This is analysis of a UC Berkeley report on WHEN you should kick and when you should go for it on fourth down...more in formation after the article:
Hey, NFL coaches! Want to win one more game this season than you otherwise would? Below I'll tell you how. In fact, I will reveal a simple formula that will increase the odds of victory for most football teams.
And I won't just be whistlin' Dixie. The formula was tested over the offseason in thousands of computer simulations by the sports-analysis firm Accuscore -- the same Accuscore that, during the 2006 NFL season, compiled a better game-predictions record than anyone in the ESPN television, radio and Internet empire. Accuscore and I spent some time over the winter devising and testing various assumptions about football tactics, arriving at one formula that almost always improves the chance of winning. Ideally we'd like to sell the formula to NFL franchises for huge amounts of money, then spend the rest of our lives riding around in limousines. But what the heck, in a moment I'll simply give it away.
First, some history. Year in, year out, Tuesday Morning Quarterback rails against excessive punting. I document "preposterous punts" -- punts on fourth-and-short in opposition territory or on fourth-and-1 when trailing in the second half. My archive of preposterous punts now numbers into the hundreds, and shows teams that punt on short-yardage situations in opposition territory, or when trailing in the second half, almost always go on to lose.
Elaborately, I've argued that if NFL, college or high school teams went for it on most fourth-and-short downs, the additional scoring from sustained drives would more than offset the field position surrendered by occasional failed tries. I've quoted Don Shula as privately telling a powerful insider (me) that not punting would revolutionize football. I've shown that since the average offensive play gains almost five yards, going for it on fourth-and-3 or less mathematically favors the offense: and that this would hold even if nobody ever punted on fourth-and-3 or less.
Last fall, I worked through the probabilistic pluses and minuses of rarely punting, concluding, "Probabilities suggest a team that rarely punts will increase its scoring" but not increase opponents' scoring. Last fall's anti-punt column also highlighted an academic paper by economist David Romer of the University of California at Berkeley, who contends NFL teams should go for it even on most fourth-and-longs. Finally, I've railed against the two reasons coaches order punts on fourth-and-short. First, "because that's what we always do." Second, because if coaches order fourth-down tries that fail, they will be blamed, whereas if coaches order punts, the players will be blamed for the loss.
Because coaches are afraid of being blamed, my anti-punt theory has never received a systematic trial. In my own coaching of middle-school, county-league flag football, I haven't sent the punt unit on the field in two years, and have posted two undefeated seasons. But since punting happens less in flag than in tackle anyway, this might not tell us much. Until such time as some college or pro coach decides to believe what he reads in TMQ, there won't be a real-world confirmation of anti-punt thinking.
So, I asked the computer whiz-kids at Accuscore, whose software simulates entire NFL games, if they could take actual games from the 2006 season and rerun them with everything the same except one team eschewing the punt. We chose three types of games: great games between top teams (for instance, New England at San Diego in last season's playoffs), good games between good teams (for instance, Denver at St. Louis in the regular season) and average games involving average teams (for instance, we tested the home-and-home series between the 49ers and Cardinals). We defined two sets of punt-shunning rules: the hyperaggressive pedal-to-the-metal tactics advocated by Romer and a somewhat more conservative set of anti-punt rules designed by TMQ. We assigned the anti-punt tactics to the home team but not the visitor, then to the visitor but not the home team in the same pairing. Accuscore simulated about 10,000 sets of games, to wash out the effects of chance.
Bottom line: avoiding punts added an average of one point to a team's per-game scoring, without adding any points to its opponents' average scoring. Teams avoiding punting became 5 percent more likely to win -- statistically significant owing to the thousands of tries. Doesn't sound like much? One more point scored per game represents the difference between the Bengals and the Patriots of the 2006 season. Last season, one additional victory would have put the Packers, Panthers or Rams into the playoffs. A 5 percent improvement in victory likelihood translates into one additional victory per 20 games, or just shy of one extra win per NFL season. I think any NFL owner would gladly pay millions of dollars for one additional win per season.
Accuscore tested the hyperaggressive approach advocated by Romer in this paper, and also tested TMQ's somewhat more conservative tactics. The Romer rules, derived from his statistical study of an entire NFL season, are as follows. Go for it on fourth-and-4 or less from anywhere on field, even deep in your own territory; go on fourth-and-7 or less inside the opponent's 45; go on fourth-and-10 or less inside the opponent's 33 (except that inside the opponent's 33, attempt a field goal in the fourth quarter if a field goal causes a tie or gives you the lead). My own rules I'll describe in a moment.
Accuscore found when high-quality teams -- especially last season's Chargers, Colts or Patriots -- employed either Romer's very aggressive anti-punt tactics or my somewhat less aggressive anti-punt rules, their chances of winning improved by about the same amount. That both sets of rules worked for quality teams suggests good teams are more likely to be able to convert first downs -- and thus the better the team, the more that team might profit by rarely punting.
There was one exception: rarely punting slightly reduced the odds of victory for the 2006 Baltimore Ravens, which had a weak offense but the league's best defense. For the mid-quality teams, such as last year's Broncos and Chiefs, and for average teams such as last year's Niners and Titans, Accuscore found the hyperaggressive Romer strategy was volatile: making victory a lot more likely about two-thirds of the time, but decreasing the odds of victory the other one-third of the time. For the same group, TMQ's rules added somewhat to the odds of victory while almost never backfiring.
AP Photo
Stop me before I punt again!
The Conclusion: unless you are the Baltimore Ravens or Baltimore-Ravens-like, use the Tuesday Morning Quarterback Sure-Fire, Never-Fails, All-Weather, Computer-Tested, Victory-Enhancing, Call-Me-in-My-Limo Guidelines. Here they are, and sorry there was no way to simplify:
• Inside your own 20, punt.
• From your 21 to 35, go for it on fourth-and-2 or less.
• From your 36 to midfield, go for it on fourth-and-3 or less.
• From the opposition 49 to opposition 30, go for it on fourth-and-4 or less.
• From the opposition 29 to opposition 3, go for it on fourth-and-3 or less.
• From the opposition 2 or 1, go for it.
• Exception: inside the opponent's 25, attempt a field goal if it's the fourth quarter and a field goal causes a tie or gives you the lead.
A few notes on my sure-fire formula: By only disdaining the punt in situations in which the odds of success are pretty good, my anti-punt strategy takes into account player and crowd psychology. Because fourth-and-short attempts will usually succeed, players will remain upbeat, while the defense will understand that though it occasionally will be sent out with bad field position after a blown try, overall, the team will benefit from rarely punting. Romer's more aggressive strategy creates too much chance of a lustily booing home crowd, or players who think their coach is a fool after, say, a missed fourth-and-3 attempt from your own 10-yard line: and psychology is a big factor in football.
My strategy also values field goals more highly than does Romer's: field goals are nothing to sneeze at, so it makes sense to attempt them on fourth-and-long. Finally, my rules violate my own immutable law of field-goal decisions, namely, Kick Early, Go For It Late. I couldn't think of a way to incorporate Kick Early, Go For It Late into the decision-making tree without causing Accuscore's job to become excessively complex. At any rate, the simulations showed that unless you're in the fourth-quarter exception, statistically you're better off going for the touchdown on fourth-and-goal from the 2 or the 1 -- disproving the Kick Early, Go For It Late law, which can no longer be viewed as immutable. (I've sent the football gods a memo on this.)
By the way, this "revolutionary" study? It's approaching TEN YEARS OLD...
