website statistics
While most of the talk around the Packers victory over the Cowboys will center around the huge incompletion reversal (56% Dallas win probability to 81% Green Bay win probability) and Rodgers stellar play on a bad calf, there were two interesting decisions in the second half, both of which, Troy Aikman had plenty to say about.

The first decision: After scoring a touchdown to be down 21-19 with 1:41 left in the third quarter, should the Packers have gone for two? My immediate reaction was, yes. It's always better to know what you will need and have the information. Brian agreed. Here are the numbers:

Kicking the extra point: 44.9% Win Probability for Green Bay
Converting Two-Point Attempt: 52.6% WP
Two-Point Attempt Failure: 41.6% WP
Expected Win Probability Going-For-It: 46.9% WP

Aikman's defense of Mike McCarthy was that very frequently, teams "regret chasing points." He is saying that people always remember when a two-point conversion fails and the team ends up losing in a close one because of it - this is a pretty large bias. League-wide, two-point attempts convert around 48%. The break-even point to go for it was around a 30% conversion rate for McCarthy. This is not a huge swing, so little probability shifts here and there can definitely effect the decision. These numbers include team strength as well. That being said, McCarthy potentially sacrificed 2.0% in win probability by kicking the extra point.

The Packers would ultimately score again and miss the no-brainer two-point conversion, leading 26-21. The next decision came on 4th-and-2 from the Green Bay 32-yard line - on the very play that Dez Bryant's catch was eventually overturned. Aikman strongly disagreed with the decision to go for it, down five, with 4:42 left in the fourth quarter. The Cowboys did have two timeouts remaining, but a 50-yard field goal attempt converts league-wide at 69.4%. The conditions for field goal kicking were also subpar, but the Cowboys have one of the best kickers in history in Dan Bailey (although he already missed one field goal during the game). So, let's say those two factors cancel out and the 69.4% is accurate for our baseline analysis.


Before the play, the Cowboys had a 30.7% chance to win the game. A successful conversion means 50.0% win probability. Failure means 18.6% win probability. A missed field goal is essentially the same as the Packers would be looking to run the clock out. A successful field goal would mean a 28.2% chance to win for Dallas - and they would have a decision about whether or not to onside kick, but we assume they would not. A punt is also an option, which would result in an expected 20.1% win probability.

In the middle of the field, 4th-and-2 converts around 58.0%, which results in:

E[WPGo-For-It] = 0.500 * 0.580 + (1 - 0.580) * 0.186 = 36.8%.

E[WPField Goal] = 0.282 * 0.694 + (1 - 0.694) * 0.186 = 25.3% < 36.8%

The break-even point is 21.4% going for it, which means as long as the Cowboys think there is a better than one-in-five chance to convert, they should go for it (which they did). 

So, in both cases, based on the league-wide numbers, Aikman was incorrect - particularly in the case of going for it. Jason Garrett certainly made the right call - although the right play-call is a different story.

While everyone is angry about the overturned call, keep in mind that the referee needed indisputable evidence to overturn the call. It was then indisputably an incomplete pass. That being said, ever since the Calvin Johnson touchdown was overturned, there have been complaints about the rule itself. So complain to the NFL about the rule, but not the officiating. It is not the same as the previous week against the Lions where there was a judgement call and the flag was picked up.

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

Leave your comments

Post comment as a guest

0

Comments (7)

  • Guest - John

    great commentary as usual, Keith. Is there any way that we can send a messenger to personally deliver this article to Aikman, or better yet, get color commentators who know what they are talking about? then again, it is kind of fun to make fun of them (he and Simms).

  • Guest - lions fan

    As a lions' fan, I did get a little bit of pleasure at how the game ended. As a fan of football, games, and sensible rule sets, I found it ridiculous to the point of wondering why I would watch such a game.

  • Guest - James

    The 4th down bot said that going for it was wrong, in favor of punting.

    http://nyt4thdownbot.com/play.html?gameid=01112015_DAL@GB&playid=20150111009385

    In that situation, punting seems to make no sense. I realize that the bot is a bot, but some logic to help it discern when punting doesn't seem like a good idea would be good (such as when punting probably results in a touchback). Only 5% of the data includes that situation, making it seem rare enough to have to dig into the situations that caused that choice.

    If a team punted, it would likely end up on the 20. That would give a 0.78 (ish) WP for the Packers (or a .22 for the Cowboys, basically the FG attempt value). The 4th down calculator says that the Cowboys would have a .32 WP, though. I'm really curious about what kind of situations coaches opted to punt from the opponent's 32 when behind in the score. A punt that is highly likely to cause a touchback should be evaluated a bit differently, I think.

  • Yeah, I saw that too and thought the Bot/Brian's older model had an issue with the win probability following a punt as well - I think his newer model has definitely improved upon this. His listed 32.0% win probability, mine was significantly lower at 20.1% (part of this is Green Bay being favored, which isn't taken into consideration in the Bot's decision). Similarly, the Bot does not take timeouts into consideration either.

    The expected opponent field position on a punt from there is around the 10-yard line, though, not the 20.

    A lot of situations inside the last five minutes of the game are very difficult to model due to lack of data and huge changes in behavior/strategy.

  • Guest - Daniel Davis

    "the referee needed indisputable evidence to overturn the call. It was then indisputably an incomplete pass. "

    No, he *needed* indisputable evidence. If there's one adjective you can't apply to that call, it's "indisputable".

    You can make a bad argument that covering 5 yards with 3 steps, changing the ball from one hand to another, lunging forward, and planting the free hand to further advance the ball somehow does not constitute acts common to the game to advance the ball. But it simply doesn't pass a laugh test to call that bad argument "indisputable".

  • Guest - Daniel Davis

    For your calculations, you need to take into account that Rodgers completed something like his last 10 passes, and that's before any necessity of going for it on 4th down. The Cowboys defense was not going to stop Rodgers. They would drive down, as they did without really needing a score, run out the clock, and score.

    That's what the Cowboys needed to do. Go for it, but slow play it so if they make it, they would get the last score. Make it a game of the Cowboy's offense against the GB's defense, where the Cowboys had the advantage.

  • Guest - John gatto

    Dallasvwasvrighrtbtobgovforvtwo,gutnhotvwithvthebplaynitbchose;dramatically,avsinglenpointnonlymprolingscthecagonynforbtruevfahns,andoesbnothingbtonibspievthecdefehsevtovresistvgivingbupcavfstalbfieldbgoal..fanbeasevuscwhatgthenhiddenbdyhamidbofbfootgallmiscabout,ultimately2wasvrithtbthinkihg,
    Butvavbootlegborcabhalfbacknpasscwouldnhavecdonevthebjob.Boisevwouldbhavenmadevit,socwouldvhave Belichek