Tuesday, February 17, 2004

PART IV: Paul DePodesta: If
You're "The House", You Can Be Fearless  

In previous entries on the Paul DePodesta presentation to non-baseball managers on change management (off-line as I write this), I've covered a few of his techniques & insights. This post continues that series, and is not the last.

Let me note that yesterday, the Los Angeles Dodgers, the first organization to hire an official team statistician (Allan Roth), chose this highly numerate young man to be their new G.M.

If you haven't read the previous & don't want to, here's some table setting:

Paul DePodesta, the Assistant GM of the Oakland Athletics featured (not extensively enough) in Michael Lewis' Moneyball, is one of the most interesting of the new statheads in baseball front offices. What separates him from the others is not that his stats are better or deeper, it's that he understands the key, final, most-difficult-to-master concept in the Management By Baseball model: Change.

He understands that to successfully manage and push change, you have to change he changes you deploy, even as you're concurrently fighting to install them in the first place.

To understand the system he was working within better, he kept asking naive questions. And in return, he got a lot of opinions.

The response to all this questioning was somewhat expected. There were opinions layered on top of opinions. A lot of people were saying, "I think it's because," or "maybe it's this," or even "that's the way we've always done it". My industry is comprised on human capital -- the players are our assets. So subjectivity plays some role. But the enormity of the subjectivity was staggering. [snip]

Opinions are great -- don't get me wrong. They're great for starting research projects. Then you go study and see if you can prove the opinion or not. But when placing multi-million dollar bets on future outcomes, opinions are wholly unsatisfactory. Opinions as conversation starters are fine. Opinions as conclusions are very bad.

I started research projects to discern the objective "why". I wanted to know why certain teams won and why otyher teams lost; why certain drafts produced big stars and others didn't.

This was the naive question put to work.


He knew he wanted to instill analysis into the mix. He knew he'd need measures, But like many managers outside of baseball who are trying to develop and direct changes sensibly, he was faced with "information overkill"...too many unintegrated numbers and no established systematic way to balance and blend them. DePodesta threw himself at an old sabermetric drill, using Markov technqiues to find probabalistic relationships between baseball events and run scoring.

I was able to figure out that a man on first with nobody out is worth "X" runs and a man on second with two outs is worth "Y" runs. From there I was able to jump to understanding what it means to have someone who can hit a lot of doubles. What was the value of that event and others? I went a step further and asked who the people were who could add these value-enhancing skills to our team.

The result is a model that can assign some marginal value to each thing that can happen on the field. Then, you can apply that model to calculating a rough value for each player, as well as shape in-game strategies and plan training effort (to maximize training in the components that have the highest rewards).

And to go along with the probabalistic model, he imported an analogy to buffer the fear most people have of change.


The analogy was a casino analogy: "Be the house".

Any individual bettor might win any individual bet against the casino (the house). But the odds are in the house's favor. Chain together enough events, stack in enough betters making enough bets, and probabalistically it's overwhelmingly likely the house will come out ahead.

Because baseball has such a long season (so many events), if you can just identify and grab enough small edges, you can "be the house" and win, as DePodesta aimed for, 60% of your games. It means failure (40% of the games are losses) is permissable and therefore no cause to panic, because you're going to (at least you set yourself up to think this way) win in the end.

Every season we play 162 games. Individual players amass over 600 plate appearances. Starting pitchers face over 1,000 hitters. We have plenty of sample size. I encouraged everyone to think of the house advantage in everything we did. We may not always be right, but we'd be right a lot more often than we'd be wrong.


In many organizations (clearly not all), there are collections of things that were once processes, alive, evolving, that have ossified into habits, supported by opinions. As in baseball, they are supported not by rigorous examination or testing (that may have been done when they were first established), but by habit, involuntary gesture. In anthropology, we call these "survivals" like the behavior (almost involuntary on the part of the speaker) of saying "God bless you" in response to someone sneezing. More often than not, the speaker doesn't truly believe that at the moment of sneezing, the sneezer is subject to being possessed by an evil spirit, the reason people in the 12th century were trained to say "Gode blesse ye" to prevent such nefarious doings. (Aside: Perhaps I'm being optimistic when I say "more often than not". I go to a wide-demographic site called Iwon, and they have polls for their visitors. One was about evolution versus creationism. The results indicated about 20% believed in evolutionary biology, about 20% believed God created all life the way it was in 4000 B.C., and about 40% believed the totally impossible Baked Alaska that really it was both at the same time. So perhaps a plurality of American believe, along with the Bundi of Borneo -- yes, the bones in the nose dudes -- that sneezing makes one subject to possession by evil spirits).

So putting a yardstick to presumptions, finding out where the support for behaviors is a habit and not the result of ongoing examination, is a fruitful way of turning up opportunities to be aggressive.

And "being the house" can be productive, too, in many settings. Not all shops are like this, for example, the auto mechanic place up in Everett that mostly does tune-ups for Porsches and Audis. Close-enough is no-good, only perfection is acceptable because the equipment wasn't made to be even slightly tolerant. And, because there are not a lot of transactions over which to accumulate an edge.

But more likely than not, you have some processes in your own organization that are repeated frequently and where perfect is not perceptively better in any way to "close enough". These are great targets for re-tooling, because if you can figure out a way to "be the house", that is, install a strategy that's a winner over time, you can reap constant rewards. And you can buffer staff fears about occasional "losses" which always accompany change but must not be allowed to halt it.

BUT, please note, you have to keep measuring results. Unlike a casino, the environment shifts subtly even when your assumptions were totally correct. But sometimes, your assumptions are going to be flawed, because data can tell you only so much.

As baseball researcher Don Malcolm will be the first to remind you, the model DePodesta created for the Oakland A's (more on that in the next entry in this series), worked well, but not well enough. The model paid tribute to in Moneyball was heavily re-tooled. Yes, the As still have more than their share of unathletic bodies, but their success in the last couple of seasons is more attributable to three "ace"-quality starting pitchers and defense that was improved over the original model that rejected defense as an overpriced hood ornament. Through observation, DePodesta learned that attributes like defense and team OBA were not linear functions charted against success, but non-linear one (for example, you can yield on defense only so far before the model decays and goes China Syndrome).

In my next entry in this series, I'll continue this exploration of DePodesta's insights & techniques, specifically related to how he worked to make the change happen, the (very successful, and contrary to the way I usually try to do it) organizational development approach he took.

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

free website counter