Monday, January 18, 2010

The Experiment: Rethinking NFL Coaching

Consider a thought experiment for a moment. Say we've walked into the world of Star Trek: The Next Generation, and are watching the android Data. As part of his studies of the human psyche, he's examining the world of sports and has taken a particular interest in American football. Specifically, in an attempt to improve his leadership abilities, he's trying to figure out what makes a successful coach. To this end, he's constructed a holodeck program to simulate a NFL game. In order to isolate the variable of coaching, he's programmed both teams with identical players.

The difference between the teams? Team A is run by a typical NFL coaching staff, with a head coach (Coach A), position coaches and offensive, defensive and special-teams coordinators. Team B is run by a 30-year old who doesn't have any actual football experience, but has spent hours on end playing every installment of the Madden franchise since its inception and watching every possible NFL game for decades. To make it possible for him to control the entire team, it's programmed in that he sends in his instructions through Madden's "Coach" mode (playcalls, substitutions and timeouts, but no actual control of players on the field), while Team A handles their gameplan in the traditional model. Now, when these two teams face each other in Data's simulation, which would you bet on?

Every bit of common sense would say to go with Team A. They're run by guys who have been there and done that for years. In fact, because they're involved with an NFL team, these guys have been selected as some of the best coaching talent out there. They work crazy hours and develop elaborate game plans. In short, they're the professionals. Of course most people would take this highly skilled team over a single rank amateur. In fact, most people wouldn't bother running the experiment at all; they'd tell Data it's a foregone conclusion.

One nice thing about computers (and androids), though, is that they evaluate situations based on information and programming instructions, not simple a priori assumptions about common sense. Data would run his program, and I'd bet it would come to a surprising conclusion; namely, Team B winning.

Why would I bet on Team B? I'm not discounting the value of experience; in fact, I recently read Malcolm Gladwell's book Outliers (which I've written about before), and I largely buy his argument that much of the success we typically credit to indviduals' innate qualities is, in fact, largely due to their experience and circumstances. That's actually what I'm building on here. The 30-year-old running Team B has his own experience, and it happens to be in a very different arena than that of the traditional coaches running Team A. My guess is that it would prove superior.

Perhaps the easiest area to explore this idea in depth is the area of clock management. Joe Posnanski wrote a very interesting piece on the subject today, concluding that "Coaches in the NFL have no idea how to use the clock." He makes a compelling argument that fans understand clock management better than coaches, and I think he comes up with part of the reason why; fans spend more time thinking about the clock than coaches. However, that's only part of the explanation, I think. The key factor is experience.

No one coached more NFL games than George Halas, the legendary coach of the Chicago Bears. He finished with 318 wins, 148 losses and 31 ties in 497 regular-season games. Second all-time in games coached is Don Shula of the Baltimore Colts and Miami Dolphins; he finished with 328 wins, 156 losses and six ties in 490 regular-season games. Using Pro Football Reference to rank coaches by NFL games, no active coach comes even close; the highest-ranked is Tennessee's Jeff Fisher, who's 17th with 136 wins and 110 losses in 246 regular-season games.

Now, think about that for a moment. The only area where you really get to work on clock management is in games. Sure, pre-season games help, and you can do a bit during practices, but there are always so many more pressing things teams are doing in practice, and it's never quite the same environment. Moreover, it's only a fraction of those games (the close ones) where you really have to worry about clock management; it's probably the last thing on a coach's mind when he's either winning or losing by a lot. Even in those games where clock management becomes an issue, an NFL coach is worried about lots of other things as well; keeping his players happy, working with his coordinators, thinking about playcalling and motivating his team. Considering all that, I'd guess that NFL head coaches really haven't had that much experience managing the clock.

What about the coach of Team B, though? It actually wouldn't be that hard for him to surpass Team A in clock management experience. For one thing, Madden offers the option to play shorter quarters, which allows you to get many more games in. The shorter quarters don't affect clock management much, because you still go through the same situations at the end of games. Even if you play with full-length quarters, though, Madden is much shorter than a regular NFL game thanks to a lack of commercial breaks and an ability to call plays quickly. Moreover, in Madden, you can play any number of games in a row in a sitting; in real life, you only get 16 per season (plus any playoff games), and you have to wait a week between each game. Thus, serious Madden gamers like our Team B coach spend much more time actually managing their teams in game situations than real NFL coaches.

Don't quite buy it? Well, consider this. If a Madden gamer played 16 full seasons, he'd have coached 256 regular-season games, ten more games than Fisher. In 31 seasons, he'd pass Shula with 496 games. In 32 seasons, he'd pass Halas with 512 games. 32 seasons is hardly an unreasonable figure; I've played through around 10 in the last six months, and I'm anything but a hardcore Madden user (which our Coach B would be). A casual Madden gamer like myself could hit 32 seasons in three to four years of gaming; a hardcore one could do it much faster. Moreover, Madden has far less distractions than an actual game (no motivational speeches required, no assistants to keep happy, no fans going nuts, no way to argue with refs), so Madden coaches probably spend much more time thinking about clock management than real coaches. Thus, Coach B would have far more experience actually managing the clock than Coach A.

Of course, there's much more to successfully coaching a game than clock management. However, many of the other components are similar. Consider the issue of when to use challenges, another area fans and writers often take NFL coaches to task for. Using the same argument seen above, most Madden users probably have more experience with challenges than head coaches. The same applies for offensive and defensive play selection and audibles. These can be worked on more in practice, so NFL coaches have more experience with them than they with challenges, but they're still very different in a game than they are in practice. Even if we counted all the hours Coach A spent running plays on the practice field (not as many as you might think, as practices also involve a lot of individual drills and small-group situations), though, our hardcore Madden user Coach B would likely still have more experience with playcalling than Coach A. He would have spent more time in game situations, and that gives him a better chance for success.

Now, you can make a case for Coach A if you insist that clock management, challenge management and playcalling in Madden can't be applied to actual football. If they're completely separate things, then Coach A has a distinct experiential advantage. I don't think they are, though. Again, the strongest case can be made in the area of clock management; the clock operates under the same rules in Madden and "real" football, and you can take the same measures to either stop it or keep it running. Challenges and playcalling are a little tougher, but keep in mind that Madden is programmed to operate under the same rules as "real" football, bases its plays on real plays and bases its players on real players. The quality of the simulation has improved over the years, and you could argue that the latest versions give reasonably realistic results in terms of playcalling; frequently, the same offensive plays work against certain defensive sets in both reality and the game.

There's something else that would help Coach B: innovation. As I wrote in my piece on the subject at The Good Point, innovation has always faced an uphill battle in the NFL. Few coaches are willing to really innovate for fear of looking silly; however, once an innovation is discovered, it quickly goes from laughingstock to widely-used strategy (see the Wildcat offence). Those innovations generally aren't that far out there, though; for example, the Wildcat was largely based off the old single-wing offence and was run successfully in college before it was brought to the pros.

The reason we never see anything really bizarre in the NFL is thanks to the backgrounds of the people involved. Most coaches in the NFL have gone through a very similar career arc, starting off as players, then becoming low-level assistants, then coordinators and then head coaches. They may have slightly different schools of thought based on their personalities and the coaches they've worked with (hence the concept of coaching trees), but all those ideas are really just branches of the same tree. You see this in many businesses; people from the same backgrounds tend to approach problems the same way. If you come up with a multi-dimensional business problem involving manufacturing a new device and present it to different groups, accountants will tend to look at it in terms of the costs involved, while research scientists will look at it from the perspective of what can be created and engineers will look at it from a perspective of what's feasible to create. None of these approaches is necessarily right or wrong, and you have to incorporate elements of all of them to find success.

The problem is that the NFL's small group of coaches from similar backgrounds means everyone approaches certain issues in similar ways. (Chris Brown had a good piece on this at Smart Football). There's also the risk of looking silly if you try something radical, which motivates many to stick to the tried-and-tested paths even if they do have other ideas. For example, consider the case of fourth downs. Bill Belichick got plenty of criticism for daring to go for it on fourth-and-two in his own territory this year, even though many of the numbers we have on fourth-down conversion percentages support him. That was a highly unconventional call by NFL standards, but it's still well within the bounds of what is sometimes done; that was late in the game with a short distance to go. What you would never see in today's NFL is a coach who consistently goes for it when faced with fourth down and five yards or less, regardless of field position and time left; that's far enough outside the box that no one would dare to try it for fear of looking silly.

The nice thing about Madden, though, is its judgement is based on results, not aesthetics. I've consistently gone for it on fourth and five or less and found lots of success doing so. You can do plenty of other things that you wouldn't likely see in the NFL too, such as running five-receiver sets on every play or prominently featuring flea-flickers and end-arounds. Madden users have a variety of different backgrounds, and ones very different from the typical NFL coach, plus the game design encourages experimentation and innovation. It also leads to a more statistically-based school of thought; if certain ideas work more often than not, a player will keep using them. Naysayers would argue that these ideas wouldn't work in the NFL, but it's impossible to know because no one has tried. The nature of Madden promotes innovative thinking, though, and that would be yet another advantage for Coach B.

I'm sure this idea still seems reasonably silly to many of you. After all, Madden's just a game, and games by nature are designed for entertainment. However, the increasing realism of the game in recent years has brought it closer to the level of a simulator. There are plenty of other fields where simulators are considered to provide valuable training experience, particularly for pilots and astronauts. Do you think it's really tougher to deliver a useful simulation of football than a simulation of landing an F-14 on a carrier?

Now, I'm not saying we could replace NFL head coaches with Madden gamers and instantly see improvements across the board. There are many areas of football where a current NFL coaching staff like Team A's would have an advantage, particularly in developing players from raw draft picks to seasoned veterans, motivating their players to give it all they have and juggling their players' egos. Over a long period of time, such as a complete season, I think those advantages could nullify Team B's greater expertise in playcalling and clock management. However, for the one game considered in Data's experiment, I'm taking Team B.

The broader question, though, is why can't we have the best of both worlds? What if we kept the current backgrounds of our NFL head coaches, but got them to frequently play Madden and encouraged them to try new strategies? I'm quite sure we'd see them get better at clock management, and I think they would improve as playcallers and innovators as well. At the end of the day, this will probably just remain a hypothetical, but it's an interesting one to consider. Is the way we currently develop coaches the best way to do things, or should we encourage them to take a look at the virtual world? After all, football is just a game.

No comments:

Post a Comment