Using the "Week That Was" data

For several seasons now we have been publishing the weekly actual "matchup stats" for the games just played, including such data points as Play Success Rates, Big Gains, Average Yards per Play (counting penalties), and overall Effectiveness.

Sample 'Week That Was' game report from the 2003 season

Green Bay at Detroit Matchup Rush PSR Rush Big Rush Yds Rush EFR Pass PSR Pass Big Pass Yds Pass EFR Stats Score Green Bay 41% 6% 3.3 76 47% 10% 6.6 80 21 Detroit 43% 6% 3.0 68 32% 3% 4.2 48 3 THE WRONG TEAM WON! According to the linemakers the Pack were anywhere between a 6 1/2 and 7 point favorite. Sure didn't play like it to me, but da numbers say that the Green Men were actually the better team. Lucky Lions.

The question becomes whether the charts above can provide any predictive power for upcoming games in assessing how teams are likely to fare. Our contributor Tim has already written a research piece looking at one of the obvious areas for measurement, namely what happens in the game following one where the "Wrong Team Won" (the stats score had a different winner than the actual score of the game, as in the case above).

Tim found preliminary evidence to indicate that 'Wrong Team Won' LOSERS have been good bets the next game as underdogs -- see the Full Article for the details and respective charts.

The next area we will explore is in cases where the stats score shows a wide margin (which we defined arbitrarily as 17+ points), whereas the actual score was reasonably close (which we defined as under a touchdown), but with the "Right Team" winning.

Case Study #3:
Teams losing a close game, while being soundly defeated on the "stats score"

Analysis: Underdogs in such a situation have been very successful against the spread, with a 44-25 (64%) record. Oddly enough the performance of dogs is mediocre when getting 3.5 to 7 points, but huge when getting a field goal or less, or more than a touchdown. We would not however be swayed by that and would emphasize all dogs for the time being.

Favorites meanwhile are a remarkable 0-10 against the line in the week following a loss of 1-2 points when the stats score had them being beaten by 17+ points. Whether this is a true anomaly or a sign of something resembling the "horrible defeat syndrome" where a tough, tough loss has carryover effects is unclear. The chance of randomly seeing an 0-10 pattern is however 1 in 1024, so a rare enough event that it might be worth noting when it occurs.

When the margin of victory in a lopsided stats score is a touchdown or more there are less obvious implications for the ensuing week, except that teams coming back as favorites off a loss of 7-13 points when beaten by 17+ on the stats side have been poor wagers (27-45) -- perhaps suggesting that the prior game was always in hand for the other team but the final score was closer than it actually was?

Case Study #4:
Teams winning a close game, while being soundly better on the "stats score"

Analysis: Here again there appears to be some value on the underdog, with a 39-25 (61%) overall record, and an excellent 34-17 mark when the team in question returns as a dog getting a touchdown or less.

On the other hand, returning as a big favorite after this 'pulling out a tight one that should have been easy' has been a negative, with a poor 19-30 history against the point spread.

There are no noticeable effects for games won by a touchdown or more with a healthy stats score edge.

Now, the logical progession would be to next look at teams that won in blowouts where the stats score was close, but surprisingly this is a rare occurrence (i.e. reversing the definitions used above to see instead games where a team won the stats score by 1-6 points but won the game by 17+ produces just a sprinkling of cases over a five year span).

Future research into this area will look at the details of the rushing/passing stats in relation to the final scores.