Two coaching decisions, a review

Scenario 1: Your team is up 28-23, though the other team has moved the ball quite efficiently all game. There is 10:56 left in the fourth quarter, and you have fourth and goal from the one yard line. A field goal puts you up by eight points; a touchdown probably ices the game. (“Checkmate,” as Urban Meyer would say.) What do you do?

Scenario 2: Your team is up 21-17. The other team has the ball on roughly your two-yard line. Thirty-six seconds remain; they have just run the ball on second down so the clock is moving. They have no timeouts, but you have all three of yours. The other team has just quickly driven the field to get into this position. Question: do you call timeout to preserve some time for yourself in the chance that they score a touchdown on third or fourth down? Or do you leave the pressure on them to execute on those two downs over thirty-six seconds. What do you do?

Analysis (and identities of the coaches) after the jump.

Scenario 1 – What actually happened: Mike Leach, with his team leading 28-23, called a quarterback sneak on fourth and one that was stopped. His team eventually lost to Houston, 29-28, in the waning seconds. Unsurprisingly, he has taken a lot of heat for not kicking the field goal. Indeed, it looks like he has even recanted, saying, “If I put the field goal team out there, at that point in the game that’s the better thing to do,” Leach said. “And I didn’t do it.” But should he feel bad?

I’m not so sure. I will say this for certain: if it is the first half, I think we can all absolutely agree that he going for it was the right call. Fourth and one is quite literally almost always worth it from just about anywhere on the field, and especially so on the goal line. The analysis goes something like this, though see the link above for more: Afield goal attempt has an expected value of about 2.5 points (the chance of making it multiplied by three). The value of “going for it” is a bit murkier, but from that part of the field the chance is worth about four points, give or take some of the yardage values. You arrive at this number by taking the chance of success times six, plus the subsequent chance at a good PAT times one, plus factoring in the opponent’s horrible field position if you fail. These numbers are fairly quantifiable, and even if they aren’t to an exact degree we know that four, the expected value of going for it, is higher than two-and-a-half. (Again, people often treat field goals as if they were automatic.)

The question is whether things change that late in the game, in the fourth quarter. Going back to Urban Meyer, I think that depends on the flow of the game and the type of opponent. And the biggest criterion is whether you think there will be more scoring. My sense is that Leach’s impulse was that eight points was probably not enough to straight win the game at that point (Houston wound up with over 600 yards of offense), though, with perfect knowledge in hindsight, it turns out that it would have been. But these decisions are made ex ante, and it still strikes me as the right one.

Now a lot of people who are understandably upset, maybe Leach himself, say, “Well, even if I agree with the call to go for it the playcall was bad.” There might be something to this but it strikes me generally as a copout way to say “I just disagree because it didn’t work.” Now the QB sneak itself was pretty ugly, so maybe it could have been practiced better, but it’s not like the team doesn’t work on the sneak or that isn’t the most direct way to convert those plays.

In any event, I’m curious if there are any comments on the call to go for it. I am looking for real analysis though, not just monday morning quarterbacking. I am legitimately interested in the best way to view these decisions as time begins to get more scarce late in games. I think the data supports the idea that the fourth and one on the one is quite literally always the right call in the first half.

Scenario 2 – What actually happened. Danny Hope, Purdue’s first year coach, called a timeout with 36 seconds left. It turns out, however, that Notre Dame’s head coach, Charlie Weis, had signaled in the “spike play” to be run on third down. As a result, because we know what Weis’s decision was going to be, the result of Hope’s timeout was to give Notre Dame two chances to score — one on third down and another on fourth — rather than one (a spike on third down and just a fourth down play). Many people excoriated Hope, including the announcers, and well, me, via twitter. But now I’m not so sure I was right. Several people pointed to my own article, which makes clear that Notre Dame had no need to spike the ball on third down — thirty-six seconds is plenty of time to run two plays. And here’s Hope’s defense of his call in his own words:

“I want to retract my statement from last night. Last night, I wasn’t sure it was the right call, but now that I look (back) at it, I’m sure it was the right call, in spite of what a lot of the experts and the critics think.

“You know, it was second-and-goal and they ran it. I was on the headset with Coach (Gary) Nord and any time you get down there, on first- or second-and-goal inside the 5-yard line, the probability of getting a touchdown is pretty good. … I don’t know what the statistics would say, but it’s very significant, the chances that when you get down there, you’re going to score.

“If that was going to be the case, we needed to get in a position to get a field goal to go into overtime or a touchdown to win. We couldn’t do that if there wasn’t any time left on the clock. I was on the headset with Coach Nord and said that if they threw a pass, one of two things would happen, that they might get a touchdown or they might throw an incompletion that would stop the clock. But if they ran the ball and it’s not a touchdown, the clock’s going to keep running, and I don’t want to run out of time.

So I got the official’s attention (and said) that if they ran the ball and it wasn’t a touchdown, that I wanted to call a timeout as soon as I could to save as much time as I could on the clock. I don’t have any reservation about that at all.”

I think this is quite persuasive. Now, again, with hindsight, the only scenario that would make the timeout a bad idea actually was the one happening: that Charlie Weis would make the amateurish call to waste third down by unnecessarily spiking the ball on third-and-goal with over 30 seconds left, which was plenty of time to run two plays. But how could Hope know that? I suppose you could say you could wait to see if your opponent was going to do something so lucky for you, but that might waste precious seconds.

And it offends my sense of order to say that Hope erred by not presuming that his adversary would do something so suboptimal, even if [insert Charlie Weis joke]. Unfortunately for Hope, he saved Weis from himself, but otherwise he left Purdue in a better position than it would be otherwise: ND still had two chances to score, and Purdue still had more time than it would have had otherwise to put together a drive. Although it didn’t work out for Purdue, that Charlie Weis after all these years doesn’t know his clock management well enough to see what a huge waste of a down spiking the ball would have been with that much time is his problem, not Hope’s. They just need to play better defense in that situation (i.e. not giving 25 yard cushions on third and 18 to give up the 20 yard comeback….).

But, as with the Mike Leach scenario, comments and criticisms necessary. And, for this second one, apologies to Coach Hope as I initially thought it was the wrong call. I retract.

  • Andrew

    I think Leach’s decision was a good one. I agree that the play call might have been better, but there’s no reason to think a sneak like that wouldn’t work (esp. as they’ve had success with sneaks in the past). It was a good call given how Houston had been playing: they had a good shot at overcoming 31-23, as well. They did what they had to win with the score 28-23, but who’s to say that they couldn’t have recovered an onside kick or something to get a field goal and win it from 29-31, if Leach had decided to kick a field goal.

  • bobby-o

    “I think the data supports the idea that the fourth and one on the one is quite literally always the right call in the first half.”

    You give the expected value of going for it as 4 points; what is the expected value if there are only a few seconds left in the first half, and the 4th down play is the last play of the half? In that case, you would not have any bonus for giving the other team the ball on their own 1; the expected value would just be the value of the TD+PAT against the odds of making it.

    My intuition would say that this expected value is still greater than the 2.5 of the FG, but obviously less than the usual 4. I am curious how the clock factors into it.

    On a related note, I am curious how similar analysis would impact the last part of the LSU-MSU game, especially how LSU should play given the limited clock after the turnover on downs.

  • http://walkonboy.blogspot.com/ Walk On Boy

    The always-go-for fourth and short is a great idea in theory. And logically it is optimal in the long run, however when it doesn’t work out people shut it down in the short run. So next time, fans and more importantly coaches may be a bit hesitant to go with what is statistically better the vast majority of the time. The opposition and scheme have something to do with the play call as well. Although Tech sneaking it seems like a great idea, how many times do they run the sneak in practice as opposed to a short/medium range route. Also (didn’t see the play) but assuming that it was under center. How many under center snaps do they take in a week?

    This is the same for going for two early in the game. Just because a team is down two doesn’t mean they should go for two. There will be more chances and scoring to even out the gap. Not to say if there is a two point play that will work not to run it. Two is better than one. Oregon didn’t need to run a split gate fake against Cal, but they did and put up points.

    There are all sorts of second guessing that occurs when the result is in. The question is figuring out what the optimal play call for a particular team against an particular opponent at a unique time in a game.

  • http://fridaynighttailgates.blogspot.com Matt

    I’m a big believer in going for it 4th and anything inside the 3, I just think there are too many variables on your side that in most situations it is the correct call. I’m a Texas fan and wish Mack would have gone for it in a couple of situations early in the game on Saturday (although it obviously didn’t matter in the long run).

    NOW, having said that, I think that late in the game you probably need to kick the field goal just so that one play doesn’t beat you, it just ties you (with the 2 point conversion).

  • Mrfphy

    I disagree with your method of evaluating field goals vs. touchdowns in the end of game situation. Multiplying the value of the score by the probability of success is only valid if you have a high number of attempts, enough to get rid of statistical deviation. At the end of the game, you only get one shot and the smart answer is to take the points. (Proof: if you had a 1 in 100 chance of scoring a hypothetical 1,000,000 point uber-touchdown your odds say go for it…but I’m still kicking the field goal)

    As for the timeout call, the chances of a team scoring once they get down near the goal line are pretty high. But the chance of taking the ball back the other way with less than a minute after the ensuing kickoff is probably REALLY small. Smaller than the chance that your opponent won’t score. That coupled with the fact that a running clock puts all the pressure on your opponent is a pretty compelling argument.

    Of course both these scenarios depend on the situation, opponent, scoring patterns in the game, etc. but I’m arguing both calls.

  • Nole55

    Well, wasn’t Mike Leach right? He wanted to end the game right there, because had a fear Houston would just drive down the field and score, which they did. Now, a field goal still guarentees a minimum of a tie, and Houston still had to make the 2 point try, but lets examine the situation. Texas Tech was in a hostile environment (or as hostile as Houston can get), Houston was moving the ball all day, and holding Tech to a field goal still gives them a glimmer of hope. The safe call is the field goal, but if he wanted to (virtually) assure victory rather than merely increases the chances of it, goin for it on 4th is the way to do it.

    A couple other notes. One, does the spot of the 4th down conversion make any different in calculations? What I mean by that is, what effect does Houston getting the ball back on their own 1 yard line have on the expected points. Also, does Houstons offensive output to that point have any bearing? If they had 450 yards or whatever, im sure the risk increases as opposed to if they had 200.

  • http://smartfootball.com Chris

    Nole55: The spot Houston gets the ball back is very important. If you’re on your own 30 that is far less attractive than being on their one yard line.

    Mrphy: I’m not sure your variance-sample size point makes sense. Yes, expected equals average, but that doesn’t mean that you throw the analysis out the window unless you’re going to have 1,000 4th and 1s. You just extrapolate from these past scenarios to the current one, to ask whether this call, in isolation, is the right one. I’m pretty sure it was, though again the late game aspect could, maybe, change that.

    walkonboy: Point taken, but I don’t think your other example supports your conclusion. Two-point conversions have a very different payoff matrix, because the gain for going for it from the two (not the one) is only two points, not seven. Thus even if you converted 50% of your 2-point conversions your expected value would be one point for both PATs and two-point attempts, but the PAT would be more attractive because of decreased variance. In this situation though, you’ve got a full point and a half improvement on going for it versus taking the three, thus dwarfing most of the variance. (Plus it depends where you are with the hashmarks, as the PAT is nicely centered.)

  • 4.0 Point Stance

    Chris;
    With respect to your last point, about P’sAT being centered, I’ve long felt that coaches don’t understand how difficult the field goal attempt is from the one yard line, on the hash. They think “oh it’s less than a 20 yarder, let’s kick!” but it really is an extreme angle, especially in college, and especially for a kicker going against his body.

    Twice this year I’ve noticed coaches (unfortunately I forget who) who have taken delay of games from the 1-2 yard line to set up an easier angle on the kick. I feel somewhat vindicated.

  • DT42

    With respect to the Red Raiders I think that as a coach you have to know what your team is capable of. Coach Leach is asking an offensive line that operates with wide splits and pass block more in one half then Georgia Tech does in a season to impose their will on a defense on a 4th and 1 at the goal line. I have no problem with the choice to go for it, i think however that if they are in a situation like that again they will go with plays that they have worked on in practice. You should call a play out of your red zone offense that has been worked on. I doubt that the Red Raiders practice a QB sneak.

  • Brad

    My gut says Leach made a bad call. With 10 minutes left to play at 8 points they need to get a touchdown and a field goal just to tie. You are probably only looking at between 4-5 more drives total between the two teams. At that point you goal should be to reduce variance in the outcome not increase expected value.

    Even if they drive for a touchdown on the next play and get a conversion you are still tied with 6-7 minutes left in the game and the ball. If you fail to score and they score you probably get the ball back with time to score yourself. If you score you probably give the the ball with 2-3 minutes and a 3-7 point lead. Houston had not been very good at getting TDs (only 1 out of every 5 drives) and had missed half their field goals.

  • Joe U

    When the second down play in the ND/Purdue ends and there was 36 seconds left, I immediately thought of your article on UCLA’s 2 Minute drill and Homer Smith’s logic on why spiking the ball is always a bad call. Purdue’s defense was disorganized and not prepared. Danny Hope had to call a timeout because Clausen could have run a quick fade or another play, if ND had been prepared and scored easily. The timeout let the defend and stop the next play. With 36 seconds left ND coudl have run two more running plays, let alone two passes.

    Thanks Smartfootball for making me a better football fan.

  • catfish

    Scenario 1 I agree, go for it.

    Scenario 2, I don’t think so. How often does a team get the ball back with less than :30 and drive down the field to score (FG or TD)? You are pretty much betting you get a great kick return. Odds aside, my real problem with this call is that it puts all the pressure on yourself. Calling timeout here is just giving up on your defense and signals to the offense that you think they are better than you. It gives them time to calm down and call two good plays rather than running around trying to communicate and execute, all without making a mistake. This call just makes you look like a loser, that’s my problem.

  • Mrfphy

    Chris;
    Try this: Let’s say you’re in the same position but you have to choose between a 3 point score with a 90% probability of success and a 6 point score with a 45% probability of success (I left out PATs for simplicity). These both have the same expected value of 2.7 points. So are these two alternatives equal? Long term, yes. If you only have one shot, no.

    In our case the expected values aren’t the same. This changes things but I’m not enough of a statistician to know exactly how. But I still feel like at the end of the game points is better than no points.

  • RWard

    Im a Red Raider and I support Leach, if he kicks I’m disappointed. I don’t support the play call, our backs, specifically Baron Batch had been running the ball powerfully all second half. A play action pass or some other mis direction would have been better call.

    More importany the exchange of poession was a bigger deal, we pickoff Keenum but get stopped again. The missed 4th down was too early to be the game changer

  • I R A Darth Aggie

    1. Yes, you go for it. But I would call a play that would be…unexpected down on the goal line. Or throw a fade to the corner if you have a tall receiver to go up and get it. And you’re right, FG’s are not automatic any more than PATs are. And a blocked attempt could be picked up and scored with. ESPN’s stats show that they where getting an average of 5.1 yds/carry, and 6.8 per pass so getting 1 yard shouldn’t be that hard.

    2. You let the clock run and trust in your defense to get the job done. Otherwise you have to hope that you hit a home run with the precious few seconds you’ll have left in any case.

  • James

    Well, I always have trouble analyzing coaches decision-making in late game situations because they know their players more than I do. Maybe Leach was worried about his defense’s ability to stop Houston in a late game situation? There is a trust issue there that cannot be over-looked.

    Psychologically, Leach always believes he can get the points so the decision was probably very simple to him. It was 7 vs. 3 in his mind. That also would explain his willingness to admit fault in the press conference- it was the wrong decision because it did not work; not necessarily wrong based upon his thought process at the time.

    Personally, I would have taken the three off of what I saw from Tech’s defense that night. They were pretty good, not great against a great offense.But maybe Leach knew his defense would fold?

    Anyways, it appears he has more significant issues on his roster.

    http://ncaafootball.fanhouse.com/2009/09/27/texas-tech-has-a-twitter-problem/

  • Nole55

    I stupidly left out how much time was left when trying to analyze the situation. 10:56 is a long time, so I can see why Leach wouldn’t feel safe with an 8 point lead. To be honest, the more I think about it, I would trust Mike Leach going for 4th and 5 more than him going for 4th and 1, because his play callin is more tailored to it (crazy as it sounds). I say this having done 0 homework on Leach’s short yardage plays or conversion success, but it just seems like at 4th and 1, coaches eliminate 90% of the playbook, where as 4th and 5, he’d be more open. What do you guys think?

    As for Hope, Jimmy Clausen was having his worse day of season to that point, Michael Floyd wasn’t playing, and Charlie Weiss is just a bad coach in general. Versus Florida, ya I think you pretty much assume that they will score. Versus Notre Dame, I mean, givin that they werent exactly setting the scoreboard on fire, it would seem logical to just let it ride out. Also, I didn’t watch the game, but was Charlie doing the universal sign for spiking the ball, that little hand motion thing? With the exception of Carrol in the USC-ND game from a while back, thats usually a rock solid indication that a spike was coming.

  • Ed

    I think Brad above made the most coherent statement regarding Leach’s decision, which is that limiting the variance in the game outcome is more important than maximizing your expected point outcome. Take an extreme example: down one, ball on the opponent’s one foot line, with :01 left in the fourth quarter. Expected point value says go for a touchdown. We take it to be common sense that kicking a field goal is the better choice, but it helps our decision making to realize that while the decision is black-and-white that late in the game, the algorithm used (which your brain uses completely naturally, by the way) can be applied over grey areas as well, especially if you take some time to formalize it a bit. And, just as importantly, this algorithm is not the same as simply computing the expected point value.

    As the game winds down, additional points past a certain number become less useful to the team that’s ahead, until at the conclusion of the game, every point you’ve scored over and above a margin of victory of +1 is really useless (unless you’re Jeff Sagarin). You could refer to it as Marginal Utility of Points. So the question for Leach should not be, “What is the expected return on my decision in points,” but rather, “What is the expected return in wins in this game?” The results of these two different decision-making algorithms will often diverge as you get later in the game. You’ve stated as much in the post. The calculation should be to determine the utility of being up 8 with 11 minutes left versus being up 12, while considering the probability of success in either play call. You might find that the eventual decision is the same, but the thought process is different.

  • http://residualprolixity.blogspot.com NewsToTom

    1. While I’m normally a “go for it” proponent, I thought at the time and still think Leach made the wrong call because of the game situation. A made FG forces them to score a TD and a 2-pt conversion just to tie. Down 31-29, looking at a 2-pt conversion, what’s Houston’s chance of winning? Obviously, it depends on how much time was remaining when the TD was scored, but at the time they actually did score the winning TD, I’m guessing 30%, based on 60% chance converting, OT 50-50. If they left more time on the clock for TTU to drive, then it’s probably even less. Realistically, at 31-23, UH needs two scores to be likely to win the game-just like they’d need 2 scores to win if Texas Tech goes for it and converts.

    2. I can’t divorce my analysis of this case from my “domain knowledge” that ND was (stupidly) planning to spike the ball on 3rd down. Purdue’s D did seem at least a little discombobulated, and the logic of the call makes sense, but calling the TO reflected a lack of situational awareness on the part of Hope and his staff.

  • dr

    I like to think that the ticking clock in scenario 2 is one of the best defenses that a team can have. The offense is in a rush, the coaches are farther from the players than at any other point on the field, and the home crowd should be making their lives hell. I think all of this compounded with nerves will probably earn the offense a false start penalty. Even without a penalty, I like the idea of not letting the offense gather themselves and forcing them to finish the game in two plays with the clock running.

  • stan

    1. Go for the TD. Borrow from Homer Smith’s formulation on the two pt conversion — you only take a shot at the lower payoff (two pts) when you can accurately estimate the likely possessions remaining and what the losing team has to do. I.e. you have to be late in the game and you have to know that the one pt conversion won’t be needed. How does that inform Leach’s decision? Simple — With that much time left it was impossible to know what combination of points would be needed. Far too many drives and point combinations were possible so the smart play was to take the choice which offers the highest expected payoff. He had no way of knowing whether the “sure” three points would be decisive. With all that time, Houston could easily have put up a TD and a FG and everyone would be moaning about going up by only 8 when he could have been up 12.

    A caveat about expected payoffs — national averages can be extremely misleading. The payoffs can vary greatly depending on the quality of the offenses and defenses on both sides and the actual odds of making the particular fourth down play.

    2. As for Purdue — the first point is that Weiss was nuts (again, see Homer Smith re: spiking). I have no problem with the TO saving time. He could have had 25 secs to get in FG range. That’s a possible 3 or 4 plays with a timeout. Plenty enough to get 40 yards and a kick.

  • PaulsHealthBlog.com

    Leach’s call was a good one. It is asking a lot for a team to drive the length of the field and score. Too many chances for mistakes, errors, etc.

    If anything, put the loss on the Texas Tech defense.

  • Nate

    I think Mrphy is right that the expected values of the decisions aren’t the only important consideration, but not for the reasons he stated. This isn’t a sample size thing. It has to do with the relationship between scoring points and winning games. The goal of a football game is not to score as many points as you can, it’s just to score more points than your opponent. In some situations expected value (points) is an acceptable measuring stick for expected victory (win/lose), but in others it’s not. Early in the game, a decision that (over time) will score the most points is almost always going to be the decision that will win you the most games. Late in the game that may not be the case, because points come in packets of certain sizes. At the very end of the game, points can have wildly different values. An obvious example is if you are down by 4 with the ball with 1 second left. Even if you are close enough to kick a field goal with a high chance of success, you’ll still throw to the endzone because points 1 through 3 are worthless, whereas points 4 and 5 are enormously valuable. Kicking the field goal may maximize your expected points, but it gives you no chance of winning the game. Or in the Texas Tech-Houston game, if the score were 29-23 instead of 28-23, a field goal starts to look a lot better (success results in a 2-score lead), even though expected value considerations haven’t changed.

    So with all that in mind, did Leach make the right call? Hard to say. He maximized his expected points with a significant amount of time left (10 minutes) in a game that was likely to have multiple additional scores. From the expected victory point-of-view, the fourth point he would have gained (i.e., one more than a field goal) was extremely valuable, but the fifth through seventh were somewhat devalued.

    If going for it was a mistake, I don’t see how it can be that big of a mistake. He gave himself more than a 50-50 shot at icing the game, and even when that didn’t work out Tech was probably still a favorite. That means he probably gave himself something like an 85% chance of winning the game (If the chances of converting the TD OR winning when you failed were both 50%, Tech would have had a 75% shot. It seems like both were probably higher than that, so that’s why I’m throwing 85 out there.) Is it possible that the field goal would have yielded a chance much better than that? I don’t see how.

  • Kevin

    Obviously an interesting topic. Can we have some more scenarios to discuss please..

  • Mrfphy

    So far I’ve argued simply that the expected value model is not necessarily an accurate compass for which to make the field goal vs. touchdown decision. This is because the probability of scoring is more important than the overall outcome in points, especially late in the game. This leads to the argument stated many times above by others that the important thing is to outscore your opponent and win the game, not score the maximum number of points. There’s obviously some fuzzy gray area in the second half where the strategy changes from scoring a maximum number of points as a way of beating an opponents who will score an unknown number of points to assessing the current scores and making a decision based on expected outcomes for the opponent. Given this knowledge, here’s a scenario:

    A defensive battle is playing out in the form of a scoreless tie late in the game. Both teams are known for having bad offenses averaging no more than two touchdowns a game. (Late is arbitrary here because I won’t be assigning exact probabilities to scoring chances) Your opponent has a great track record of creating turnovers and has yet to produce one in today’s game. Your team is driving down field and has reached the point where the probability of a successful field goal attempt has plateaued. (I’m betting that if you graphed field goal tries you’d find that the difference between a 15 and a 25 yard try aren’t very different) It’s 3rd and long…an obvious passing down. Given the argument that you’re not trying to maximize points you are simply trying to outscore your opponent, are you better off kicking the field goal on third down!?!?

    I don’t find it a big stretch to believe that the probability of winning the game with a field goal could be higher than the combined probabilities of all the things that could go wrong (including a turnover on third down that would leave you scoreless). Yet a coach would be killed for doing this. Where’s the hole in the logic? Or do the number of unknowns and variability of in game probabilities prove that there’s no best answer and that it’s best to go with your gut?

  • Jeff

    Leach absolutely should have gone for it on 4th and Goal at the 1. He was a yard away from a likely game-sealing score. There were three far bigger errors the Red Raiders committed that gave the game away:

    1) Abandoning the run. The Red Raiders averaged 5.5 yards per carry, but only rushed 6 times in the fourth quarter. Everyone knows running the ball drains the clock, and Tech really had no reason to start flinging the ball all over the place.

    2) Not allowing the Cougars to score sooner. I’m probably in the vast minority here, but a year ago against Texas, I was screaming at the TV for Tech to let Texas score (which they were obviously going to do), so Tech would have more time left to conduct a final drive (which they happened to do, in case you’d forgotten). It quickly became clear Houston was driving to score, with some 32,000 delirious fans cheering them on, and Tech could do little to stop them. Houston strolled down the field, crossed the goal line and kicked off to the Red Raiders, giving them less than a minute to drive most of the field. This wouldn’t have been catastrophic, except for

    3) Tech failing to utilize the sidelines to their advantage. The only possible reason I can think of for Tech not throwing to the sidelines every time (so their receivers could get out quickly and stop the clock) is that getting out of bounds only stops the clock temporarily (as when a first down is made) at the college level. If this isn’t the case, why would you ignore one of the most basic end-of-game strategies when you have less than a minute to score and no timeouts left?

    All of these decisions were far greater factors than the fourth-down try. Passing up the field goal is simply the “Bill Buckner Red Herring” – the obvious excuse everyone points to, but ultimately isn’t a valid excuse for the outcome of the game.

  • http://www.hailtopurple.com/jhodges/ JHodges

    Regarding scenario 1:

    Someone else brought this up, but if the ball was on or close to the hash and at the one yard line, that is a very tough angle, especially if it’s on the opposite side of the field as the kicker’s kicking foot. I’ve seen multiple solid kickers miss such “extreme angle” short kicks.

    And if they did indeed miss the FG, Houston would get the ball at the 20 yard line instead of at the one.

    This fact basically reduces the expected value of the points when going for the FG.

    And then there’s the old football mantra of “if you can’t convert a fourth and one by running the football, you don’t deserve to win the game.”

    I’d go for the TD just based on that, but adding in the fact that they were on the road and had a chance to seal the game just adds more to the argument.

  • Paul

    I still think that Hope made a bad decision to call that timeout. The probability might indicate that Notre Dame is going to score in that situation, so I understand the reasoning, but what is the probability of Purdue getting into field goal range with only 30 seconds left?

    If Hope doesn’t burn that timeout, and ND has once chance to score from inside the five yard line, the odds are good that the Irish score. How much does the probability increase when they have two chances? And, more importantly, how does that increase in probability compare to Purdue’s chances of getting into field goal range with almost no time left?

    Even though ND is most likely going to score in that situation, I think stopping them once at the five yard line might still be more likely than Purdue getting into field goal range with 30 seconds left. It seems to me that Hope’s best course of action would have been to do everything possible to stop ND from scoring (i.e., not calling that timeout), instead of banking on his team’s chances of getting into field goal range with such little time left.

  • Kevin

    I agree with Paul here. You are going to have to toss it up deep once, maybe twice and hope it lands in your own teams hands and hope it’s far enough to kick. This time out gives the opponent an ADDITIONAL play to score. Period.
    The exeception for me to call time out with that time on the clock is if the defense was in misalignment.
    I’d rather let the defense win it in one play defending our endzone. Statistics or not that’s my call (and I teach high school math)….

  • Bobby

    Paul and Kevin, your conclusions relies on the fact that ND gets an extra play. But that is only true if you KNOW, ahead of time, that Weis will (stupidly and unnecesarily) spike it. Jokes aside, you can’t know that. Hence Hope’s choice was between some time and no time, and he chose some. His D had to make two stops regardless.

  • Linus

    I don’t have the box score in front of me but I’d be shocked if 10+ minutes remaining equaled 5 or 6 drives. That’s less than 2 minutes per drive. Pretty optimistic, even for excellent offensive teams.

    My decision if I’m Leach hinges on how much confidence I have in my offense. I’m not talking about the confidence on 4th and 1, because with such a small sample size, luck has too large an effect. I mean, if I kick, and am up 8, do I really believe that, in 10 minutes Houston is going to score twice, and we’re not going to score even once?

    I would think, with the offense I’ve got, it’s more likely that, even if Houston scores 10 in 10 minutes, we’re going to score at least once and win. And the PRESSURE is on Houston, because they have to score TWICE in 10 minutes.

  • Mark

    Bobby nailed it. From Purdue’s coachign perspective, they had to stop two plays. They wanted time left after the two plays so they could try to do something IF Notre Dame scored. There is no way they could have known that Charlie Weis and his staff were going to run around begging their upperclassman quarterback to spike the ball at the 2 yard line on third down with over 35 seconds left. Of course it makes the coaches look bad AFTER the fact, but during the game, it is ABSOLUTELY the right call. Purdue’s kicker kicked a 60 yard field goal in the first week, by the way. Getting 40 yards, in 24 seconds, with 2 timeouts, is not impossible. Although taking the sack after the first play pretty much gutted those chances.

  • Ben

    The thing about all these analyses is this: in the end, it all comes down to how much risk the coach is willing to take. That is the ultimate truth about any decision theory. It doesn’t give you the single right answer; it just helps you understand why you choose A instead of B. Maximizing points is generally preferred (not always!) up until the last part of the game (“last part” itself being defined by the coach as a form of risk). As the game progresses, it can become more valuable for either team to guarantee points depending on the situation. Ultimately what the coaches are praised or shunned for is their determination of risk, especially since each of those decisions carries with it some form lack of (statistical) confidence in the team – which we fans would never express in those situations ;-)

  • http://www.onthemoon.webs.com vonpish

    Your blog is very nice…i’m sure it can inspire other teams…+.-

  • Mike

    I agree Mike Leach made the right decision. I don’t think that the decision is the same for every coach and every team. Leach’s team is a strong offensive club that is built around being aggressive. That is just Tech’s style of play and what the players are recruited to do. However, I don’t like his play call. More than anything he needs to keep his team within it’s comfort zone at such a critical point in the game. He needs to have a few plays on the back of his play chart specifically designed for “4th and 1 in the clutch”

    The decision itself should always be based on your team’s style of play and the flow of the game. Jim Tressel would have taken the field goal and felt good. He knows that his well-coached special teams unit will cover the kick well and not give his opponent great field position. A coach like Pete Carroll or Nick Saban should feel more confident in their defense than Leach.

    As for the play call, if you do go for it you have to play to your team’s strengths or at least call a play that your team practices often and runs well. Every coach should have situations like this planned ahead of time. The plays should of course be plays that are more indicative of Leach’s style (passing) and perhaps some sort of tendency-breaker (direct snap or draw)

    The situation reminds me of the Ohio State at Purdue in 2002 (back when Tressel seemed like a wizard on the sideline) With the Buckeyes trailing 6-3, Tressel was faced with a 4th down and long 1 from the Purdue 37 with 1:45 remaining. Tressel’s conservative tendency and strong running team would lead one to believe he would try to power ahead for the first down. Purdue counted on this, bringing both outside linebackers on a run blitz. Tressel left both backs in to max protect, and Krenzel took advantage of single coverage and hit a deep strike to Michael Jenkins for the Score. A coach broke his tendency and made a great call of a play his team had obviously practiced (GREAT blitz pickup) and because of that, his team got the job done.

  • Darrell

    If I am bypassing the FG in scenario one, I’m running a play action bootleg with the fullback in the flat & the QB has the run/pass option. In scenario #2 I would have let the clock run and made the other team beat me….

  • Kevin

    I can concede the 2 play scenario. Timeout gives them 2, or they could run two. However, I’m not giving their OC more time to think or their offense time to line up and set up their best play so I can keep 15-20 seconds. My defense can line up and have it’s call in 2 seconds. Stop or not, we win we lose.

  • Brad in KY

    Example of going for it on fourth and short:

    This past Sunday, the Patriots converted a 4th and one from about their own 28 yard line with the lead in the second half against the Falcons. Gutsy.

  • Brad

    @Linus

    That is exactly my thinking. I took a look at the box score and Houston was taking about 4 minutes to score a TD drive. Plus TT could take steps to bleed the clock. I think you are probably looking at 2 more drives each maybe a 3rd for Houston depending on how much time goes off the clock. That is not a lot of drives to average out an expected value.

    Houston had only scored TD’s 1/5 of thier drives and had missed 2 field goals. I think the odds of them scoring twice without Tech scoring was pretty low. Tech had moved the ball well when the didn’t fumble.

  • AndrewC

    I was at this game, and originally thought that Leach made a bad decision. Then after thinking about it, I agreed with Leach on going for it, but not necessarily the playcall- thought they should have done something besides the sneak. Now I think that the sneak was probably the right move as well.
    It basically came down to answering 2 questions “is it reasonable to think that Houston will have multiple chances to score again?” and “Do I have a high probability of success if I go for it on 4th down?” Both questions can be answered with a “yes”.

    Though Houston’s hadn’t scored an outrageous amount of points in the first 3 quarters, they had moved the ball very effectively. The Tech defense was getting tired as well. I would say that 8 points would have held if there were only 5-6 minutes left in the game, but not 10+.

    And even though you wouldn’t expect the Tech line to be good at QB sneaks, their line dwarfed the UH defensive line from a size perspective, and they had been effective at running against UH all night. Add to that the fact that Tech ran a successful QB sneak vs Texas the previous weekend, and Leach had all the evidence he needed to assume that a QB sneak would be successful.

  • Linus

    Right, Brad. Look at the possible scenarios (assuming Tech makes the field goal). 1) Houston drives down and scores a TD, makes 2-pt conversion; 2) Houston drives down and scores a TD, misses 2-pt conversion; 3) Houston drives down and kicks FG, expecting to get the ball back (say they only used 2 mins, so 8 mins remaining); 4) Houston’s drive is stopped, either on 3rd down or 4th down. Now, of those 4, which one leaves Tech behind? None of them. And 3 of them leave them in the lead, with the 4th one being statistically less probable than the others.

    AND, all of them most likely leave Tech with another possession (or two) to score. As I said, if I have confidence in my offense, I’m thinking that, given two possession, we’re GOING to score. I understand the point made above about having confidence in them on the 4th and 1, but like I said, one play is just too small a sample to not have luck unduly influence the outcome.

    Question on the Purdue-ND scenario: would those who think Purdue goofed still think so if Purdue had driven down the field and kicked a field goal? In other words, do we think it was a bad call just because it didn’t work?

  • OldSouth

    Why is anyone concerned about sample size in estimating the EV of a field goal with respect to Tech’s decision? If your EV for going for it is ever >3, your decision is made.

    I also think experience and knowledge of your kicker and previous kickers is an adequate substitute for sample size in any field goal situation. Few if any kickers will ever kick enough field goals in a season to get a sufficient sample size for any test of statistical significance. n is lucky to break 20 for some teams.

  • Rob

    3rd scenario:

    You are behind 23-7 in the 4th. Your team scores a TD with 8:41 to go. Do you kick the xp or go for 2?

    Follow up question – if you are the AD and your coach kicks the xp, do you fire him on the spot and not even let him finish coaching the game or wait until after the game?

  • Eduardo

    I don´t think that being in early 4th quarter alter the result of the expected value. You have plenty of time for another posession and the other team have a great chance of 2 posessions. If they are playing well, an 8 point lead does not end the game.

    I think that the “play safe” option would be totally right only with 5 minutes or less, depending on the timeout situation.

    And, if you doomed to lose, it´s better losing being aggressive. You have to do everything to win not for “don´t lose” the game. In longterm, you will be winning more than losing play this way.

  • Eduardo

    OldSouth,

    The problem is that, if you are in a early part of the game, you can reason only with the expected value, but, if you are late, maybe you can make a better option choosing an option with less EV but less variance either (the FG), because you don´t need 4 for points (wich is your prime objective), you need only 3.

    To answer the question, you need to make an Monte Carlo Simulation and watch how many times you team would end winning or loosing in each situation. The problem is how you will do these simulations. Probably using data from the same match you are analising, or other matches with same situation. It´s not easy.

    All these can help you to choose and defend your choice, but the fact is that it will be a close call. And you will be second-guessed if it goes wrong.

    Greetings,

    Eduardo

  • Exiled_in_VT

    On Leach’s decision, it reminds me a bit of baseball analysis of bunting.

    It has been shown that bunting always reduces the number of runs you expect to score that inning. Once this was discovered, a whole lot of stat heads said you should thus never ever bunt under any circumstance.

    Then, someone pointed out that in many situations (late in the game, down 1 run or tied) you are not trying to maximize the number of runs you score in an inning, you are trying to maximize the probability of scoring at least 1 run in that inning.

    It’s not an exact analog, but it shows in order to evaluate the effectiveness of a decision, you first need to determine what objective the decision-maker has.

  • Exiled_in_VT

    I forgot to point out that in certain cases bunting does improve the probability that a run will be scored in an inning (especially if it’s in the National League and your pitcher is up to bat).

  • http://www.powerlineblog.com/archives/2009/09/024563.php Otis

    Scenario 1 reminded me of a Powerlineblog.com post I read on a Redskins’ win. The gist is, that if you kick the FG you are putting the game in the hands of your defense alone to stop the opponent (albeit a better position with an 8 point rather than a 5 point lead). By going for it you are affording yourself two ways to win. Your offense can win by punching it in, and your defense can win by stopping a 99 yard drive. And under the two ways to win option, your D is put in a great position b/c the opponent’s offense is backed up. The likelihood that you get a stop and get the ball back, close to FG range, is pretty good. I think it’s a no-brainer. In a “track meet” game you go for the points.

  • KungFuPanda9

    Kick the field goal, get three points, make it an eight point lead. Heck, even if you consider your math probability of 2.5 it’s still insurmountble considering your opponenet has to score a touchdown and two point conversion. The conversion alone makes the eight point lead even larger because the odds of making that play is considerably higher than the PAT.

    As for Purdue and the ill-timed time out. Bad call. Let the clock be your friend, unless your defense is so pitiful they aren’t capable of making two stops.

    Working the numbers and stating the probabilities are theoretically interesting. But reality has made it plain that both of these calls were incorrect. In fact, over time, they would be incorrect more times than not.

  • Patrick

    As other people have noted, as the end of regulation nears, the expected value of points (EVP) and expected win % (EW) diverge. In theory (excluding impressing polls), expected win total should be the deciding factor in all coaching decisions.

    In order to make the optimal decision, Coach Leach would need to weigh the EW of and respective probabilities of the following:
    A. Go for it and score touchdown on 4th. Up 12-Houston gets ball at approximately their own 30 yard line)
    B. Go for it and fail to score touchdown. Up 5-Houston gets ball at approximately their own 1 yard line)
    C. Successful FG. Up 8-Houston gets ball at approximately their own 30 yard line)
    D. Unsuccessful FG. Up 5- Houston gets ball at the 20 yard line.

    If he feels that P(A)*EW(A)+P(B)*EW(B)>P(C)*EW(C)+P(D)*EW(D) then he should go for it; if he feels otherwise, he should kick. According to advanced NFL stats, P(A)=.68, therefore P(B)=.32 I can’t find anything about FG probability, but assuming the Tech Kicker makes the kick 95% of the time, we get

    .68EW(A)+.32EW(B) vs .95EW(C)+.05EW(D)

    The EWs are practically impossible to predict, however it stands to reason that:
    1. EW(A)>EW(B)>EW(D)
    2. EW(A)>EW(C)>EW(D)
    3. EW(B) vs EW(C) is ambiguous as one could argue the 29 extra yards of field position and the chance to record a safety help Texas Tech as much as the field goal.

    If one feels that EW(B) is greater than OR EQUAL TO EW(C), then going for it was NECCESSARILY the right decision.
    -setting EW(B)=EW(C), we get the equation .68A+.32B vs .63C+.32B+.05D
    -remove .32(B) from each side
    .68A vs .63C +.05D
    .63A+.05A vs .63C+.05D
    .63A>.63C; .05A>.05D therefore .63A+.05A>.63C+.05D

    In fact there would need to be a substantial difference between EW(B) and EW(C) in order for kicking the FG to be worth it. For example, if
    EW(A)=95%
    EW(B)=70%
    EW(C)=88%
    EW(D)=65%
    going for it (EW%=87) is still superior to kicking the FG (86.85). That’s a pretty optimistic EW for kicking the FG (or pessimistic view of going for it), yet rolling the dice is still (marginally) superior. Playing around with the numbers can yield interesting results.

    http://spreadsheets.google.com/ccc?key=0AkPHyEHn9kQSdE9VTXRsMkNMYnFVZGJmVF8tRjVreFE&hl=en