Admittedly, alpha is a bit of an obsession for us at AllAboutAlpha.com. We view alpha as the “magical”, un-replicatable, skill-based, value-add produced by human creativity. In a world where nearly everything can be explained by systems, just about the only thing we can’t synthetically and cheaply replicate is human ingenuity. In other words, what we can’t explain using known variables we chalk up to “alpha”. This applies to financial markets, business, education, the arts, and even – as this article points out – to American football.
Analytic Investors (see previous posting) calculates an “alpha” for NFL teams at the end of each season of American football. They define this alpha as “the extent to which each team exceeded or underperformed its expectations”. Essentially, they attempt to identify how well a team performed given the raw materials available to it. While a comprehensive list of these raw materials isn’t listed anywhere, the “market” (i.e. bettors) collectively defines them via predicted margins of victory. We might refer to this alpha – this unexplainable quality – as “heart” or perhaps “skill” displayed by the head coach in making the most with what he was given.
In their words:
“A team’s alpha is determined by measuring how much the particular team, on average, exceeded their predicted probability of victory. For example, if the market expects a team to win eight out of sixteen regular season games and they end up winning nine games (one more than expected), then the teams alpha is 6.25% (or 1/16th).”
While the results of this study show that successful teams do tend to produce positive alpha, this is not always the case. In certain circumstances, high expectations were not actually met and a team with a winning record ended up producing a negative alpha. For example, the Denver Broncos were 9 and 7 on the season, but produced a -1.3% alpha. In fact, alphas for the group of 8-and-8 teams ranged all the way from 13.9% to -6.4%.
So who’s to blame for low alpha? Well, like hedge fund investors, NFL team owners often fired their managers for producing low two-year alphas. Three of the six teams producing the lowest two-year alphas ended up firing their managers. Perhaps even more telling is the fact that managers had a higher propensity to resign from low-alpha teams, regardless of their actual win-loss records. Of the three resignations since 2005, all were from negative-alpha teams, even though the combined win-loss of these teams was nearly even (23-25).
A Trading Algorithm
So how does Analytic help its clients profit from its insight? It proposes an “adjusted point spread” for each game:
“The key to deriving the proper point spread adjustment is the difference in alphas between the two teams. The team with the higher alpha will always have the spread adjusted in their favor. This happens because if a team has a higher alpha, the market has systematically undervalued them relative to their opponent. Thus, to account for this mispricing, the spread must be moved in their favor. If the match-up contains two teams with the same alphas, there will be no adjustment to the spread.”
The NFL provides a useful analogy for exploring alpha-generation in asset management and can be taken in many directions. For example, one topic we’d like to see addressed next year is the extent to which coaches’ salaries are linked to alpha-generation.
Read Full Article (free – email/name only – registration required)