The “mission” of College Football by the Numbers is to find and promote new ways of evaluating college football teams, games, players, and strategies. Like every other football fan, expert, player or coach, we collect and analyze data, but with two important differences. First, we tend to use data that are collected consistently and comprehensively. In sports, these are known as statistics. I watch LSU and say, “dang, that defense is fast.” That’s data. Then I measure acceleration, top speed and range for every player in college football and find that the LSU defense has a relatively high average in all three. That’s statistics. You can disagree about the former – you might not think the LSU defense looks fast – but the latter is a fact.
The advantage of the standard approach is that it can quickly collect data on thousands of variables. It's pretty easy to watch a guy and decide if he's fast. My data set cannot tell me if a player was fatigued or limping, if the pass interference was really pass interference, how often the left tackle was really holding, if the coach gave a rousing halftime speech, if the players were focused, or if the stadium was loud. Anything we can observe can be measured consistently and comprehensively in theory, but some are more practical (e.g., crowd noise) than others (e.g., the rousing-ness of the speech).
The standard approach doesn’t collect data consistently and comprehensively. We don’t remember every play of a game (my computer does), let alone of the season (my computer does), only the highlights and a few relevant numbers - thus the need for a Heisman moment. Even worse, we tend to notice and remember the data that fit our preconceived notions. For example, few fans have a good grasp of how good their team really is because, if optimistic, they remember the good and forget or justify away the bad, and if pessimistic, they remember the bad and ignore the good.
The second major difference between our approach and the traditional approach is that we are explicit about our methods. Once you’ve collected data you have to decide what to do with those thousands of mis- and selectively remembered bits of data. Does it matter if the team plays up-tempo? When? How much? The kind of statistical analysis that we promote has the same challenge, but because we make our analysis explicit (e.g., we put a point value on home field advantage) we can test and compare approaches, modify our methods and improve our results. That’s the scientific method. A quick glance at human history is enough to see it works.
Some people think they make decisions based on their gut or instincts. First, there is no pre-historic evolutionary advantage to successful sports betting, so I can tell you confidently that instincts are not involved. And Les Miles is not gastrointestinally gifted. Instead, what we often refer to as a gut feeling is really a manifestation of our subconscious. Our conscious minds are terrible calculators but our subconscious minds are much more effective. Experts drawing on years of experience have access to an amazing analytic tool, and they are largely unaware of it. What they spit out to justify their position is the best post-hoc mumbo jumbo their conscious minds can concoct. But even these subconscious calculators are limited and imperfect (see Les Miles), and we cannot build on knowledge stored in one person’s subconscious. It’s not progressive knowledge. The goal of CFBTN, then, is not to replace traditional methods for analyzing college football, but, as far as possible, find expressions for that knowledge on paper and develop new tools to extend it . . . and try to be entertaining along the way.