## Panthers Punt Away The Win

On Saturday, I watched as my Northwestern Wildcats punted on 4th-and-1 from midfield, up three with 1:30 remaining. A conversion would have guaranteed a win, allowing the Cats to kneel down. By punting to Michigan State, I was certain NU would lose in dramatic fashion once again. Luckily, the Cats defense came up big, making what seemed like their first fourth-quarter stand of the year. Ron Rivera had an eerily similar decision against the Bucs this week.

With just over a minute left, the Panthers had the ball at the Tampa Bay 49, up 21-13. The Bucs had used all three of their timeouts, so a conversion wins the game. Not only do the Panthers have one of the best goal line backs in the league in Cam Newton, but they also have the human bowling ball, Mike Tolbert, who was born and bred for those situations.

League-wide success rate on 4th-and-1 is 74% and as mentioned, with the Panthers superior short-ground game, their estimated success rate is likely even higher. A conversion means 99.9% chance of winning if the opponent has no timeouts -- barring a Philip Rivers-esque fumble on the kneel down. A failed 4th-down attempt means a 93% chance of winning. Not only does the opponent have to score, but they have to convert a 2-pt conversion as well. A punt, which likely nets about 30 yards, results in a 96% chance of winning the game. Either way, the Panthers are sitting pretty. And while the expected win probability (98% after adjusting for timeouts) does not seem like a significant deviation from 96%, notice the break-even point: 50%. In fact, after adjusting for the timeouts, the break-even point drops to close to 42%. That means that the Panthers would have to believe they could only convert the 4th-and-1 four times out of ten or less.

While it's necessary to evaluate Ron Rivera's decision on the process, which is outlined above, let's look at the outcome. Tampa Bay drove downfield in 50 seconds on six plays, scoring a touchdown and converting the two-point conversion, sending the game to overtime. The Bucs got the ball first in OT and never looked back, scoring a TD on their opening drive to Dallas Clark. So, after punting on 4th-and-1 up eight points, the Panthers never ran a real offensive play (they did kneel down once at the end of regulation).

On another note, I cannot express enough disdain for John Fox's decision to kick a field goal on 4th-and-Goal from the 1 in a 7-7 game, early in the second quarter. The numbers say to go for it unless you can't convert at least 32% of the time. Peyton Manning can convert that well over 32%. The CBS commentator, though, went on a long rant about him making the correct decision to "take the points, because points usually win football games." When is Chip Kelly coming to the NFL?

Keith Goldner is the creator of Drive-By Football, and Chief Analyst at numberFire.com - The leading fantasy sports analytics platform.  Follow him on twitter @drivebyfootball or check out numberFire on Facebook

### 16 Responses to “Panthers Punt Away The Win”

1. Steve Hall says:

Hah! I knew I'd read about that Solomon Wilcox rant here this morning. The whole time he was talking, I was thinking, "you play for the MOST points!"

2. Steve Hall says:

Hah! I knew I'd read about that Solomon Wilcox rant here this morning. The whole time he was talking, I was thinking, "you play for the MOST points!"

3. Pat says:

From the AP

http://scores.espn.go.com/nfl/recap?gameId=321118029

"But after a 40-yard field goal by Connor Barth, the Panthers' offense couldn't put the game away and had to punt."

No one ever HAS to punt

4. Jeff Clarke says:

Did anyone else see the Browns-Cowboys game. The Browns had 4th and goal from the 1 with 1:30 to go down 4. The announcers described it as "decision time" and said it was a "tough call". In some fairness, they did say you have to go, but even the fact they would hesitate at all says so much about their mindset.

5. Jeff Clarke says:

Also did anyone see that the Jags went for it on 4th and 10 from the 50 in a tie game in Overtime? I thought I'd never see a coach go in a situation where I would punt, but I'm not sure about that one.

6. Chase says:

Sorry if I've missed this before, but how do you adjust for timeouts?

I'm curious because when Tampa Bay trailed the Panthers 21-13 with just 32 seconds remaining, and had a 3rd and 4 on the Carolina 41-yard line, the WP model says they had a 5% chance of winning. Since they were out of timeouts, is it lower?

(I ask because I'm using this example for the Fifth Down tomorrow.)

7. Anonymous says:

Jeff: Yes, saw that in Browns game, drove me nuts as well.

Another brilliant comment late in Chargers-Broncos: "Keep in mind, the Chargers have already recovered an onsides kick". Yes, because the odds of recovering a surprise onsides kick are exactly the same as recovering an expected one, so that's a very insightful point to bring up minus the context.

8. Anonymous says:

I would be interested to see you analysis of 4th downs by CIN Marvin Lewis. based on the game against KC, he seems to get it.

9. Jon Jackson says:

Your analysis would be valid, if the 4th and 1 was not against the TB best in the league run stopping defense, lowest yds per carry, and most run stops behind the line of scrimmage.

TB probably would have stuffed them. Analysis should be situational dependent. The WP for a 4th and 1 fail is 0.93(your numbers), 3% less than a punt.

Also coaches like their jobs, remember the Atlanta, NO game last year in OT when Mike Smith went for it on 4th and 1, and then NO kicked the game winning FG.

10. Keith Goldner says:

Jeff-
Saw that in the Browns/Cowboys game. I was thinking, "Only Pat Shurmur would kick a field goal here... oh wait". Also, loved the Jags going for it.

Chase -
You need to do some manual adjusts for timeouts since in the model, average timeouts left are assumed. I would say you are right, that it is less than 5% although, probably not dramatically. Timeouts have a lot greater effect on defense than offense, but obviously there is an effect.

Anon-

Cin decisions this week were really good for the most part. Loved the fake punt and few fourth down attempts. They did punt on 4th-and-6 from the KC 39 in the 3rd, but they were up 15 and that's not a must go for it situation. Going for it, though, is the better option EP = +0.64 vs EP Punt of +0.04. Interestingly, up 15, the win probability actually increases by 1% on a punt.

11. Keith Goldner says:

Jon-
I have TB as the No. 8 opponent-adjusted rushing defense and CAR as the No. 13 opponent-adjusted rushing offense. Yes, these are baselines so it's definitely appropriate to adjust for opponents/situation further, but don't think that would shift it dramatically. Saying TB "probably would have stuffed them" when 4th-and-1 is converted around 74% league-wide is just no accurate.

12. Anonymous says:

Forget the number stuff. It was time for the coach to say to the offense GO WIN THIS GAME. The defense/offense/coaching staff has clearly demonstrated numerous times this season that they can't/won't put teams away and win a game in the 4th quarter. Choosing to punt just reenforces the idea that the team can't win.
I thought that the kneel down to end the game was poor and demonstrated a give up by the coaching staff. Sure there was a small chance that they could have done something. But they had one timeout and about 20 sec or so. That's time to at least try to win the game. Time to run 3 or 4 plays and maybe get to try a long field goal.
Les

13. Anonymous says:

Keith, I'm afraid I agree with Jon. While it's been well-demonstrated that NFL coaches punt too often, to argue that there's a substantial difference between a WP outcome of 0.98 vs 0.96 is a bridge too far.

It feels like you're applying gambler's logic: in gambling, you'd take a 2% advantage every time, because your probabilistic model is exact. In the WP model, the numbers are fitted to average data -- the model itself is an estimate.

The strongest argument that you can make from the WP numbers is that there was no obvious answer: Rivera truly had to make a decision.

-Andrew

14. Anonymous says:

There are many many things wrong with this post. Here are just a few:

2 of the panthers previous run plays went for 0 yards and -2 yards. Cam scrambled for 11 on a 3rd and long, which obviously had a difference defense that you would see on 4th and 1. When the coach looks directly at the previous few plays, the see that running has failed. With those players, against that defense, at that point in the game. League wide precentages are meaningless, but even worse is the hand-wavy 'but the panthers are even higher".

ah, the hubris of the stat heads. Actually thinking their 96% vs 97% doesn't have a +-10% on it.

The punt was the right decision, and talk of that "break even" point is meaningless. Why is it meaningless? because IT IS ALREADY IN THE TOTAL WP, which were effectively identical.

Also, what are the decimal places on those numbers? I doubt the difference is > 1%.

FYI, you can bootstrap an estimate on that league wide 74% success rate. I bet the spread is huge.

15. Anonymous says:

One notes that the WP quoted above of 97% and 93%, all have somewhere around a 20% probability of a field goal.

in fact, the probability of a field goal attempt is zero in those situations.

16. Jon Jackson says:

I just watched the game and Tampa stuffed Carolina on the 3 previous Rb run plays. the statisticians says 0, 0, -2 but that is generous, more like -1,-1,-2.

The system that I am developing(granted it will take probably another season before the appropriate weighting of the variables is achieved) uses no less than 20 statistical parameters for each team, all adjusted for strength of opponent, and normalized for league averages.

Right now I achieve approximately 1 more win over vegas spread per week or about 56%. This is for all 13, 14, 15, or 16 games being bet on. Real value probably can be obtained for about 3 to 4 games per week.

I'm working on it, and the data set is small, 27 weeks. As I said above, It'll be at least a season.

All those dependent on these guys for a system, I say watch as many games as possible, stare at the stats and develop your own system.