The Numbers: A Rebuttal (7/18/11)
Monday, July 18 at 4:41 PM ET
Early into the site's existence, Bill Simmons' brainchild Grantland.com ran a piece by generally well-respected, frequent Wired contributor Jonah Lehrer (who, despite looking like he is 13, is actually a year older than I am), which essentially denounces the "sabermetric" movement of the last decade or so, while hyping "intangibles." The article is written well enough and has become extremely successful due to the buzz it has generated and the eyeballs that have viewed it on Grantland One would presume that the founder of a website that analyzes future outcomes using probabilistic simulations would have a strong opinion on such a topic. One would be right. While I don't agree with Lehrer's overall premise, my take is probably a little different than most would expect. The notion that interest in numbers and interest in the game outside of the statistics are mutually exclusive is inherent in every discussion like this. It's also absurd.
First of all, let's address some points of clarification that I alluded to with the parentheses above. Supposedly Bill James coined the term "sabermetrics" in reference to (straight from the term's wikipedia page) "the analysis of baseball through objective, empirical evidence, especially baseball statistics that measure in-game activity rather than industry activity such as attendance. Lehrer seems to apply that definition in the broadest sense, meaning that any reference to numbers with respect to baseball, especially if the same analysis ignores anything that "can't be quantified" is a form of sabermetrics. In that context, sabermetrics has been around as long as baseball itself. But a column such as this, believes it is a new phenomenon without a valid premise.
I fear that Lehrer, who typically writes on psychology and neuroscience, lost himself in the story-telling objective, trying too hard to apply his normal expertise to sports without removing his personal bias (which is something along the lines of "JJ Barea is a short, scrappy guy with little hype or attention who played really hard and with such great heart and clutch that the Mavericks won"). Simmons, individually and now with Grantland.com, is attempting to bring to light the rich stories of our athletes and sports personalities (amidst other, similar pop culture interest) in much the same way that they were presented in the early 20th century before news, highlights, blogs, debates and "The Decision" consumed our sports culture. I get it - and I like it. In fact, often ignored in the discussion of SABR relative to sabermetrics is how important it is to that group to preserve the stories and history of the game - numbers or otherwise. And James, the most prominent name amongst those who have considered baseball statistics, is as much concerned with the anecdotes as the analysis. It's that psychological background into how people tick and the journeys leading to the outcomes we are familiar with that can add to the appreciation for and interest in sports. I love hearing the stories about Dave Cowens driving a taxi in the middle of his career. I do. But, while notable and despite Lehrer's attempt to rationalize our interest in them, those stories are trivial when it comes to wins and losses.
Personally, I would not consider myself a sabermetrician because a) I've yet to find a consistent definition of the category and b) I am not and have never been a member of SABR (this is of no offense to that organization with which I have worked in the past - in an attempt to be more well-rounded, I simply prefer to keep my affiliations and memberships separate from my work). Instead of labeling my approach, I consider myself someone who tries to make the most efficient and rewarding decisions for the broadest group of people I can affect in both the short term and long term... just like everyone else.
Everyone tasked with making decisions attempts to achieve the most desirable outcome - optimal if you will - from that choice. To consider it a psychologically, as quickly as we can, we breakdown all of the possible outcomes from the decisions we make, weigh their likelihoods and determine which choice brings about the best chance of the best solution. Experience, mental capacity to comprehend the issue and knowledge of the issue all go into achieving results. Some are better at that than others. The best have combine a wealth of experience and the help of technology or others to assist with what the brain cannot compute as quickly.
A baseball (sticking with baseball to remain consistent, but this applies to all sports) manager, general manager, umpire or player is doing the same thing. For managers, general managers and players, the goal is - or at least should be - the same: Given my constraints, what can I do now to improve my teams' chances of winning as many games as possible? If the manager, general manager or player has any other objective, he shouldn't be doing what he is doing. Each decision that individual makes that is at all related to the game should be completed with that objective in mind. The problem is that it is impossible for the mind, in the allotted time, to comprehend everything that warrants consideration with each decision.
That's where the numbers, and more importantly, technology, can help. Numbers help us solidify the likelihood of events and technology helps us account for interactions of all possible variables. I am a significantly better handicapper and fantasy sports player now with the aid of simulation than I ever was before because it can quickly account for every possible interaction of every player throughout the entire game or season. I'm also much better at planning events (yes, I applied probability theory to our wedding - you did to, though it was likely in your head, not in excel), routing travel, creating a healthy diet, managing my finances and scheduling my day with the use of numbers and technology. Could do any or all of these things without explicitly utilizing numbers/probabilities and technology? Technically, yes. But it would probably take me the rest of my lifetime to simulate one game 50,000 times by hand (with or without the use of "sabermetric" numbers). With technology, I can speed up that process to the point where it takes less than a second to do all of that and provide me with results that can help me manage risk and make efficient decisions. The most likely event does not always happen, but at least I have a sense of how likely it is.
So while those who are not as analytically inclined prefer to put us "nerds" in a separate category, I feel we are one and the same. There is no mutual exclusivity here where we have some secret nerd agenda that you can't handle. We all want our teams to win and we think about the same things that can affect the team's chances of winning (who should bat where? when should a pitcher come out of the game? is a top minor league prospect worth giving up for a veteran left-handed specialist?). The difference is that those of us who are comfortable with and trust the right (right meaning adequate sample size, removal of bias and an understanding of how past figures can project to future success - this is the part where most slip up and where those who can correctly manipulate the numbers get frustrated with others who try to force irrelevant numbers into the conversation) numbers, especially with the aid of technology, reach optimal decisions quicker and with far greater frequency than others.
Over one hundred years of baseball data suggests that batting Drew Stubbs, a career 32.9% on-base percentage and a 28.6% strikeout-rate, first and Edgar Renteria, clearly in the downswing of his career with a 30.7% current obp following years of 33.2%, 30.7% and 31.7%, second is a terrible strategy. However, over on hundred years of good ole' boy baseball stubbornness says to bat a fast guy first and a good bunter second. So Dusty Baker, the Reds' manager who prefers sticking with convention and not ruffling feathers over giving his team the best chance to win, sticks with the latter. Regardless of the tone of that last sentence, I don't presume that he does this to deliberately hurt the team; he is trying to win games. He does it because he does not know any better and does not have the ability to comprehend all that goes into building the best lineup (no one does without the help of technology and the numbers). But if he knew with certainty that batting Ramon Hernandez first and Joey Votto second instead would give him a 2.5% better chance to win (that's a made up number, but it's probably close to the truth if Votto is followed by Chris Heisey, Jay Bruce, Drew Stubbs, Brandon Phillips, Scott Rolen and then Edgar Renteria), one would assume he would change his lineup to the more efficient one.
Just about every business utilizes what Lehrer has categorized as "sabermetrics." In retail, businesses need to find the optimal re-order points at which new products are needed. From there trucks use the most efficient possible routes to get where they are going. This saves time and money. It used to be done by "gut" and "feel" (i.e. in the minds of the management teams), but it is much more responsible to do it with technology. The same can be said for anyone in the finance industry, in manufacturing, in energy, in medicine. If a logistics manager knows that one of his truck drivers has a medical condition that forces him to take more frequent breaks than others, he does not ignore this; he builds it in to his plan. A portfolio manager is interested in the personal background of a new CEO, which can be added to, but not dominate, the valuation of the new organization. Efficiency is king. Risk management wins. We finally have the technology to achieve optimality (or at least much closer to optimality than ever before). Jump on board because this is the future.
Which brings me to the notion of "intangibles." If the term "sabermetrics" seems confusing, think about "intangibles," which, by definition seems to be impossible. How can we categorize something that is not capable of being categorized? To me, everything is tangible - otherwise, it wouldn't be. I get questions all the time about things like "human spirit," "heart," "clutch" as if we intentionally ignore certain aspects of a player to deliberately undervalue him. It's not like we look at a guy's 40 time and project what kind of quarterback he will be. And even given that, it's not like speed is "tangible." We have a sense of its relative value for a player and, in a good sense, it is quantifiable, but that does not give it tangible qualities that are vastly different from other qualities the player has. Every player on every team has attributes that allow him to contribute to his team's likelihood of winning or losing. Obviously, some of these attributes are more obvious than others, but we still have to account for all of them.
Joe Montana does not have numbers that dominate the historical statistical leaderboards, but his teams won games. There is a (tangible/quantifiable) reason (Montana rarely made mistakes and could get the ball to his playmakers in specific situations where that was necessary). On the contrary, Terrell Owens' teams failed to win as often with him on the field as off; despite his gaudy numbers (Owens' personality demanded the ball at a higher rate than anyone in the league, which hurt his team's ability to diversify its offense to add balance - not to mention the high frequency of dropped passes). Players rarely change over time (and when they do, it's minimally and generally along an expected curve for age, health and experience). The more we know about a player, the more we learn about his attributes and what he contributes to his team's chances of winning.
Because we are on the topic, another notion that I find unfortunately amusing, is the attribution of "inventing" statistics to certain people. People can popularize the use of certain statistics, but that does not mean that they invented those statistics. It's not like someone was sitting around 15 years ago and said, "So if I take Tom Glavine's walks allowed and add them to his hits allowed and then divide by innings I get WHIP. Wow, I just invented whip!" As long as walks, hits and innings pitched have existed, so has WHIP. Numbers represent observations of historical facts that are part of the public domain. Use as you see fit.
Lastly, I can address the specific (I would say arbitrary) examples provided in the column. The Dallas Mavericks did not win the NBA Finals because they ignored the numbers, threw JJ Barea into the mix and just winged it. Dallas is one of the most number and technology savvy teams in the league. They even employ 82games.com founder Roland Beech to sit on the sidelines and advise on matchups based on empirical data. Plus, it was a thorough analysis of coaching candidates that identified Rick Carlisle as being the most likely to make optimal decisions. Do I believe that the Mavericks had a greater chance to win the series than we gave them (33.8%)? No. They hit their 33.8% chance (Congrats!). And they did so by doing exactly what I (the numbers nerd) said they should (see NBA Playoffs Recap publicly on my blog and on the radio on Dallas) - which was to increase possessions/tempo (and what better way to increase possessions efficiently than by playing with two point guards?).
The poorly constructed and reviewed reference to Aaron Rowand's contract in the column is even more egregious. First of all, the contract was signed after the 2007 season, so I'm not sure what 2010 stats have to do with that decision. Secondly, Lehrer presents OPS as if all of the "sabermetricians" got together and determined that OPS (which is actually a big mathematical faux pas) is the end-all-be-all tool for evaluating position players, which is far from the truth (personally, all I am ever concerned about is the strength-of-schedule, ballpark, health, age and experience-adjusted probability that I need to take the next step in the plate appearance decision tree). When he signed the deal, Rowand was a 30 year-old centerfielder, who in-depth statistical analysis evaluated as one of the best, if the not the best, defensive players at a crucial defensive position. His hitting was never spectacular, yet was typically around average for the league, which made him slightly above-average for his position. He had been very healthy up until that point. And, he was going to a pitchers' park after playing in a hitters' park for his entire career. Yes, he got injured last year and that hurt his value, but that could not have easily been foreseen. It should come of no surprise that Rowand's offensive numbers declined. It should also come of no surprise that the Giants won last year's World Series on the strength of their pitching and defense.
Ultimately, everything in sports comes down to the numbers. Wins and losses reign supreme and some form of points yields those achievements. It's been that way for as long as sports have existed.
July MLB Performance:
Even though it's been more than a week since the last blog entry, we have only had about a week's worth of games and, even since the players, managers and I got a chance to take a little break, we are still seeing similar performance to what we highlighted in June and in the beginning of July. All playable picks on the whole are profitable (+$400, RL and ML have been great, while O/U struggled a little before going 7-0 last night), with our top picks doing well. Normal+ picks thus far are 12-5, including going 5-1 on the ML since our last post. We have now hit 15 of our last 16 normal+ ML picks and are 39-16 on the season with such plays. Building on a strong June, "Upset Watch" picks, where we like a +100 or worse underdog to win outright more often than not, are 6-4 so far this month, which can provide great value...
... maybe trusting the numbers isn't such a bad thing.
As usual, if you have any of your own comments about this article or suggestions about how to improve the site, please do not hesitate to contact us at any time. We respond to every support contact as quickly as we can (usually within a few hours) and are very amenable to suggestions. I firmly believe that open communication with our customers and user feedback is the best way for us to grow and provide the types of products that will maximize the experience for all. Thank you in advance for your suggestions, comments and questions.