Monday, May 24, 2010

Quarterback Inconsistency Due to Short Seasons?

In "The Wages of Wins," the authors argue that "quarterbacks are like mutual funds," as past performance cannot determine future productivity.

For instance, the authors decided to "grade" each quarterback, based on that quarterback's productivity compared to the rest of the league in each season, to compare how quarterbacks did from year to year. Each quartile received its own letter grade; i.e., the top 20% of quarterbacks in a given year received A's, the next 20% received B's, and so on.

Often, quarterbacks are rather inconsistent, as evidenced by Kerry Collins:
As a rookie in 1995, Collins ranked in the lowest 20%, receiving an F. In 1996 his performance improved and he received a B. The next two seasons, thought, Collins declined, earning an F in both campaigns. Then in 1999 he earned a C, followed by a B in 2000. After declining again, this time to a C, in 2001, Collins earned an A for his efforts in 2002. Unfortunately his success was short-lived, as his performance was worth a D in 2003. In 2004, the final year we examined, Collins earned a C.
The authors conjecture that quarterback performance varies widely because their success is so dependent on on how the other players block, catch, and play defense.

Each NFL season consists of 16 games. I wondered if batters in baseball, whose batting average is thought to be independent of their teammates, would look similarly inconsistent when examined in 16-game chunks.

Someone with more time and inclination could look at a larger sample, but I looked at Tony Gwynn's 1997 season, in which he won the National League batting title with a .372 average (I used data from the ever-useful Baseball-Reference). That year, Gwynn played 149 games. The below table summarizes Gwynn's performance in the first 144 games, divided into nine 16-game periods.

PeriodAt-batsHitsAveragePercent change

Gwynn's lowest period average is .306, still far above the league average (historically around .260). But if he were judged like NFL quarterbacks are, we'd wonder why Gwynn had such a huge drop-off from period 1 to period 2 (equivalent to comparing one season against another in the NFL) before making a huge rebound in period 3. The latter part of the sample, though, doesn't much look like a mutual fund, as the deviations from one "season" to the next aren't very large.

Whether other batters are more or less streaky than Gwynn in '97 remains to be seen.