A week before the 2019 Super Bowl between the New England Patriots and Los Angeles Rams, I attended a predictive analytics event at UC Irvine that discussed predictive modeling. Both faculty and students presented their predictive methods to project the final score of the Super Bowl. It was quite fun and interesting to see how the wide world of data can be used to predict future outcomes and how the varied data sources were combined to formulate estimates.
The next day I was reading through some articles discussing the varied betting opportunities offered by Las Vegas and other sports betting websites that covered a range of events, not just the winning team and final score. I started browsing through the list of 450+ “prop bets” such as how long the national anthem will last, the first beer commercial, and many other non-final outcome related scenarios.
As I was reading through the list of prop bets I asked myself, “I wonder if I could come up with my own predictive model and apply it to these?” I quickly realized that no, I cannot come up with my own predictive model in time for Super Bowl application. It’s really hard. However I knew I could access NFL performance data and run some basic analytics to guide me toward the most reliable decisions.
So, I scanned the list of prop bets and picked 18 of my favorites. I wanted to test my own “hunch” predictions, based on many years of being an avid sports fan, and compare them to the accuracy of objective data-driven decision making. Just for the fun of it, I decided to include a column of randomly chosen outcome selections as well.
I first made my “hunch” predictions based on my existing knowledge and arm-chair expertise.
Next I pulled, organized and filtered performance data from Pro Football Reference. I was able to analyze data specifically focused on each prop bet so that my decisions were statistically supported. The depth of data available on this site was incredible and it provided everything I needed to make decisions based on data-driven insights. I didn’t always agree with the numbers, because I felt they didn’t take into consideration a lot of the “intangible” variables surrounding the matchup. But, for the sake of the exercise, I was committed to letting the numbers speak for themselves.
Lastly, using a random selection generator, I filled the column predicted by chance alone; sincerely hoping the random selections did not perform better than me.
After the Super Bowl ended and the official stats were released the next day, I marked the actual outcome for each prop bet and began comparing the three selection methods. Shockingly, my hunch did not have the highest prediction accuracy. In fact, it was not even close:
| METHOD | CORRECT PREDICTIONS | ACCURACY % |
| Data Support | 15 / 18 | 83% |
| My Hunch | 10 / 18 | 56% |
| Random Selection | 8 / 18 | 44% |
Not surprisingly, the predictions supported by data were the most accurate. Admittedly, this was a hit to my ego. However the one bit of solace I could find was that I beat out the random selection method… barely.
This was definitely an interesting exercise. Even though I use and trust data to guide decision making on a daily basis, when it came down to something like sports betting I genuinely felt that my knowledge internal understanding would come out on top. I trusted myself during this process more than I did the data, which I think is a great example of the dangers of making decisions based on hunch or intuition.
Had there been significant ramifications to these decisions, I would have been in serious trouble by trusting my hunch. As if I needed more convincing, this exercise validates a saying we have around the office… “Perception is unreliable, stick with data.”
If you’re interested in seeing the logic behind each decision, I’ve outlined my notes below. It’s a long read, but an interesting read.
OPENING COIN TOSS: HEADS or TAILS
MY HUNCH: Heads
In Kansas City for the AFC Championship game, the visiting Patriots got to call the opening coin toss. They elected “heads” and were correct. The game went to overtime where again the Patriots got to call the toss and again they correctly selected “heads.” Having won the toss and got the ball they won the game sending them to the Super Bowl. At the Super Bowl, the Patriots were designated the visiting team so just like in Kansas City they got to call the opening coin toss. My gut told me they were going to pick heads again and that they would be correct for a third time.
SUPPORTED BY DATA*: Tails
From a statistical standpoint, coin flip probability is always 50/50 split between heads and tails. Calculations can be run to determine the likelihood based on a streak, for example a streak of 3 heads in a row is more likely than a streak of 4 in a row. However this doesn’t change the overall probability of a single coin flip on its own.
That said, for the sake of this exercise I tried to find some statistical pattern to justify choosing one or the other. I looked at past Super Bowl coin flip results segmented by time zone, inside or outside of a dome, AFC or NFC making the call… but ultimately there was nothing even remotely convincing.
*Since the data suggested an equal chance between heads and tails, we went with the old adage, “Tails never fails.” – Because presumably at some point there a series of outcomes justified the saying. Right?
CHANCE SELECTION: Heads
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: TAILS
PREDICTION ACCURACY:
My Hunch – 0% (0/1)
Data Support – 100% (1/1)
Random Selection – 0% (0/1)
////
FIRST TEAM TO SCORE: PATRIOTS or RAMS
MY HUNCH: Patriots
This was Jared Goff’s first Super Bowl in his second season as a starter and no doubt the biggest game of his career. It will take him some time to shake off the Super Bowl nerves, and the Rams coaching staff will likely go conservative in the first quarter to help the young quarterback build his confidence. On the other side, this is Tom Brady’s ninth Super Bowl and throughout the playoffs the Patriots offense has shown no signs of weakness. Patriots score quickly and easily.
SUPPORTED BY DATA: Patriots
Throughout the regular season, the Patriots scored points on their first drive of the game 50% of the time. The Rams scored only 44% of the time. In the playoffs, The Patriots scored 4 touchdowns on 5 first-quarter drives, whereas the Rams scored only 3 times out of 5 first-quarter drives; and were held to field goals on each of those which means they gained less overall yards.
The data shows the Patriots gain more yards and have a higher first-quarter conversion percentage than the Rams, throughout both the regular season and the playoffs.
CHANCE SELECTION: Rams
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: PATRIOTS
PREDICTION ACCURACY:
My Hunch – 50% (1/2)
Data Support – 100% (2/2)
Random Selection – 0% (0/2)
////
FIRST SCORE OF THE GAME: TOUCHDOWN or ANY OTHER SCORE
MY HUNCH: Touchdown
In the AFC Divisional round, the Patriots had 3 first-quarter drives and scored 3 touchdowns. The following week in the AFC Conference Championship, the Patriots covered 80 yards and scored a touchdown on their first drive of the game. I don’t see this trend changing, so I’m betting the Pats roll and put up 7 to start the game.
Another outcome I can see is that the Rams go big on their first drive to get an early advantage. They have a group of receivers that present serious vertical, down-field threat. So a big play from the Rams wouldn’t surprise me either.
SUPPORTED BY DATA: Any Other Score
Looking back at the eight career Super Bowls played by Tom Brady, he has never scored a touchdown in the first quarter. That’s a shocking zero-for-eight. Furthermore, in those eight games the opposing team scored a first-quarter touchdown only twice. Historically speaking, a first quarter touchdown for a Patriots Super Bowl is unlikely.
As for the Rams, they haven’t scored a first quarter touchdown in either of their past two playoff games this season; the only points coming from field goals.
RANDOM SELECTION: Touchdown
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: ANY OTHER SCORE (FIELD GOAL)
PREDICTION ACCURACY:
My Hunch – 33% (1/3)
Data Support – 100% (3/3)
Random Selection – 0% (0/3)
////
JERSEY NUMBER OF FIRST PLAYER TO SCORE: 26.5 – OVER or UNDER
MY HUNCH: Under
I have a Patriots touchdown to open the scoring, and the offensive playmakers happen to have predominantly low jersey numbers. The WR and RB numbers for the Patriots are 11, 13, 14, 15, 18, and 26. Even the kicker’s number is under (3) so with a field goal I’m covered as well.
SUPPORTED BY DATA: Under
Between both the Rams and the Patriots, 11 of the 15 offensive players with the highest scoring likelihood (including kickers) wear jerseys with the number 26 or below.
RANDOM SELECTION: Over
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: UNDER
PREDICTION ACCURACY:
My Hunch – 50% (2/4)
Data Support – 100% (4/4)
Random Selection – 0% (0/4)
////
FIRST TEAM TO SCORE WILL WIN THE GAME: YES or NO
MY HUNCH: No
While I have the Patriots scoring first, I believe Rams defensive coordinator Wade Philips will adjust the team’s defensive strategy to limit the Patriots scoring as the game goes on. The Rams have the Defensive Player of the Year (Aaron Donald, DE) and a solid pass rush that can put pressure on Tom Brady, which is the key to beating the Patriots. Plus, in my heart I want the Rams to win so I see them starting slow but finishing strong. Patriots score first, but Rams win the game.
SUPPORTED BY DATA: Yes
Since 2002, the first team to score goes on to win the Super Bowl 67% of the time.
RANDOM SELECTION: Yes
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: YES
PREDICTION ACCURACY:
My Hunch – 40% (2/5)
Data Support – 100% (5/5)
Random Selection – 20% (1/5)
////
EITHER TEAM WILL SCORE WITHIN THE FIRST 6.5 MINUTES: YES or NO
MY HUNCH: Yes
Both the Patriots and the Rams have powerful offensive weapons and the Patriots have been making a habit of scoring on their first drive. Even if a touchdown doesn’t happen early, I have a hard time seeing neither team at least getting into field goal range.
SUPPORTED BY DATA: Yes
Throughout the regular season, the Patriots scored within the first 6.5 minutes 59% of the time; the Rams 50% of the time. In both playoff games leading up to the Super Bowl, the Patriots scored within the first 6.5 minutes; both were touchdowns.
RANDOM SELECTION: No
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: NO
PREDICTION ACCURACY:
My Hunch – 50% (3/6)
Data Support – 83% (5/6)
Random Selection – 17% (1/6)
////
EITHER TEAM WILL SCORE IN THE FINAL 2 MINUTES BEFORE HALFTIME: YES or NO
MY HUNCH: Yes
There’s usually a sense of urgency leading up to halftime. Teams open up a bit to gain momentum going into the break, so I think a field goal is a safe bet.
SUPPORTED BY DATA: Yes
Since 2002, a team has scored in the final two minutes before halftime 76% of the time.
RANDOM SELECTION: YES
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: NO
PREDICTION ACCURACY:
My Hunch – 43% (3/7)
Data Support – 71% (5/7)
Random Selection – 14% (1/7)
////
TOTAL TOUCHDOWNS SCORED BETWEEN BOTH TEAMS: 6.5 – UNDER or OVER
MY HUNCH: Over
The past three Super Bowls played by the Patriots (’15, ’17, ’18) have been over. The Patriots score five touchdowns against the Los Angeles Chargers, who had one of the top defenses in the league. Most experts are predicting the score total score to be over 60 points, so I feel at least three touchdowns each is reasonable to assume.
SUPPORTED BY DATA: Under
While throughout the season the Rams averaged 32.9 points per game, the average dropped to 28.8 during away games. The Patriots averaged 27.2 points per game, and only 21.6 away.
Since 2005, Super Bowls recorded under 6.5 total touchdowns 64% of the time.
RANDOM SELECTION: Over
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: UNDER
PREDICTION ACCURACY:
My Hunch – 38% (3/8)
Data Support – 75% (6/8)
Random Selection – 13% (1/8)
////
FIRST PENALTY OF THE GAME: PATRIOTS or RAMS
MY HUNCH: Rams
The Patriots have 38 players with Super Bowl experience compared to the Rams having only four. The Rams are a younger team with a less experienced quarterback and head coach. Typically these teams, due to the pressure of the situation, are prone to make mental mistakes.
SUPPORTED BY DATA: Rams
The Patriots have avoided committing the first penalty in their past three Super Bowl appearances.
From 2011-2017, the Patriots are called for 13% fewer penalties than their opponents during the regular season and 25% less during the playoffs. In 2018, the Patriots committed 15% fewer first-quarter penalties than the Rams.
RANDOM SELECTION: Rams
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: RAMS
PREDICTION ACCURACY:
My Hunch – 44% (4/9)
Data Support – 78% (7/9)
Random Selection – 22% (2/9)
////
MORE FIRST DOWNS RECORDED: PATRIOTS or RAMS
MY HUNCH: Patriots
The Patriots have a tendency to control the ball and control the clock. Rather than going for big plays I see them sustaining long drives which will ultimately lead to more first downs.
SUPPORTED BY DATA: Patriots
Over the previous three games, the Patriots have averaged 30 first downs per game compared to 25.7 for the Rams. In their past three Super Bowl appearances, the Patriots have recorded a total of 29 more first downs than their opponents.
RANDOM SELECTION: Patriots
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: PATRIOTS
PREDICTION ACCURACY:
My Hunch – 50% (5/10)
Data Support – 80% (8/10)
Random Selection – 30% (3/10)
////
TOTAL SACKS BY BOTH TEAMS: 3.5 – UNDER or OVER
MY HUNCH: Under
Both teams have been successful running the ball, so fewer passes means fewer sack opportunities. Plus, both offensive lines have been playing well and Tom Brady has yet to be sacked once during the playoffs.
SUPPORTED BY DATA: Over
Over the past seven Patriots Super Bowls, there has been an average of 6 sacks per game.
Over the past two playoff games, the Patriots average 3 sacks per game and the Rams average 1.5 sacks per game.
RANDOM SELECTION: Under
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: OVER
RUNNING PREDICTION ACCURACY:
My Hunch – 45% (5/11)
Data Support – 82% (9/11)
Random Selection – 36% (4/11)
////
LONGEST FIELD GOAL MADE: 47.5 – UNDER or OVER
MY HUNCH: Over
Every point counts during big games, and sometimes teams have to be aggressive and attempt field goals when they normally would punt. This game in Atlanta will be played inside a dome, so wind won’t be an issue. I’m predicting a field goal of at least 50 yards is attempted and made.
SUPPORTED BY DATA: Under
Since 2002, the average Super Bowl field goal is 37.4 yards with only two games (2002, 2004) recording a field goal of over 47 yards.
During the 2018-19 playoffs, the Rams average field goal is 33.5 yards and the Patriots average 35 yards.
The Rams kicker has a career 58% success rate for field goals 50+ yards, and the Patriots kicker has a career 71% from the same distance.
RANDOM SELECTION: Over
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: OVER
RUNNING PREDICTION ACCURACY:
My Hunch – 50% (6/12)
Data Support – 75% (9/12)
Random Selection – 42% (5/12)
////
MORE RUSHING YARDS: SONY MICHEL (NE) or TODD GURLEY (LA)
MY HUNCH: Sony Michel
For some reason the Rams are using their backup RB more than Gurley, which is limiting his overall touches and yards. On the other side, the Patriots are very effective with their running game and Michel is the lead running back. In this situation I’m going with the hot hand.
SUPPORTED BY DATA: Sony Michel
Throughout the regular season, Todd Gurley has averaged 89.4 yards per game compared to Sony Michel averaging 71.4 yards per game.
However, Todd Gurley injured his knee late in the regular season which has negatively impacted his performance during the playoffs. Over the past two playoff games, Todd Gurley has averaged 62.5 yards compared to Sony Michel averaging 121 yards. This is largely due to Todd Gurley’s minimal role in the offense, having 16 carries in the Divisional round and only 4 carries in the Conference Championship. This is compared to Sony Michel having 24 carries and 29 yards over the same two week period.
RANDOM SELECTION: Sony Michel
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: SONY MICHEL
RUNNING PREDICTION ACCURACY:
My Hunch – 54% (7/13)
Data Support – 77% (10/13)
Random Selection – 46% (6/13)
////
MOST PENALTY YARDS: PATRIOTS or RAMS
MY HUNCH: Rams
Due to the level of Super Bowl experience in favor of the Patriots and their tendency to have penalty calls go their way, it’s safe to say the Rams will rack up more penalty yards.
SUPPORTED BY DATA: Rams
Throughout the regular season, the Patriots averaged 46.5 penalty yards per game compared to the Rams average of 54.9 penalty yards per game. Furthermore, the Patriots averaged 8.1 yards per penalty compared to the Rams 9 yards per penalty.
RANDOM SELECTION: Patriots
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: RAMS
RUNNING PREDICTION ACCURACY:
My Hunch – 57% (8/14)
Data Support – 79% (11/14)
Random Selection – 43% (6/14)
////
TOM BRADY (NE) WILL THROW AN INTERCEPTION: YES or NO
MY HUNCH: No
Tom Brady has found success all season throwing short passes out wide and across the middle. The variety of offensive weapons he has and the amount of time he has in the pocket to run through his options, I think he has a clean game.
SUPPORTED BY DATA: Yes
Tom Brady has thrown at least one interception in three of his last 4 Super Bowls. The Rams are tied for 3rd in the league (T-Patriots) with 18 regular season interceptions, and have recorded two interceptions in their two playoff games this year.
RANDOM SELECTION: Yes
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: YES
RUNNING PREDICTION ACCURACY:
My Hunch – 53% (8/15)
Data Support – 80% (12/15)
Random Selection – 47% (7/15)
////
JARED GOFF (LA) WILL THROW AN INTERCEPTION: YES or NO
MY HUNCH: Yes
Jared Goff is a young quarterback going up against an experienced defense in a huge game. He’s going to make a mistake and the Patriots defense will capitalize.
SUPPORTED BY DATA: Yes
The Patriots defense is tied for 3rd (with the Rams) during the regular season with 18 interceptions. The Patriots also have two interceptions over their last two playoff games, and adding to the pressure is the six sacks recorded by the Patriots defense over the past two playoff games.
With the injury suffered by primary running back Todd Gurley, the Rams have had to shift their offense to be more pass-heavy. During the Divisional round, the Rams only had to pass on 38% of the offensive plays (Gurley, 16 rush att) and didn’t turn the ball over. During the Conference Championship they were forced to pass on 58% of their offensive plays (Gurley, 4 rush att) and Goff threw an interception.
RANDOM SELECTION: Yes
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: YES
RUNNING PREDICTION ACCURACY:
My Hunch – 56% (9/16)
Data Support – 81% (13/16)
Random Selection – 50% (8/16)
////
PATRIOTS SCORE A TOUCHDOWN IN THE FIRST QUARTER: YES or NO
MY HUNCH: Yes
The Patriots had no problem moving the ball against the Chargers last week and scored three touchdowns on their three drives in the first quarter. I see this confidence and high level of play carrying over into the Super Bowl.
SUPPORTED BY DATA: No
Out of his previous eight Super Bowl appearances, Tom Brady and the Patriots have never scored a touchdown in the first quarter. Throughout those games, the Patriots have punted on 67% of their first-quarter drives. The 2018 regular season numbers haven’t been great either; first quarter drives have been – punt 43%, touchdown 30%, and field goal 11%.
RANDOM SELECTION: Yes
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: NO
RUNNING PREDICTION ACCURACY:
My Hunch – 53% (9/17)
Data Support – 82% (14/17)
Random Selection – 47% (8/17)
////
RAMS SCORE A TOUCHDOWN IN THE FIRST QUARTER: YES or NO
MY HUNCH: No
As previously stated, I think the Rams play conservative to start the game and pick up steam along the way. Potential for a field goal in the first quarter but unless there’s a big play, I don’t see a first quarter touchdown.
SUPPORTED BY DATA: No
Throughout the 2018 regular season, the Rams only scored a touchdown on 32.7% of their first-quarter drives. In the playoffs they haven’t scored any first-quarter touchdowns, with 2/3 drives ending in either a punt or a turnover.
RANDOM SELECTION: Yes
We coded the outcome options and used a random number generator to make selections for each prop bet.
ACTUAL RESULT: NO
RUNNING PREDICTION ACCURACY:
My Hunch – 56% (10/18)
Data Support – 83% (15/18)
Random Selection – 44% (8/18)
