Eight weeks ago, the Vikings were undefeated and ranked No. 1 in DVOA. Now they're 6-6 and probably done. What went wrong?
06 Dec 2004
Guest Column by William Krasker
During the 2004 season at footballcommentary.com, I have been analyzing notable coaching decisions on a regular basis. In this article, special for Football Outsiders, I reexamine a few of those decisions, and also discuss some new ones.
In the Week 12 game between St. Louis and Green Bay, St. Louis trailed 21-10 with 3:50 left in the 3rd quarter. Facing 4th-and-7 at the Green Bay 24-yard line, the Rams lined up for an apparent field goal attempt. However, St. Louis coach Mike Martz had called for a fake. Holder Dane Looker passed the ball to kicker Jeff Wilkins, but Wilkins was stopped immediately for a loss, and the Packers took over on downs.
In my original analysis, using the footballcommentary.com Dynamic Programming Model, I concluded that the fake field goal gives the Rams about the same probability of winning the game as an actual field goal attempt. I won't repeat the analysis here. Instead, I want to clarify some of the assumptions, and explain why my original conclusion was misstated.
The Rams had the option of going for the first down openly. Necessarily, then, if Martz chose a fake field goal, he must have thought that it gave a higher probability of success. Rather than accept that judgment, in my original analysis I used probabilities that would be appropriate if the Rams had openly tried for the first down.
Strictly speaking, then, the conclusion of my original analysis is that going for the first down openly gives the Rams about the same probability of winning the game as attempting a field goal. Any conclusion about the fake field goal per se requires additional assumptions, which I didn't want to make at that time.
Actually, as I look over where the Green Bay defenders were positioned before the snap, it appears that the particular play Martz chose was doomed before it started. Perhaps the real lesson from that play's failure is that if you call a trick play in a crucial situation, you need a procedure for calling it off prior to the snap if it appears likely that it isn't going to work.
I'm reminded of the "fake field goal" New England used successfully against St. Louis in Week 9. On 4th and goal from the 4 yard line, the snap went directly to kicker Adam Vinatieri, who passed to Troy Brown for a touchdown. As an aside, I don't actually think this play should be called a fake field goal. A fake field goal is a play, other than a field goal, run from a field-goal formation. The Patriots were not in field-goal formation; at least I have never seen them kick a field goal with a wide receiver standing next to the sideline. But that's just terminology. I mention this example because if a St. Louis defender had been anywhere near Brown, New England presumably wouldn't have run the play at all. They would have called timeout, or taken a delay-of-game penalty, and then kicked a field goal.
With 1:18 left in the Week 9 game between Oakland and Carolina, the score was 24-24, and Oakland had 1st-and-goal at the Carolina 3-yard line. Oakland had three timeouts, Carolina none. Before I describe what actually happened, I'll work through (perhaps in more detail than you need) what should happen if both coaches are rational.
I'll begin with some useful information. Place kickers miss short field goals about 2% of the time. Kickoffs are returned for a touchdown roughly 0.7% of the time. If it's do-or-die for the return team, so that they are willing to try multiple laterals, the probability might be as high as 1%, but I won't go higher than that unless a marching band comes onto the field.
It's generally believed that Carolina should let Oakland score in this situation, but that's not quite true. Certainly, if Oakland scores a touchdown, Carolina has the opportunity to tie the game again with a touchdown of their own. But once the game clock gets so low that Carolina's probability of scoring that subsequent touchdown is smaller than Oakland's probability of missing a field goal, Carolina should try to stop Oakland. That's certainly the case if there is only enough time left for a kickoff return. The general result is that if the game clock is below some amount t*, Carolina should try to prevent Oakland from scoring a touchdown.
Now, if Carolina is going to try to stop Oakland, then Oakland shouldn't try to score a touchdown, because the risk of a turnover more than offsets any potential benefits. So, once the game clock dips below t*, the solution is for Oakland to let the clock run down to 0:02, call timeout, and kick a walk-off field goal. Carolina can then win only if the field goal misses (probability 0.02) and then Carolina wins in overtime (probability 0.5). Oakland thereby reduces Carolina's probability of winning to
While the game clock is above t*, Carolina prefers that Oakland score a touchdown, rather than continue to let the clock run; but of course Oakland shouldn't take the bait. The conclusion is that with rational coaches, Oakland will take a knee twice to reduce the game clock to 0:02, call timeout, and kick a field goal.
So what actually happened? Oakland coach Norv Turner incorrectly had his offense try to score a touchdown on both 1st and 2nd down. Carolina should have let Oakland score at that point, but coach John Fox had his defense try (successfully) to stop both attempts.
Oakland finally saw the light and set up for a field goal. Even then, Oakland erred by stopping the clock with 0:09 left, guaranteeing that they would have to kick off to Carolina following the field goal, and virtually doubling Carolina's probability of winning the game.
Toward the end of the Week 6 game between Kansas City and Jacksonville, a common endgame situation arose. With 3:52 left, Jacksonville trailed 16-14, and had the ball at their own 33 yard line, 4th down and 1 yard to go. Jacksonville coach Jack del Rio decided to go for it rather than punt, and was criticized for the decision even though the Jaguars ultimately won the game.
If the Jaguars punt, their main hope for winning the game is to force the Chiefs to go three-and-out. In that case, after the exchange of punts, Jacksonville can expect to get the ball back somewhere near where it is now. So whether they punt or go for it, Jacksonville's aim is to get 1st and 10 at around their own 33 yard line.
I think it's safe to say that many, if not most, coaches would punt in this situation. It's natural for the coach to focus on the desired scenario -- we'll force them to go three-and-out and then get the ball back -- without thinking about exactly how likely that scenario is. In addition, punting postpones the moment of reckoning. Coaches (and some commentators) prefer decisions that are virtually certain to keep a team's hopes alive. Phil Simms once proclaimed prolonging the game to be a basic principle, and St. Louis coach Mike Martz gave a similar, if less concise, explanation for playing for overtime in last year's Divisional-round game against Carolina. But the object is to win the game, not prolong it. The starkest example of the distinction between these criteria arose last year when Houston played Jacksonville. The Texans, trailing 20-17, had the ball at the Jaguars' 6-inch line with 0:02 left in the game. Many observers were shocked that the Texans would go for the touchdown, and risk losing the game right there. But it's obviously the right thing to do.
Returning to the Kansas City-Jacksonville game, if the Jaguars go for it, there is only about a 30 percent chance that they fail to pick up the first down. And even if they fail, they could still win by holding Kansas City to at most a field goal. If they punt, there is roughly a 30 percent chance that they don't get the ball back at all, in which case they have no chance to win. (Indeed, 18 of Kansas City's 50 possessions up to that point in the season had consumed more than 3:52.) Even if the Jaguars do get the ball back after punting, it will be with much less time on the clock. It's clear, therefore, that del Rio's decision go for it was correct.
With 14:17 left in the 4th quarter of their Week 11 game against Detroit, Minnesota scored a touchdown to close the deficit to 19-13 prior to the try. The Vikings elected to attempt a two-point conversion.
According to the Model, a two-point conversion in that situation is appropriate only if the probability of success exceeds 0.46. It's unusual for a coach to call for a two-point conversion in a situation in which such a high probability of success is required, but this isn't the first time Mike Tice has done it. In 2002, against New Orleans, he went for a two-point conversion when trailing by 1 point with 0:05 left in the game. That decision makes sense only if the probability of success exceeds about 0.5.
In my Week 11 article, I also mentioned that Minnesota had chosen to kick an extra point in Week 10 versus Green Bay, in a situation in which going for two required only a 0.38 probability of success. For that I accused Mike Tice of inconsistency. My criticism was unfair, because it's certainly possible that Minnesota's success probability on two-point conversions varies from opponent to opponent.
I generally assume that the average team's probability of success on a two-point conversion is 0.4, but in fact there is considerable uncertainty about that number, mainly because teams attempt very few two-point conversions. During the 2003 regular season and playoffs, there were 68 attempts, of which 31 were successful. In 2002, teams were 47 of 93, and so far this year the results are 26 of 50. I hesitate to go farther back, because probabilities can change over time, but if we combine these three years we get 104 of 211, or 49%. Even if we assume that the success probability was the same for each of those attempts, so that there is a well-defined quantity being estimated, the standard deviation of the estimate is about 0.035, leading to a 95% confidence interval of
That inference isn't legitimate, though, because there is variation from team to team. The probability of success on a two-point conversion depends, at least to some extent, on both the team attempting the try and on the opponent. In addition, teams are presumably more apt to attempt two-point conversions when they believe their chances are high. For example, trailing by two points with a few seconds left in the game, every team would attempt a two-point conversion. But leading by five midway through the 3rd quarter, I suspect that only a coach who feels good about his team's chances would go for two. The data are therefore presumably subject to a "selection bias," meaning that teams with a high probability of success are over-represented. Consequently, there is good reason to believe that 49% is an upward-biased estimate for the average team. But there is also reason to believe that 40% understates the success probability for many teams.
Obviously, from data on actual attempts, there is no hope of estimating team-specific probabilities for success on two-point conversions. There would be only a handful of observations per team. However, it should be possible to draw some inferences from a team's success in other short-yardage situations.
In a recent article I described situations in which the team that trails can benefit from an intentional foul prior to the last minute of the game. In some cases this tactic gives the team its only chance to win. So far this season there have been many cases in which a team could have used this tactic, including two games in Week 11 alone. Here I want to describe a situation in which the team that leads can benefit from a penalty.
Let's begin with a quiz. How much time can the offense run off the game clock between plays? If your answer is "40 seconds," you're forgetting about penalties. Suppose there is a penalty while the ball is dead between plays. If the game clock was running, it stops for the penalty. Then, except in the last 2:00 of the first half or the last 5:00 of the second half, the game clock restarts (Rule 4-3-1) with 25 seconds on the play clock. This sequence occurs in virtually every game. In particular, by letting the original 40-second play clock expire, the offense can take 65 seconds off the game clock in total, at the cost of five yards. (A false start just before the play clock expires works just as well.)
This is a valuable tactic in certain situations, and every team should be prepared to use it when appropriate. Consider for example the Week 4 game between the Jets and Miami. The Jets, leading 17-9 with the game clock running at 6:40 in the 4th quarter, punted on 4th and 15 at the Miami 43 yard line. A 5-yard penalty wouldn't have mattered -- punter Toby Gowin sailed the kick beyond the end zone for a touchback -- so the Jets should have let the play clock expire and killed another 25 seconds.
Actually, there is nothing explicit in the rule book sold to the public (except a timeout by the defense) that prevents the offense from repeating this over and over. To be sure, "more than two successive penalties, during the same down, after a warning is unsportsmanlike conduct." (Supplemental Note 2 to Rule 4-3-10.) But that just increases the distance penalty; the game clock still restarts. I'm assuming there is some other rule that could be invoked. Perhaps it's the vague Supplemental Note 4 to Rule 4-3-10 ("Certain acts of delay may involve stopping the game clock immediately"), or maybe it's something in the supplemental material to the rule book that's not readily available to the public. (I'll have more to say about that later.) But it's amusing to consider what could happen otherwise: The team that receives the second-half kickoff could take a knee on first down to start the clock, run out the third quarter with repeated dead-ball fouls, take a knee on second down to begin the fourth quarter, and then take the clock down to 5:00 with more dead-ball fouls. (At that point the offense would be about 10−14 inches from its own goal line, so the last few half-the-distance penalties would be pretty tricky to pace off.)
(Ed. note: Not to mention Joe Theismann would throw a hissy fit like you wouldn't believe.)
The Jets have a clock-management guru. Maybe they need a coin-toss guru.
The Jets won the coin toss prior to their Week 9 game against Buffalo. Presumably because the wind was from the southwest at about 25 miles per hour, coach Herman Edwards elected to defend the west goal. In my original article I pointed out that since the teams change ends after the 1st and 3rd quarters, each team gets the wind for half the game regardless of who kicks off for either half. Therefore, there is no advantage to electing to take the wind in the first quarter. However, my criticism of Edwards' decision puzzled some readers, who assumed that the Jets were entitled to receive the kickoff for the second half. Actually, Rule 4-2-1 says that after the pre-game coin toss,
The winner of the toss must choose one of two privileges and the loser is given the other. The two privileges are
(a) which team is to receive; or
(b) the goal his team will defend.
For the second half, the captain who lost the pregame toss is to have the first choice of the two privileges listed in (a) or (b).
The Bills, as coin-toss losers, had their choice of privilege to start the second half. Naturally, they chose privilege (a) and elected to receive. Thus, Buffalo received the kickoff to start both halves, and the effect of Herman Edwards' decision was simply to reduce the Jets' expected number of possessions in the game. According to the Model, the decision lowered the Jets' probability of winning the game from 0.5 to 0.464. This is actually worse for the Jets than spotting the Bills a one point lead (0.47).
If the game is being played in a hurricane, and the storm's eye is forecast to pass through at the end of the first quarter and reverse the wind direction, then it might make sense to take the wind. But in every realistic case, if you win the pre-game coin toss, you should elect to receive.
An interesting situation arose just before halftime of the Week 12 game between Baltimore and New England. The Patriots, leading 3-0, took over at their own 16-yard line with 1:08 left. Content to run out the clock, the Patriots had Tom Brady take a knee on first down. The Ravens refused to cooperate, though, and belatedly called their second timeout at 0:56. On 2nd down, Corey Dillon ran out of bounds, effectively giving the Ravens an extra timeout, and setting up a sequence of events in which the Patriots were fortunate to give up only a field goal.
My first reaction when I read this was, Why didn't New England have Brady continue taking a knee? The Ravens can call their final timeout, but if the Patriots are careful to use the play clock, Baltimore will get the ball back with only about 0:10 before halftime.
In his post-game press conference, New England coach Bill Belichick explained that by running Dillon wide, he hoped to use extra seconds and leave Baltimore with next to nothing when they got the ball back. That's fine.
However, from Dillon's post-game comments, it appears that during the timeout that preceded the 2nd-down play, the New England coaches didn't mention staying inbounds. Dillon said that he should have looked at the clock and figured it out himself, but when staying inbounds trumps other considerations, it's the job of the coaches to emphasize it.
A related example arose in the Week 7 game between Arizona and Seattle. On 3rd down with 2:00 left in the game, Arizona's Emmitt Smith scored on a 23-yard run which, following the extra point, gave the Cardinals a 25-17 lead. Since Seattle had no timeouts, Smith could have downed himself after picking up the first down, and the game would have been over. The moment he crossed the goal line, Smith put Seattle right back into the game. Again, it's not Smith's responsibility to figure this out. It's the coach's job, and the two-minute warning would have been the ideal time to go over the situation with the players.
The development of good tactics, and the evaluation of tactics observed on the field, sometimes require a clear understanding of the rules. However, unlike the other major sports, the NFL does not make its official rules available on its Web site, nfl.com. Instead, there is a Digest of Rules, which is difficult to locate on the site, and so nearly useless that it's not worth the effort. If you are really interested in the rules, the NFL forces you to purchase the book Official Rules of the NFL. Even then, you won't have all the information needed to understand what's happening on the field. "For reasons of space," the copyright page of this rather small book says, it omits certain "supplementary material." The omissions are not always innocuous. For example, editions prior to 2004 omitted the infamous Note 2 to Rule 3-21-2, which says that when a player is holding the ball to pass it forward, "any intentional forward movement of his hand starts a forward pass, even if the player loses possession of the ball as he is attempting to tuck it back toward his body."
The rules and all supplementary material should be online where they would be accessible, free, and searchable.
My final remarks concern the on-screen displays that the networks use during games. Every network shows the score, the time remaining, the down, and the number of yards to go for a first down. However, only ESPN shows the play clock at all times. On the other networks the play clock pops up only when it gets down to five seconds. That isn't good enough to allow a proper evaluation of how the team that's ahead is managing the clock. I hope the other networks follow ESPN's lead.
1 comment, Last at 20 Sep 2006, 7:12am by Klassenfahrt