The comeback, done the right way

Submitted by M Squared on

The notion of the comeback is distinctively unfamiliar to a program that is first in all-time wins and winning percentage.  But as we continue to rebuild in order to regain that top status, we will find ourselves, unfortunately, in the comeback position in many games this coming year.  A lot of comebacks require a 2 point conversion along the way, and that in turn requires some decision making.  Various interesting decisions arise as to 2 point conversions.  I want to address a specific kind here – one that, in my opinion, just about every football coach gets wrong. 

 

The Setting: It is the first game of Coach Rod’s era at Michigan.  There are about 9 minutes left in the game, Michigan is down 25 – 10 against the Utes but marching down the field with surprisingly crisp execution.  Then, Steven Threet hits Junior Hemingway in the corner for a 33 yd touchdown.  Michigan has cut the 15 point deficit to 9.  The score is 25 – 16.  What now?  Before the touchdown, everyone’s thinking we need a 7 point conversion, an 8 point conversion, and the defense not to give up any more points.  So far, so good.  But does the order matter?  Yes!  Not only does it matter, almost every coach chooses the wrong order!  In his first game, Coach Rod also chose the wrong order, and went for the 7 point conversion before the 8 point conversion.  Ultimately, it didn’t matter in that game, but that doesn’t mean it won’t matter in other games. 

 

The Principle: It’s better to have a small chance to win than no chance to win.  The 8 point conversion attempt must come first. 

 

In Coach Rod’s first game, Michigan happened to score the second touchdown with 6:26 left in the game.  So, Michigan had ample time to overcome Coach Rod’s mistake.  As noted, choosing the right order in that game would not have made a difference.  But suppose instead, that Michigan scored the second touchdown with 26 seconds left in the game.  Now, if Michigan fails to convert the 2 points, Michigan is out of luck.  To recap, MI scored TD + the kick after with 9 mins left.  Then, MI scored w/ 26 seconds left, failed the 2 point conversion, game over.  

 

Switch up the order.  MI scored with 9 mins left, but now goes for the 2 point conversion right away and fails.  The score is 25-16 with 9 mins left.  The deficit is substantial but not insurmountable.  To compare, 25-23 with 26 seconds left is much, much more substantial (i.e., almost no chance to win).  It is better to have a small chance to win than no chance to win.   

 

Essentially, in any comeback where the coach believes that a 2 point conversion is necessary to win the game, it is better to attempt it earlier than later.  Put another way, the likelihood of converting the 7 point conversion and the 8 point conversion is the same in the two scenarios.  It is more probable than not the 2 point conversion will fail.  It is better to give the team more time to address that likely failure than less time. 

 

I can think of two major objections to my position.  First, attempting the conversion earlier and failing deflates the team’s momentum.  Second, delaying the conversion is better because maybe the team won’t have to make that decision (i.e., fortuitous events such as a pick 6 make it unnecessary). 

 

This post seems ridiculously long already, so I will stop here instead of diving into further analysis.  I will note that I have thought about those two objections, and, one, I do not think they outweigh the force of my argument, and two, I do not think they are the rationale of coaches when they select, what I believe, is the wrong order. 

 

 

Comments

Blue in Yarmouth

August 25th, 2009 at 8:36 AM ^

to say that coaches who know far more than I about football are making the wrong decision, but I agree with your rationale. I have thought similarly in those situations as well but I would have to hear what a real coach thinks before I said he was wrong. Also, I never heard of a 7 or 8 point conversion before........Just sayin'

M Squared

August 25th, 2009 at 9:28 AM ^

...let me assure you that I know that I don't possess half of the football knowledge in Rich Rod's pinky fingernail.

Football coaches are experts in football, but that does not make them experts in math or game theory. I don't claim to be an expert in those fields, but given that football coaches aren't either, it's interesting to discuss whether they've got it right after all.

In recent years, many football coaches have started to use a 2 point conversion chart. That chart was not made by another football coach, but someone outside football. So, discussions outside football can inform discussions inside football.

Or consider, for example, the attempt to go for a first down on fourth down. Is it me, or do Pete Carroll, Charlie Weis, and Urban Meyer seem to go for significantly more of these than any pre-2000 college coaches? It sure seems like it, but I could be wrong. I take no position on this issue, but clearly those 3 coaches think that the preceding era of coaches had it wrong (if in fact these 3 do go for it more often).

BiSB

August 25th, 2009 at 8:50 AM ^

But I think typical coaching strategy says to take the 7 on the first go-around, and go for two on the next score. 99.9% of coaches would have done the same thing, including Lloyd (not to mention my high school coach).

The one additional reason you didn't mention is that you never know what's going to happen later in the game, especially with nine minutes left. There are a number of ways the score can change where is won't just come down to the binary choice you indicated. Just look at the Wisconsin game: Michigan went for two with 10 minutes left to go up 3. It seemed like the right choice at the time, but in hindsight, had they taken the one point, Wisconsin would have needed the Woody Hayes-esque "three point conversion" at the end to have a chance.

You take the guaranteed points when you can, and take the risks only when you need to.

That, and your objection seems a little late. We're all into the "acceptance" stage of last season, and you're dragging us back into "anger" or "depression"

Mgoscottie

August 25th, 2009 at 10:10 AM ^

I think if you're down by 8 or 15 with 9 minutes, that scoring three touchdowns and losing one of the extra points isn't really a big concern.
With that in mind, if you assume that you will miss the two point conversion, I think it would be better to do so on the first touchdown than the potential game tying touchdown.

steve sharik

August 25th, 2009 at 8:51 AM ^

...because two point conversions at different points of the game can either be good or bad decisions. There is no right time or wrong time to go for two in the situations you bring up. Yes, there is more pressure on a team to make it at the end of the game, but is there any more pressure then as compared to when you have to go for it after a 2nd OT? I don't believe so.

I think the best way to make those decisions is to go with your instinct as to when you think your team can make it. If you're on a roll and the other team is reeling as hasn't been able to stop you lately, go for two (if you need the points). If, however, you get some kind of a gift (like a defensive TD) and are trying to climb back into the game, why kill your momentum by going for two with a stalled out offense?

The reason these "when to go for two" arguments are never settled is because there is no right answer.

M Squared

August 25th, 2009 at 9:14 AM ^

...situational setting has to play into the decision making - absolutely. If the reason for the TD is that a crucial player on the opponent's defense has gone down w/ an injury, then I agree, stick w/ the kick after because the offense may likely just march down the next possession and again after that.

Given a large enough sample size, however, you would think that this kind of stuff would even out (i.e., it's just as likely that an offensive player on our team would go down w/ an injury). So w/ all things equal, it is surprising that coaches do not try the 2 point conversion earlier and deal w/ the failure of the attempt earlier rather than later.

jg2112

August 25th, 2009 at 9:34 AM ^

You wrote:

The notion of the comeback is distinctively unfamiliar to a program that is first in all-time wins and winning percentage. But as we continue to rebuild in order to regain that top status, we will find ourselves, unfortunately, in the comeback position in many games this coming year.

I respond:

Michigan, during the last few years of the Lloyd Carr era, was more likely to win a game that it trailed going into the fourth quarter, than to win a game it narrowly led going into the fourth quarter.

Michigan has ample experience coming back to win games. Just off the top of my head:

Ohio State 1988 (Kolesar catch).
Bowl game v. Alabama (Kolesar catch).
vs Iowa - 1997.
vs. the Lolphers, 2003 and 2004.
vs. Virginia, 1995.
vs. Illinois and Michigan State, 2007.

And, uh, Wisconsin 2008?

MGrad

August 25th, 2009 at 9:35 AM ^

You kick it first to make it 8, and keep the team focused on one score to force OT. Basically, you inspire both sides of the ball to just "make A play", getting an emotional lift.

If you miss the two point conversion, and it's now 9, you need 2 scores, which is much tougher in 9 minutes against a team that had 350 yards and 36 minutes of possession in the game.

El Jeffe

August 25th, 2009 at 9:46 AM ^

The way the game actually played out, your point makes sense (though if UM had picked one off and returned it for a score, making the game 25-23 with 5 minutes left, things would be quite different, as suggested by Blue in South Bend below).

Look at it this way: the 2-point conversion is a high-variance play. This is because the formula for the variance of a Bernoulli random variable (1 if success, 0 if failure) is P(1 -P), where P = the probability of success. In the NFL at least, the conversion rate is about 50%, so let's say it's that in college, just for argument's sake. Thus, the variance = 0.25.

Kicking Competency Lopata was 27-29 on PATs for the year, so let's assume that probability would have held in game 1. 27/29 = 0.931. So the variance of the PAT = 0.931(.069) = .064, obviously much lower than the 2-pointer.

I believe that it is better to take the lower-variance (or "sure points") and live to fight another day, rather than needlessly exposing yourself to the risk of getting behind the eight ball (there may be a mixed metaphor in there somewhere).

What if RR had gone for 2 at first, missed it, and then UM had gotten a safety? Then UM would have been down 25-18. If they had scored, they would have had to kick the extra point just to tie, instead of to win (i.e., if they had kicked the PAT on the first TD, then a safety would have made it 25-19).

There are dozens of scenarios you could play out like that, all of which, I think, suggest that in advance of knowing how the last 9 minutes will unfold, it is better to go low variance first, and save high variance for when you really need it.

M Squared

August 25th, 2009 at 10:03 AM ^

...I DID address in my post the very point that you raise as to fortuitous subsequent events. (It's in the second to last paragraph.)

Yeah, you might be right. It is interesting to me that most people's instinct is to go for the 7 points first. The reason I felt otherwise is that those fortuitous subsequent events tend to be really unlikely (e.g., pick 6 or safety, as you note, as well as other possibilities).

You wrote, "I believe that it is better to take the lower-variance (or "sure points") and live to fight another day, rather than needlessly exposing yourself to the risk of getting behind the eight ball (there may be a mixed metaphor in there somewhere)."

I probably wrote poorly, but I was trying to address this specific issue. The "sure points" are present in either scenario. I would disagree as to "needlessly exposing yourself" to risk part of your statement. The 2 point conversion must be attempted, barring some event such as a pick 6 or a safety, etc. Maybe our disagreement can be resolved after all by an empirical analysis of how often a team needing a 2 point conversion gets a pick 6 or safety, thereby obviating the 2 point attempt later.

My guess would be that this occurs less than 10 % of the time. IF that is true (I really don't know if it is), it's better to have a chance to remedy the 2 point failure than have no chance at all when it fails w/ no time left. Anyway, it appears that the weight of the opinion is against my argument. I can respect that.

El Jeffe

August 25th, 2009 at 10:34 AM ^

See? A civil discussion. About football, even. Nice.

I think I (and you) are focusing too much on wacky scenarios like The Machete's pick 6 against Wisconsin, or a safety, or virtually any shotgun snap to Jimmah in the 2007 ND game. I take part of the blame for that. Generally speaking, a lot of regular ol' football stuff can happen in 9 minutes, so I would rather take the sure points.

Which brings me to my second point: you say

The "sure points" are present in either scenario.

But that's not true. In your scenario, the sure points would only be there on the second TD if you made the first 2-point conversion. That is, if UM had gone for 2, missed it, and the game had played out exactly the same way, then it would have been 25-16, and even a TD and a 2-pointer would not be enough to tie.

So in my scenario, there are unconditionally sure points. In your scenario, the chance to have sure points is conditional upon a high variance, relatively low probability of success event happening first.

So, I think mathematically, unconditionally sure points > conditionally sure points.

M Squared

August 25th, 2009 at 11:18 AM ^

I find that people thinking or posting about Michigan football during the college football withdrawal period, like the month of March for instance, are far fewer than I would have imagined. Thus, it would be nice if we die-hards could get along...but it seems a lost cause on some of the other Michigan boards.

I understand what you're saying about the "sure points." I do. But you're just looking at it post facto. I was trying to say that when examined prior to the sequence, they would be considered "sure points" in both scenarios. So, a coach calculating his odds at that point in time would correctly apply the same percentages (e.g., likelihood of 2 pt conversion is 42%; likelihood of PAT is 98%). I'm probably explaining this quite poorly.

You are right that under your scenario there are unconditional sure points, but I'm arguing that you would still lose if the 2 pt conversion fails. Over a large sample size, your approach may actually result in more points because the team would have taken the sure points more often. You may be right about that; I'm not sure. But my argument is that ultimately, the team would still lose if the two point conversion is not successful. Thus, under your approach, there would be a higher incidence of closer losses but not wins.

In the end, the "more points" can't overcome the failure of the two point conversion, which is the same conversion rate in either scenario. Under my approach, at least this is my crazy argument, the team will have more chances to overcome the deficit with time on the clock.

wile_e8

August 25th, 2009 at 10:07 AM ^

There are dozens of scenarios you could play out like that, all of which, I think, suggest that in advance of knowing how the last 9 minutes will unfold, it is better to go low variance first, and save high variance for when you really need it.

I kinda agree with the OP, which is why this sounds backwards to me. I would think, with the dozens of scenarios that could play out in 9 minutes as opposed to 30 seconds, you would want more time to recover if the high-variance play doesn't go your way. If you wait until the very end of the game, you're essentially rolling the dice on a 50/50 event with a very poor chance of recovery if it doesn't work out.

It all comes down to personal preference I suppose, because there are examples supporting each case. But given the choice, I don't know why anyone would pick the one that has the most likely scenario being equally likely to fail as succeed with no options to recover after failure.

mgovictors23

August 25th, 2009 at 9:49 AM ^

I agree with what was said earlier, it's all up to the coaches when they want to do it. I think for the most part they do it when their instinct tells them to do it.

jwfsouthpaw

August 25th, 2009 at 11:18 AM ^

In college football, momentum is pivotal; I remember the announcers for the Wisconsin game remarking in the third quarter that the momentum had swung in Michigan's direction; seconds later, Minor broke his big touchdown run, and the famous INT happened seconds after that. Momentum swings games.

Now, let's take your Utah example. The team is down 25-10 in the fourth quarter and scores a touchdown. 25-16. What do you do?

If you convert the PAT, the team is within 8 points -- just one score. Your players are much more confident that you can erase that deficit, and perhaps just as importantly, the other team knows that their cushion has disappeared. Momentum is clearly with Michigan in this case. And should Michigan score another TD (as actually happened), I think that momentum has an even bigger chance of carrying over to the 2-point conversion (alas, this did not actually happen).

If you attempt the 2-point conversion and fail, you still need two scores to overcome the deficit, which is much more difficult. The players, of course, know this, and so does the other team. The other team can run its offense and defense without worrying that one mistake could lead to a tie. Here, the momentum might still be with Michigan, but not by nearly as much.

That's just my take. Your point is definitely valid, but I think the psychological factor is significant in making that decision.

chally

August 25th, 2009 at 11:41 AM ^

I can't say Rich Rod did it right or wrong because it depends on numbers I don't have.

The problem I have with your approach is that you assume a coach who is down by 8 will call the game the same as if he were down by 7 (i.e., using time freely and attempting to score late into the game to prevent the other team's comeback), rather than as if he were down by 9 (i.e., attempting to score as quickly as possible knowing that his team needs two scores).

If your assumption is right, so is your analysis (for the most part).

Say a two-point conversion is a 30% probability event. If the team goes for two first, 30% of the time they will be down by 7 and use conservative strategy A (which has some non-zero probability of success). 70% of the time, they will be down by 9 and use risky strategy B (which has some other non-zero probability of success).

If a team down by 8 employs strategy A just the same as if they were down by 7 30% of the time, the outcomes will be identical. The other 70% of the time, they must employ last-ditch strategy C (which will almost certainly have a lower probability of success than strategy B).

That's basically your argument. I object to these assumptions, because I believe that coaches are aware that a two-point conversion is a low-probability outcome. I am confident that a coach behind by 8 with nine minutes left is aware that he will likely need to score more than once, and is therefore calling the game with the intent to score as quickly and as often as possible.

For example, assume a coach takes the extra point to go down by 8 points with 9 minutes remaining. With 5 minutes remaining, his team drives the ball into safe field goal distance (say, to the 20), but stalls leaving a 4th and 3. I'm inclined to believe that almost any coach with a competent kicker would take the field goal there, just as if they were down 9.

If a coach down by 8 calls the game the same as if he were down by 9, then the logic plays out differently.

If you go for two first:
30% of the time you have an x% chance of tying.
70% of the time you have a y% chance of winning outright by scoring twice, but no discernable chance of tying.

If you go for one first:
z% of the time you will score a touchdown first. 30% of those occasions you will tie the game.
(1-(z*.7))% of the time you will need to score twice (either you score a non-TD first or you will fail the two-point conversion), thus, you have a y% chance of winning outright (still employing strategy B).

Since (1-(z*.7))% > 70%, the difference between the strategies is that when you go for 1 first, (1-(z*.7)-.7)% of the time, you have y% chance of winning rather than x% chance of tying.

Therefore, you should always go for one first where y > (x* % chance of winning in OT). The numbers here depend on the riskiness of your alternative strategies A & B.

In short, you may be right but (1) its probably for the wrong reasons, and (2) it is contingent upon data that we're basically speculating about. If I were a coach, I would hire someone to actually do a statistical analysis of games to tell me what to do. Maybe coaches do that, in which case they are probably more right than you.

StephenRKass

August 25th, 2009 at 11:48 AM ^

I didn't read the comments & posts fully, but scanned them and there appears to be something missing, namely consideration of the personnel available. I think that with the receivers and qbs available in 2009, RR makes a different decision & goes for the 2pt right away vs. Utah. Specifically, I would be much more willing to attempt at 2pt conversion with DR under center, or even DR & Tate side by side, along with 4 - 5 receivers. Hike to Forcier, then either pass immed. or lateral to DR & he passes or runs. The defenders are spread to cover all the receivers, giving DR more of a chance to juke and get in. With the qb situation in 2008, I can really see why RR would be hamstrung.

Jeff

August 25th, 2009 at 12:02 PM ^

When down by 15 points I don't think the order matters. If you miss the 2-point conversion then there is no way to win unless you score a field goal or a third touchdown.

However, coaches could improve their strategy when down by 14 points. In that case they should try a 2-point conversion first. If that is good, then they kick after the second touchdown and win. If it is not good, they go for 2 after the second and can still tie.

If you score 2 touchdowns and keep the other team from scoring then the probabilities work out like this (using 50% on 2-pointers and Lopata's 27/29 on kicks):

2-point strategy conventional
Win 47% 0%
Tie 25% 90%
Lose 28% 10%

Since a game can't end in a tie, the question becomes how badly do you want to avoid overtime? Using these stats the the conventional strategy is better only if you will win in OT more than 72% of the time.

Probably a 2-pt conversion is more like a 40% event. Under that assumption, the conventional strategy improves a lot but still requires better than a 57% likelihood of winning in OT.

Tim

August 25th, 2009 at 12:42 PM ^

Theoretically, if you're down by 15 points with enough time for at least 2 scores, my personal philosophy would be to go for the 2-pointer first, so you know whether you need 1 or two scores to tie it.

If you go for the 7 first and miss the 2-pointer the second time around, you're out of luck, as you're less likely to have been planning that you need a third scoring drive.

Saluki

August 25th, 2009 at 1:20 PM ^

...trumps all others in this situation.

RR needed to keep as much momentum as he could going. Regardless of the realities of "knowing what is needed" by going for 2 first, RR figured they would be doing well just to score again, and momentum would be key to making that happen. "Let's worry about getting 2 in the unlikely event that we can make it in the end zone again."

Wolverine 98284

August 25th, 2009 at 2:03 PM ^

It's a discussion right?

Momentum is a finicky thing. 1 pt or 2pts? It's a gamble all the way around when applied to momentum. At this point in the game, M had some. Taking the safer 1 point did not hurt or help it. Attempting and missing a 2 point conversion may have killed momentum, making it would have boosted it. In any case, an aggressive gamble would have been impressive and in and of itself and could have pushed the team.

joelr222

August 25th, 2009 at 3:19 PM ^

Any decision is easier to make in hind sight. With that much time left, you defense could score and change the game. I'm just saying any logical coach would make that decision in the moment, at least those that do not possess a crystal ball that is.