Sunday, July 25, 2010

PART I - Advanced Experimentation 410:
The Zdurienck Supremacy, or Containing an Implosion  

Change is inevitable. Change for the better is a full-time job. --Adlai E. Stevenson

There's a powerful example of Baseball's superior management abilities in the horrendous cherry-pie time of the 2010 Seattle Mariners. In spite of the obvious fact that Ms General Manager Jack Zdurienck's plan was an awful, unmitigated black hole from which no light escaped, it equally shows his innate genius and why he's a noble exemplar to follow.

┬┐How can a total black hole represent genius management? For two reasons, the first of which I'll discuss in this Part I.

In 2009 the Mariners surprised everyone who knew something about baseball by finishing the season at 85-77. The underlying facts were a little gristlier...the Ms were outscored by their opponents 640-692, which the Pythagorean thumbnail indicates should generally create a win-loss record closer to 75-87, 10 wins below their actual result. This projection was farther below an actual result than that for any other team in the 2009 majors. A large deviation like this in either direction is generally accepted by analysts to presage a following-year performance that will conform more closely to the Pythagorean projection than the win-loss record itself.

NOTE: This Pythagorean projection suggestion for the following year is not meant to be universal, nor is it universal in practice. For example, the 2009 San Diego Padres had a record of 75-87, but that lacklustre record reflected 8 more wins than their 67-95 projection based on their runs scored and allowed. This suggested that the Padres' overperformance probably would go away in 2010 and all things being roughly equal, 67-95 ball would be more indicative of their prospects than 75-87. But au contraire, while all things are roughly equal for them, but they have had a significant performance UPtick: as of today, if you project out their current W-L ratio to a whole season, they have been playing 94-68 ball, way far better than either their actuals or projected.

Pythagorean projection is a general rule. The exception is never the rule and all rules involving human behavior have exceptions.

The Mariners, to build on their 2009 success (which would be staying on the good side of .500, with a dream of making the playoffs) needed to do something to change the equation, to overcome the status quo. Dabbling at the edges might work but probably not. Opponents, especially within the division, get a lot of time to analyse and adapt to what a competitor is doing. But a truly needy (for success after a loooong hiatus) organization probably requires a bold initiative, probably innovative. The need for innovation wasn't mandatory, but in a tight division such as the American League West, with no apparent punching bags nor apparent Balrogs, bold could work well enough to gain a flag or blow away the competition. And if you can do that for a single season, it pays business dividends for years (not to mention the side-effect of career dividends).

It's risky though, because if it fails, you might look like a dope, and in Baseball, the most transparent endeavour in North America, everything that could make you look like a dope will wriggle to the surface and be beaten to death and well beyond on talk radio and in the blogosphere. And remember your Deming: the bolder the initiative...the farther it is from standard operating procedure...the higher the potential gains are but also the uglier the potential crater if your thesis doesn't work out.

The bold initiative was to try out what I believe is an unprecedented approach to team-on-the-field construction. Recent interest in the statistical analyst world has drifted, probably more from the need for new guys to find and exploit a niche than from actual passion, away from hitting and pitching (both well-stocked with successful analysts), and towards the under-explored & intrinsically fascinating region of defense (notable exception: Chris Dial at Baseball Think Factory who slud over to defense to be able to refine his pitching projections). Defense had been outed as "overvalued" in the book Moneyball, and the cognates from that point of view took root even though the Moneyball A's had moved back to a more moderate position on defense by the time the book came out in paperback. But one of the significant factors that appeared to be a foundation of the M's success was outstanding defense overall and in many places on the field.

Before I detail how what happened to the 2009 Mariners and how the team's front office decided to experiment, I need to lay out three systems concepts managers need to internalise.

There are three numeric concepts managers need to know about in general, but specifically to understand why the attempt to build on the Mariners' 2009 success for 2010 has tanked so horribly. Interestingly enough, many professional statisticians, even in baseball, behave as though they don't understand them when it comes to turning concept into action.

Concept One is The Additive (All-Wet) Assumption.
This is the Management By Wishful Thinking seduction of assuming that if you double some input quality or quantity, you are likely to get double the quality or quantity of the output. It's seductive because we are all taught and virtually all master the sterling example thereof, 2+2 = 4. Its allure is reinforced by the many obvious examples where it holds just fine. Team home run output plays out remarkably closely to the home run abilities of the individual players on the roster. If you replace a batter who consistently hits about five homers a year with another who consistently hits about 20 homers all other factors being stable, it's a reasonable thumbnail presumption to think the team has the potential for 15 more homers for the season. Total Team Homers are the additive result of the homers hit by each individual batter playing for the team (well, duh...only Walt Davis would take me to task for suggesting such a thing, but I'm going to contrast that common sense assumption in the next paragraph). Other tallied events that are almost completely based on individual accomplishment fit this Additive Assumption quite decently.

But many events don't fit The Additive Assumption well at all, especially out towards the extremes. One is team defense. This is one of hardest concepts for people who haven't dug into defensive stats to understand. Team defense adequacy of turning batted balls into outs is not the additive result of each fielder's ability to turn batted balls into outs.

Here's an example. Say I have the best, rangiest SS in history ... his incremental value in preventing hits is greatly affected by who is playing 2nd base and who's on 3rd base. When you see the best SS playing with 50th percentile third-basemen and second-basemen, his range enables him to take some grounders or pop-ups "away" from adjacent infielders who could had them -- if you have ever seen the SS cut in front of the 2B on a grounder up the middle and take a ball either could have had, you already know this. When one or both of the adjacent basemen truly stinks, "all" of the SS' potential for range is more fully actualized. He's going to go to the limits of his lateral range and get balls the stiff next to him wouldn't get to if SS Deity didn't. When the neighboring basemen are 50th percentile defenders, SS Deity is not going to prevent as many hits because those neighbors are going to get to some chances that SS Deity could have gotten to, so that SS isn't free to exercise all his range, ergo his full potential isn't added into the sum of team defense. When the neighboring basemen are Deity-like as well, SS Deity's ability to add incremental defensive hit-removal value is further diminished, because he and his neighbor-Deities are to a greater degree are getting to balls a neighbor would have had anyway. There are many outfield examples of this interplay, too.

At some point on the scale, certainly at the 1974 Baltimore Orioles' point, the very presence of three neighboring glove deities changes what individual batters and team managers try to do against them. In that sense, there may be non-measurable run-prevention advantages, but it's not directly the addition of additional effective fielding range.

The All-Wet instances of the Additive Assumption are traps for managers, both in Baseball and Beyond Baseball. The Mythical Man-Month is one of the best-known non-baseball examples...the all-wet assumption that if two programmers can develop solid working code in six months, then why not hire six programmers because you'd get it done in two months? It's a classic failure of the Additive Assumption, and one that's played out absolutely as regularly now as it was 35 years ago when the All-Wetness of the Assumption was outed publicly through the non-Baseball managerial corps by a rabid Mets fan.

Concept Two is a corollary to Concept One...Critical Mass
What I call Critical Mass is when you collect so much of a single attribute, the result takes off in a way that's very far from the amount you add. An example is the 1978 Milwaukee Brewers where the team's offensive potency was higher than sum of the individual players' offensive potential. The team's line-up was a Lumber Company, and the good and even middling Major League batters on the team tended to outperform what could have been expected for them. I have a suspicion as to why this happened which may or may not be correct, but happen it did. The team had such a Critical Mass of hitting that it disrupted what could have been assumed by the Additive Assumption.

Concept Three is another corollary...Event Horizon
The Event Horizon is quite like Critical Mass, but on the low-end of the scale. Like an auto engine you're tuning so it idles at lower and lower rpm, at some point, there won't be enough fuel to keep the engine turning over...cutting fuel rate by 25% won't cut the idle rpm by 25%...the engine will just stall and not run.

The Turn of the Century Oakland A's hit the Event Horizon with their defense...sometimes they had so many DH/1b/3b types on the field at once, their defensive engine stalled out enough that it undermined the superior-ness of the offense they had assembled. The 2010 Mariners, I believe, hit the Event Horizon with their offense, but I'll cover that in a subsequent post.

These three concepts are critical to understanding both why the Mariners chose their experiment and why it apparently failed. In the next entry, I'll show you how these applied, why the Ms tried a bold experiment, why it was this particular experiment, why it went so wrong and what managers Beyond Baseball can take away from this to prevent their own Cherry-Pie Times.

This page is powered by Blogger. Isn't yours?

free website counter