Alex Egol – RSAC Writer
I first want to apologize for my freezing cold take on the last Patriots blog post I wrote. The numbers were on my side, but the referees were not. Now Super Bowl LIII, between the Patriots and Rams, is upon us.
Like any Super Bowl, Super Bowl LIII has some interesting storylines to look out for. It is a rematch of Super Bowl XXXVI, which was Tom Brady’s first championship and played when Jared Goff was 7 years old. The Super Bowl will feature the youngest head coach in history, Sean McVay. McVay could become the youngest coach in history to win a Super Bowl, if the Rams are successful. That takes us to our next subject of discussion: calculating a win probability for each team.
In the 43rd and 45th chapters of Mathletics, a book we’ve already talked about on this blog, Wayne Winston discusses his use of an Excel function called NORMDIST to determine win probability given a point differential. While Vegas point differentials are often subjective in that they seek to place even amounts of money on both sides of a bet, we will still use this NORMDIST, also known as P-F-R Win Probability Method, to come up with a preliminary estimation for win probability.
The NORMDIST function takes in your assigned (x, mean, standard_deviation, and cumulative) – terms Excel geeks will understand, and outputs the value of the normal distribution function. The “x” we input in will be equal to 0.5 and -0.5, for reasons related to end-of-game ties. The “mean” will represent the point spread for the Patriots (for which we’re calculating win probability), which is -3. The “standard_deviation” we input will be 13.45, which represents the overall NFL average standard deviation for win probability from 1978-2012. I gathered this data from https://www.pro-football-reference.com/about/win_prob.htm, a fascinating Pro Football Reference Article, which will help for any clarification you may need. The “cumulative” we will input is called TRUE, which is irrelevant to our statistical discussion.
I used the same style of calculation that Wayne Winston used to find the Colts’ win probability given a 7 point spread advantage. The PFR article explains this. I have attached my calculations below.
Calculation for Patriots’ Win Probability.
The Patriots have a 58.82% chance of winning Super Bowl LIII, according to our application of the NORMDIST function over the current Vegas point spread. However, as I said before, point spread is not the most dependable metric upon which to base your calculation. We will experiment with other ways of calculating win probability.
We have discussed FiveThirtyEight’s Elo metric before, in our conversation about the Patriots’ dynasty – which was proved incorrect by the Patriots’ success these playoffs. If you didn’t read that article, https://fivethirtyeight.com/methodology/how-our-nfl-predictions-work/ is a useful link that will explain Elo in detail. According to FiveThirtyEight, the Patriots have a 53% chance of defeating the Rams in Super Bowl LIII.
It makes sense that Elo is more generous to the Rams than NORMDIST because Elo is a cumulative metric. Thus, the results of the regular season impact the Elo ratings going into the Super Bowl. The Rams won more games than the Patriots in 2018. Therefore, in our proxy for the Rams win probability, we’re factoring a metric that helps the Rams more than it helps the Patriots.
Statistics tell us the Patriots are favored over the Rams. I will tentatively choose them to defeat the Rams. But, when both teams take the field on Sunday Feb. 3 in Atlanta, there is no doubt that both teams are qualified, and capable of taking home the Lombardi Trophy.