In the Watchtower analysis of the Lions and Vikings, I stuck my neck out for the first time—and predictably, the axe fell:
The most likely outcome of the game is a close Lions win, with above-average rushing performances from both sides, and a 21-17 final score.
What went wrong? Well, there are two major components to the Watchtower breakdown: first, an analysis of past performance when one coach meets another coach. Attempting to control for the varying skill of the players, I try to find out if one coach has a schematic and/or playcalling advantage over the other—if one “has their number.” I try to identify the mechanism too—say, a certain offense is typically depressed when facing a certain defense, because that defense tends to sack the quarterback more often than that offense usually allows.
The systematic advantage of Gunther Cunningham's defense over Brad Childress' offense is one of the most consistent that I've found. Linehan's performances against Frazier (and Dungy) have been consistent, too. As a result, my two Vikings predictions were quite nearly spot-on. I was extremely confident in the respective effects I'd identified . . . but what was I saying about two major components? Oh, yes.
The second part is identifying what the “expectations” are. This is easy at the end of the season, when the teams have played ten or eleven games and their averages are fairly well-established—but at this point in the season, it’s mostly guesswork. Technically, the Lions were averaging 23 points per game after the first two games--but, that doesn't count the Calvin Johnson touchdown that we all know actually happened. Their "real" scoring average (obviously, mean of just two data points is inadequate for a projection like this anyway) is 26.5.
I assumed that that will be their scoring average at the end of the season; a reasonable assumption given the defenses they've faced, the eventual return of Matthew Stafford, etc. So, when I applied 26.5 ppg to the Vikings' 14 ppg allowed, that got me an expectation of 20 points. Apply a mild systemic advantage, and you get this:
Given an equal (or slightly lesser) level of talent and execution, and a mild systemic advantage, the Lions should roughly meet their season averages, scoring 20-24 points.
Of course, the Lions scored only ten points, so it looks like this is a huge failure. But Shaun Hill threw two end-zone interceptions from within the Vikings' ten-yard-line; if you simply switch those to touchdowns, that's 24 points on the nose. Then again, Shaun Hill turning it over against a strong Vikings D is something to be expected. It's probably why my offensive "Mitigating/Augmenting Factors" section was wholly mitigating, and ended like this:
The Vikings have held the vaunted Saints offense to just 14 points, and the much-less-vaunted-but-nothing-to-sneeze-at Dolphins offense to only 7. Even with a systemic advantage tilting the field towards the Lions, this is an extremely stout defense. It makes me very, very nervous.
On the defensive side of the ball, it was the same story: little data to establish "expectations" with, but a strong historical trend. Unfortunately, the Stefan Logan fumble that handed the Vikings the ball within striking distance skewed these results. While turnovers are a part of offense/defense interaction, this analysis doesn't factor in special teams at all--and in this case, the Vikings got a free drive AND incredible field position out of a very rare special teams mistake. In light of that, a projection of 17-20 points is not far off the mark.
You can see how that data led me down a primrose path, especially as regards the offense. I maintain that if Best (a major factor in the Lions' scoring "norms" for 2010) had not gotten hurt during the game, or if that was Matthew Stafford leading those fourth-quarter drives instead of Shaun Hill, I wouldn't look like quite the fool I do.
And yet, the egg is on my face. Anyone have any bacon?