calling the work of new analysis "Moneyball" suggests ignorance right from the get-go.

2. Seamheads don't care about batting average.

Batting average is the apple in apple pie. The pie is on-base percentage, which includes other important ingredients, like walks, the crust of the pie.  Carlos Pena can bat .198 and have the same on base percentage as Ichiro, batting .320. In that case, batting average isn't very telling. But if Pena could hit .250, he'd be a star. When Ichiro hits .280, he's a scrub. You need the apples and the crust to make a good pie, and if you ignore the crust, you're not getting the full pie.

3. Stat guys don't like stolen bases.

How can anyone not like stolen bases? Moving a runner from first to second is precious. Stat guys don't like caught stealings. Giving up an out is way more precious. So if you're going to break for second, make sure you've first paid homage to the patron saint of thievery.

Beyond that, even stat guys realize that always playing by the book makes a team predictable, assuming the other team has the same book. (They do. It's on their Nook.) Teams have to bluff ocassionally lest they become transparent.

4. Sabermetricians think they have all the answers.

All the seamheads I read recognize that most of the game is invisible. Hell, R.A. Dickey is 10-1. Dodger backstop A.J. Ellis has a .436 on base percentage. Twins rookie Scott Diamond from Binghamton University has a 2.13 ERA and a 5-1 K/BB ratio. Who had that parlay?

5. The new analysts think the whole game can be reduced to numbers.

What new analysis has shown us is exactly what can and what can't be quantified. Many old schoolers believe RBIs demonstrate how "clutch" a player is. The new analysis proved that "clutchness" is hard to quantify, and however it might be, RBIs are completely unsuited to the task.
 
6. Stat dufuses (the plural of dufus) think they can measure fielding even when their conclusions are contradicted by observation.

Hidebound traditionalists don't realize when they are observing with their hearts. Sometimes the new defensive measures help us to see that. But in 2012, that area of research is still in its adolescence and should not be trusted with the car keys. Right now, the defensive metrics are a tool to help us check our work, but they can't completely replace our eyes, even with all our inherent bias.

7. Seamheads never played the game. They can't possibly know as much as guys who did.

You've obviously never listened to the profound insights of Rob Dibble. Why should a monosyllabic cretin with a live arm see the inner workings of the game more clearly than someone who studies it?

8. The stat goobers don't understand the value of team chemistry.

Maybe the traditionalists can quantify it for the rest of us. Maybe they can explain how the squabbling 1972 A's won the World Series. And then repeated in '73. While they're at it, they might want to do the same for Billy Martin's dysfunctional '77 Yankees and the same primadonnas under Bob Lemon in '78. Those '62 Mets and 2003 Tigers had more good chemistry than Brad Pitt's high school reunion. How'd that work out for them? (NY and Detroit, not the reunion.)

No workplace functions as well when employees prefer splenectomies to each other's company. I wouldn't sign Manny Ramirez to my team with Jamie Dimon's money. But chemistry doesn't replace talent. If you want Veteran Presence and Clubhouse Leadership, hire good coaches.

9. Even with all their numbers, the sabermetrics numbskulls aren't any better at predicting who will succed in the Majors than the teams themselves.

That's because the sabermetrics numbskulls and the teams are one and the same. Every team in the league uses sabermetrics (in addition to scouting) to evaluate talent and avoid losing ground to everyone else using sabermetrics.

10. SABR monkeys don't put much stock in saves, pitching wins, RBIs and other traditional statistics.

Whoops, that one's true. Research shows those statistics correlate poorly with what they are attempting to measure.

There are almost certainly more of these, but I'm on the way to a Met game.
b

Bollocks to Moneyball

Upon his retirement, Tony LaRussa, a successful innovator and sure-fire Hall of Famer, "ridiculed 'Moneyball' and its emphasis on statistics over human scouting and observation" according to the Associated Press.

Critics have smiled on Moneyball, the movie, but LaRussa has described why I was uneasy about the book and have no plans to see the film.

Moneyball, the book, cast statheads as cowboys and scouts as Indians (feather, not dot or Cleveland varieties). The analogy is apt because the relationships are far more messy and complicated. Scouts, like seamheads, are necessary but insufficient for the proper functioning of a Major League baseball franchise.

The real debate was never between stats and scouts, it was between useful stats and misleading stats. That's why the St. Louis Cardinals employ number-crunchers, just as every MLB team does, to help reveal some of the game's hidden insights. Over the years, new analysis has exploded long-held myths, including Larussa's signature creation, the folly of saving your best reliever for ninth-inning mop-up when you could employ him to put out fires in other innings. Maybe that's why LaRussa is sore.

Or maybe he's sick of the false dichotomy. LaRussa knows better than Bill James' personal guru how Matt Holliday feels today or whether Chris Carpenter can go on short rest or if John Jay can be relied on to drop a bunt. Team number crunchers have a better handle on the general efficacy of a bunt in a given situation.

In fact, LaRussa has been most renowned for his pitching match-ups, a strategy fully infused with both scouting and stats. The guy with the binoculars on LaRussa's left shoulder whispers that the opposing batter hates the high heat thrown by the righty in the pen. The SABR dude on his right shoulder notes that the batter has a .650 OPS against southpaws in day games. LaRussa's particular genius was to be open to surprising information, synthesize it and act on it.

Moneyball correctly cast Billy Beane and his band of mavericks as explorers who had found new worlds and were benefiting from the small competitive advantages they conveyed. There is absolutely no question that statistical research can unlock mysteries and disprove old customs in the game, though those advantages are becoming more nuanced and less impactful as the practice has become widespread. Likewise, there's no doubt that observation can reveal nuances unnoticed by the statistics. But unlike stats, observation can be corrupted by human psychology -- things like confirmation bias and cognitive dissonance -- that has no power over the null hypothesis or a two-tailed test.

Here's a World Series example of the two at work: Mid-year, this blog noted the amazing performance of young Ranger starter Alexi Ogando, who twirled a 9-3, 2.92 line with a .591 OPS-against. Spectacular, but with a .241 BABIP and some other component statistics pointing the wrong way, I suggested a decline was in order. Furthermore, Ogando had reached his professional career high for innings by July. The Rangers  noted this, not to mention his 4-5, 4.48 performance in the second half, and downgraded him to the pen, where innings aren't so readily available.

A statmonger could guess that Ogando was gassed. I'm betting that keen observers like LaRussa and his staff could see it. In any case, the Cards feasted on Ogando for 14 baserunners and four runs in only seven outs. (To be fair, he was great in 9.2 innings of AL playoff service.)

The bottom line is that teams fare best when statisticians and scouts respect and rely on each other for confirmation of what they believe they've found. It's time to forget the false dichotomy perpetrated by Moneyball and just accept stats and scouts for what they are. 

b

Statheads on Crack

Sabermetrics have transformed baseball analysis and helped regular fans like us understand the game differently and better. Computer-fueled statistical analysis has punched holes in old shibboleths and established new, research-supported truths. Often, they tell us that our eyes deceive us.

The stat geeks, spreadsheet-munchers and number nerds responsible for this revolution deserve our appreciation not just for devising new ways to measure things, but to continue to devise better ways. Ten years ago, when seamheads were gestating the concept of dissociating pitching from defense, the results were often best left ignored. They have improved steadily, thanks to MLB's ability to catalog every single pitch, to the point where we could, in broad strokes, predict the decline of Ubaldo Jimenez in the second half of the season.

Sometimes, however, seamheads forget that their work is a journey and not a destination.They are beguiled by half-conceived statistical formulae whose rough edges have yet been smoothed by years of inquiry, testing and polishing. They treat the results as gospel, rather than as suggestions tempered by the adolescent state of the art.

Two years ago, Oriole farmhand Matt Wieters was touted by Baseball Prospectus as the greatest catching prospect since Zeus. (Lightning bolt for an arm, immense two-strike power.) Using their proprietary projection system to examine Wieters' minor league and college careers, it projected him to hit .311/.395/.544 with 31 home runs in 2009 and be worth 60 runs more than a replacement-level catcher. Their system, called PECOTA, compares player performance to similar players of the past and sets upper and lower boundaries on expected future results. What I've listed is the weighted mean -- or most likely -- projected rookie numbers for Wieters.

What seemed obvious to me at the time, and what you've no doubt realized, is that such expectations for any rookie are patently absurd. They fail the smell test. While we should all be open to revealed truths, we must trust our eyes until the contradictory facts congeal into a proof.

In fact, Wieters showed great promise as a 23-year-old in 2009 with a .288/.340/.412 line and nine home runs in half a season. That's still above average for a receiver, but a far cry from the projection that BP writers defended as if it were the 11th commandment. (Honor they PECOTA projection and have no other conjurings before me.) Wieters has sagged even further from that unwarranted esteem this year to .253/.326/.385 and 11 home runs in full-time duty.

I love Baseball Prospectus enough to pay good money for it and read it everyday. But they treat PECOTA as destiny and use it to project team performance -- via a Monte Carlo simulation of a million permutations -- at the start of each season. Then they update their projections as the season unfolds.

The opening season projections amount to on-screen enlightenment. Often, BP sees things I don't. They correctly projected declines for the Dodgers, White Sox, Tigers and Angels. They're a good guide to the upcoming season.

It's when Dr. Jekyll starts reading his own press releases that he becomes Mr. Hyde. Entering the 2010 campaign, BP projected that the Cardinals had something like an 80% chance of making the playoffs. My reaction then was, how could that possibly be? I understand that the Redbirds were favorites, but no non-Yankee contingent can ever be an 80% lock before a pitch is thrown. Imagine a season in which Albert Pujols and Chris Carpenter are cut down by injuries and you can see that such a projection is patent nonsense. But the BP brain trust wrote as if October in St. Louis was a fait accompli.

Two weeks ago, BP had the Braves as an 89.6% lock to play in mid-October, despite a three-game deficit to the surging Phils and just a two game cushion over San Francisco for the Wild Card. Once again my olfactory nerves shuddered. Atlanta took three pops to the chin from their division rival and fell into a flat-footed tie for the Wild Card with nine or 10 games left. BP was still offering nearly 3-1 that the Braves would take the fourth playoff spot. (Their remaining schedule is weaker than San Diego's and San Francisco's.)

I understand why BP thinks the Braves should be favored, but I doubt Atlanta residents are booking their playoff tickets quite yet. As I write this, Chief Nakahoma's nine is a game back of San Diego in the loss column with six to play, having just lost two of three to execrable Washington. How is it possible for a team to be a 90% lock for the playoffs if a 4-6 stretch puts them on the outside looking in?

Perhaps some day the statxperts at Baseball Prospectus will improve their tools and better project playoff odds. Or (more likely, in my view) predicting the future will continue to fall outside the purview of mere humans, regardless of their statistical acumen. In any case, I'll continue to treat sabermetrics as a toolbox necessary to understanding the game I love, not as perfection itself.

So I'll be watching the games this last week and ignoring the predictions.
b