[Skanedogg]

i heard brad edwards from ESPN on a local sports radio station a few weeks ago and he said something very interesting... he picked out four stats and said if team wins those four stats they will win the game 90-95% of the time. those stats were: rush attempts per game, yards per pass attempt, 3rd down % and turnovers.

i know that many cappers think those (of us, since i do it also) who try and use stats as a means of predicting games are morons, especially in the area of a "hard line formula." i personally think that stats have tobe some indicator of future games, we just are unable to put them into their proper perspective.

thus, my question, i suppose is this... what do you guys think about quantifying these stats into some sort of comparison for two teams. basically, i guess it would be under the idea of "if team A has led these factors in every other game, they will lead team B in it this game." (the one wild card, i think, is turnovers, but i think a team's turnover margin is a fair indicator of how well they take care of and take away the ball.) the follow-up, obviously, is how?

[Justin7]

You can use regression analysis to make projections for these stats (and even game scores). If a team deviates wildly from a projection, the team was lucky or unlucky... Likewise, if a team is outside the norm (i.e. 15% or 60% on 3rd down conversions), it will revert toward the mean (although crummy offenses are likely to convert 25%, and spectacular teams hit 50%, and most are 35-40%).

Just looking retrospectively though, these stats are sort of game summaries. Here's mine:
The team with more points wins 100% of time
The team that has the most QB kneels in the second half wins 100%
Win turnovers... 70%
Win 3rd down conversions 65%
more rush attempts ... you're winning.

Realize, all the indicators you mentioned are not very predictive, but more likely to be from the game state of winning.