The stat geeks are taking over Philadelphia. Isn’t it great?
It’s time to upgrade our thinking – and our platitudes – for the 21st century. Screw inches. Our beloved pastimes are really all games of probability.
And Philadelphia’s teams are finally catching up. It’s about damned time.
The two most recent hires in the big four of Philly sports, Eagles head coach Chip Kelly and now 76ers GM Sam Hinkie, are not traditionalists. The foundations for their respective philosophies, Chip’s playcalling and Hinkie’s talent acquisition, are built on foreign, impenetrable, numerical concepts such as “expected values.”
Many people have a problem with probabilities – and by extension the study thereof, that mathematical voodoo known as analytics. The term ‘analytics’ often conjures a couple of images, and neither are particularly flattering. One is that of the pasty, bespectacled weakling with a pocket protector and a TI-89, jealously crunching numbers as he fantasizes about athletic prowess he’ll never have, cheerleaders he’ll never date, and popularity he’ll never achieve. The other, more sinister depiction is the business executive in a 15th floor office furnished with exotic hardwoods and floor-to-ceiling windows, dispassionately assigning numbers to real people from high on his throne and then buying, selling, or trading them as he might deal in rice futures or credit default swaps. But stereotyping aside, I suspect most fans simply misunderstand the concept.
People intuitively deal with and base their choices on probabilities every day, frequently without knowing it. Such things govern how all of us operate. “What are the chances that the Boulevard will be faster than I-95 this morning?” Traffic reports aside, a commuter asking himself this question is going to use some form of anecdotal historical data (i.e. his memory) to determine a measure of probability that one road will be faster than the other. Of course, memories are fallible and incomplete. Applying numerical data doesn’t change the problem, nor does it ‘guarantee’ our commuter will choose the faster route on a given day. But it can provide a more accurate probability, and therefore a better chance of choosing correctly.
Every time Andy Reid said something akin to ‘I’m trying to put my players in a position to succeed’ he was really saying ‘I’m trying to increase the probability that my players will succeed.’ (Whether or not Reid ever actually did that is, for the moment, irrelevant.) Those two statements are essentially the same thing. Whether he admits it or not, whether or not he looks at the statistics, every coach makes decisions that he believes give his team the best chance to win. That’s all any coach ever does or can ever hope to do, really. They don’t play the game, after all, so much of what happens on the field is out of their control. The fact is – and this is indisputable – that probabilities factored into every sporting event that has ever been played or watched, ever. There is a percentage chance that a given play will gain or lose yards before the snap. A percentage chance that any given shot will go in before that shot is taken. We come to identify these percentages using the history of the players and/or the history of all similar plays. It may not be as easy to identify these probabilities as it is of, say, a coin flip. But the probabilities still exists.
Perhaps it scares people to think that they we are capable of creating mathematical expressions to represent the actions of people, but that doesn’t change the fact that we can. Some proof? In a packed stadium, “The Wave” usually travels at 12 meters/second. If huge crowds of complete strangers spontaneously moving in unison will tend to produce a consistent and measurable result, why is it so odd to think that an individual on a field or court could similarly behave in a consistent, measurable fashion?
Random events are inherent to sports, as they are the human condition. There’s no avoiding that. Balls bounce and winds blow and nobody can account for that. But what one can and should and must avoid is reducing his team’s chance of success through a lack of probabilistic efficiency. But we know that our memories are fallible and incomplete and prone to bias, and that is where math and statistics, probabilities and analytics play their part.
So how might such concepts be applied? Here are a couple simple but informative examples.
The Long Two
Under the tutelage of Doug Collins, who once suggested he’d rather blow his brains out than study an analytic assessments of his team, one of the 76ers’ favorite shots was the long two pointer (15+ feet from the rim). Analytics suggest that the long two point shot is the least efficient shot in basketball because it carries a decreased probability of going in without the increased reward that comes from shooting a few feet further back (Kirk Goldsberry).
In 2012-13, the Sixers attempted a league-high 1606 shots from 15-19 foot range and made just 38.8% of them, 21st in the league. Meanwhile, the team shot 36.3% from 3-point range on 1426 attempts. It doesn’t take a math wiz to see that a 2.5% bump in success rate doesn’t justify a 50% reduction in potential points scored.
I’m not saying that the team should force covered threes. I’m not saying that the long two should never be taken. These are ludicrous assumptions, but they’re the kind of assumption many people make when they hear discussions of efficiency. If you never ever shoot the long-two, it will be much easier for NBA-level defenses to key in on the shots you do take, which would then be reduced to only threes, short jumpers, and shots in the paint. Sometimes the long two is really the best shot available on the court in a given possession. Nobody disputes this. But knowing that your team only shoots the long two with a slightly higher success rate than it does from three-point range, you’re reducing the probability that your team will succeed in the long run if you don’t attempt to at least limit your long two attempts in favor of moving the ball and creating other good shots.
This isn’t rocket science. It’s not voodoo, either. Analytics are not going to fundamentally change how players actually play basketball. Shot creation is done on the court after all, not paper. Using logic, anybody could surmise that the long two ought to be a limited shot, but the numbers provide hard evidence that our inherently unreliable eyes simply cannot match on their own.
The Red Zone Fourth Down Conversion
This one’s a little bit more complicated, but it’s the same idea.
What’s your play call when it’s 4th and 1 on your opponent’s 20 with the score tied in the middle of the 1st quarter? Mr. Football Conservative says kick is the only option. Why? It’s simple. You take ‘sure’ points rather than risk giving it to the opponent. Of course, this philosophy isn’t really about risk. It’s about how you look. It’s about vanity and fear. You don’t convert that yard in the 1st quarter, the other team uses the turn of possession effectively and scores a touchdown, and you’re likely to get ripped apart by the fans and media. You’re going to look foolish.
In a play-calling philosophy that’s truly based on risk, you go for that 1st-quarter 4th down conversion nearly every time. As Tim Livingston of ThePostGame.com insightfully wrote, “For the Texas Hold ‘Em players out there, kicking a field goal in that situation is like folding pocket aces pre-flop against a smaller pocket pair. You’re conceding when you’re the overwhelming favorite. Even if the guy hits his set and wins the hand, you don’t have any regrets. You know you made the right call and that you’ll win the hand a majority of the time.”
Since 2000, field goal attempts from the 20-yard line were only made 85.1% of the time. Over that same period, redzone 4th-and-1s were converted for a 1st or a touchdown 62.7% of the time. Granted, it doesn’t sound great at first when you consider that historically you’ll give the ball back 37.8% of the time, but consider this: You’ll be giving that ball back at the opponent’s 20. When a team starts a drive at their own 20, they’ll score points about 25% of the time, they’ll punt about 55% of the time, and turn it over about 20% of the time (Brian Burke, 2009). So of those times you don’t convert, the ball will still come back to your team, on average, 3/4 of the time. Therefore, history says you’d pay for this hypothetical 4th-and-1 attempt with points at your own end less than 10% of the time.
Considering that since 2002 the margin of victory is 4-points-or-fewer in more than a quarter of NFL games, the willingness to take that 4th down gamble would be a truly considerable advantage over the course of a 16 game season. When one understands all of this, kicking a field goal suddenly looks like a far less efficient usage of that redzone possession than attempting a conversion.
Chip Kelly’s Eagles will undoubtedly blow at least one 4th-down conversion this year and pay for it with points at the other end, but I suspect Chip will sleep well anyway, because the probabilities were overwhelmingly in his favor when he made that call. Granted, Chip himself has tried to debunk the theory that he goes for it on 4th all the time, but even with an accurate kicker and strong punter, it’s a safe wager that he’ll try to convert it more often than the vast majority of his opponents.
The above example is admittedly simplistic, but it helps to illustrate the potential benefits of analytical assessments. This type of mentality could be a serious competitive advantage in the long run.
While people in this town tend to hate when their sports change, nothing about management and coaching tactics based on analytics is mutually exclusive with the “blue-collar, lunch pail” mentality that defines Philly sports. Just because you have numbers describing a player’s performance doesn’t change how you get that performance from the player. They still need to show up, practice hard, and play harder in order to succeed. Coaches and managers still need to fundamentally understand all the physical aspects of their respective sports in order to craft winning teams.
Analytics don’t take away those parts of the game. All they do is add to our understanding. They help to quantify areas of the game that were, until recently, considered “intangible.” And the metrics continue to get better, year by year. They help avoid simple coaching and management mistakes. They expose inefficiencies both on the field and in the player-acquisition market.
A manager or a coach who willfully ignores helpful information due to a stubborn adherence to a traditional, anti-statistical understanding of the game is negligent. Those who think like Doug Collins, Rubén Amaro, and Paul Holmgren, those who are willfully ignorant of analytics, are dinosaurs. The longer they stick around, the more their respective leagues will continue to innovate and pass them by.
One has to admit that the traditional approach has done very poorly in recent years. All four of our teams are middling-to-awful at the moment. Worst case scenario, the Eagles and 76ers go from terrible to… still terrible. If nothing else, Chip Kelly and Sam Hinkie represent a novelty that this town hasn’t seen in quite a while. I fail to see how that’s a bad thing.
For those who have further objections, I’d like to direct you towards Spike Eskin’s response to Larry Brown’s interview with WIP on Monday. He does a pretty good job rebutting the common arguments that tend to pop up when traditionalists like Brown discuss sports analytics.
Follow us on Twitter @TheWooderCooler
Posted on May 14, 2013, in 76ers, Eagles, Hank Mushinski, Posts, Sports Philosophy, Stats Posts and tagged 76ers, analytics, Chip Kelly, Eagles, NBA, nfl, Sam Hinkie, sports. Bookmark the permalink. 4 Comments.