Super Bowl LVII is fast approaching and the excitement is starting to build. As we near one of the largest (and highly anticipated) events in American sports, we can’t help but consider the question on everyone’s mind: Which team will come out on top?
While we anxiously await the big game, let’s harness the power of Dataiku and the Elo rating system to make a data-driven prediction. Who’s ready for some football? Sit back, grab your game-day snacks, and let’s see who the data predicts will emerge victorious.
Using the Elo rating system
One popular way of ranking teams and predicting the outcome of future matchups is by using the Elo rating system. While the Elo rating system originated as a chess-rating system, it has countless applications including its current use for many online video game rankings. It was also part of the NCAA College Football BCS rating system from 1998-2013. One additional element of Elo ratings is that it can help forecast who may win a game even if the opponents have never played against each other.
If you aren’t familiar with the Elo rating system, here’s an overview of how it works. Each player (or team) starts with a set of points (we will use 1600 since it’s is common for sports Elo setups). Each player (team) wagers some points for each game that takes place and the winner receives most of the wagered points after the match following a mathematical function. If a highly ranked team faces a much lower ranked team (suppose one team has an Elo rating of 1650 and another has a 1550 rating), then the higher ranked team has a higher likelihood of winning and will get fewer additional points with a win but lose more points with a loss.
One nuance to note is the element of home field advantage. Most sports Elo rating systems boost the home field’s Elo rating by 100 points to represent home field advantage. This is what we will use in our example. This means that unless the Away Team has an Elo rating 100 higher than the Home Team, the Home Team will have a higher forecasted likelihood of winning. 100 points is quite substantial and it generally works out that with a 100-point difference, the team with a higher rating is expected to have a 64% chance of winning vs. 50% at a neutral site. Additionally, the super bowl (along with a few regular season games) are played at a neutral site so the Elo model for the NFL has to account for the “Home Team” not actually having home field advantage.
Building the NFL Elo rating system in Dataiku
Starting with a blank project, we can quickly build the Elo rating system thanks to readily available open source information regarding sports. We chose to use the NFL scores and betting Kaggle data as a source because it has all of the NFL scores since 1966 readily available along with details of the home team, stadium that the game was played at, weather, etc. Through the power of Dataiku, we easily pulled in the CSV files and continued building the model.
A quick review of the CSV file “spreadspoke_scores” showed us that the dataset needed to be parsed which was an easy Dataiku prep step. The date parsing was needed for the Elo rating Python code (next step in the flow) to understand the dates from the Kaggle data, sort the games in chronological order, and identify future games to predict.
With the prepped data, we were then able to build the Elo rating model in Dataiku using a Python code recipe. Elo rating code samples from Elopy provided a great starting point for establishing our Elo rating for NFL teams. We added some additional code regarding when to start the Elo ratings so it didn’t begin with all the games since 1966. We decided to start at the beginning of the 2022-2023 NFL season since a team’s performance in any given year generally seems independent from the previous year.
We chose to write the Elo results to an output dataset and have the prediction for upcoming games as a separate dataset. The upcoming game predictions include both the expected spread and the percentage chance that the Home Team wins. Negative numbers in spread imply the “Home Team” is forecasted to win, while a positive number implies that the “Away Team” is forecasted to win. The terms “Home Team” and “Away Team” are put in quotes as they are extraneous terms for the super bowl as it is played at a neutral site. The only significance is seeing which team is considered Home vs. Away to interpret the results. In the case of our model, the Eagles are entered as Home.
Considering the nuance of home field advantage for all but 6 games this season, we also modified the Python script to not boost the Home Team’s Elo rating when the game is played at a neutral site.
Say “Elo” to our predicted winner
Running the Elo rating system for the 2022-2023 season within Dataiku and opening the “upcoming_game_probabilities” shows that the Elo rating model predicts the Philadelphia Eagles to win Super Bowl LVII…but it just barely favors them. The expected margin of victory is .04 points (yes, less than 1 point). The Eagles are forecasted to have a 50.2% chance of winning based on the Elo model.
This is a very small margin because the Eagles and the Chiefs have almost the same Elo rating going into Super Bowl LVII – The Eagles have an Elo rating of 1693.71, and the Chiefs at 1692.59.
Comparing recent Dataiku Elo model performance
Could this prediction be correct? Let’s compare how the model has been performing so far. Our model predicted that the Chiefs would beat the Bengals in the AFC Championship (3.7 point spread with 64% chance of a Chiefs win) and that the Eagles would win the NFC Championship (2.9 point spread with 61% chance of a Eagles win). Those both turned out to be correct with the Chiefs beating the Bengals by 3 points with the 23-20 final score and the Eagles beating the 49ers by 24 points with the 31-7 final score.
In the case of the AFC Championship, the Chiefs and Bengals had very similar Elo ratings – within 3 points (1688.91 for the Chiefs vs. 1686.21 for the Bengals) – after the entire regular season and playoff games leading up to the Championship week. The model provided a 100-point Elo boost for the Chiefs since they were playing at home for the AFC Championship, and that Home Field Advantage boost was more than enough for the Chiefs to be favored.
In the case of the NFC Championship, the 49ers actually had a higher Elo rating (the 49ers had the highest Elo rating of all teams going into Championship week) than the Eagles by almost 20 points (1699.08 for the 49ers, 1680.24 for the Eagles). However, the Eagles playing at home with Home Field Advantage still made the Elo model favor the Eagles.
Alternate scenarios for Super Bowl LVII
While exploring alternatives for this forecast, we changed the model’s start date to October 18 – right after Week 6 of the regular season – expecting that the Chiefs would be rated higher since they only lost one game after Week 6. The Chiefs did indeed get a better Elo rating, and the Chiefs switched to being favored by .32 points and having a forecasted 51.3% chance of beating the Eagles.
But it would hardly be fair to boost the Chiefs by starting after Week 6, when the Eagles have losses in Weeks 16 and 17 with Jalen Hurts out for those games. Re-running the model with those two games for the Eagles not considered (deleted from the data set) provides a forecast of the Eagles being super bowl favorites by .67 points which is quite a bit higher than the originally estimated .04 point margin of victory. The Eagles then have a forecasted 52.7% chance of beating the Chiefs which is the highest percentage forecasted for either team in all of our scenarios.
Don’t bet the farm on the Elo model
In the NFL, anything can happen during any given week. While we’ve had a lot of fun preparing this Elo rating model, it’s best to steer clear of putting all your eggs in one basket when it comes to relying on its prediction. This model works quite well for games depending on individual performance, but there are complexities with team sports – especially ones with significant variation like NFL football and its related injuries. While an individual playing chess may have a bad night’s sleep or a headache that affects their individual gameplay in one day, it’s not nearly as varied as the week-to-week variation in the NFL. Afterall, who would have thought that the 49ers would win 6 straight games to end the season and go to the NFC Championship with their 3rd string Quarterback? Additionally, when two teams have an Elo rating difference of about 1 point after 17 regular season games and 2 playoff games and the model forecasts less than a 1 point spread, its really quite far from a sure thing.
Using the Elo model in parallel use cases
The Elo rating model has defined variables and evaluates NFL matchups consistently. It has more data behind it than choosing a winner based on who has a cooler mascot, a jersey that is your favorite color, or any other factors you or your friends may choose to consider when picking the team you plan to cheer for.
You can use Dataiku for any number of questions that you want answered and you can bet that there is a Dataiku integration that will make the Dataiku project far simpler and faster than you think. Have a league of your own and want to use the original design intent of Elo ratings? Dataiku could be your friend for next year’s “Pick’em league” or Fantasy sports draft and it will save you a lot of paper and manual work with a calculator.
The Super Bowl LVII countdown is on
Thanks to the power of Dataiku, we utilized open source data and external connections to create a Dataiku project with an Elo rating system in just a couple of hours. Who will Team Amplifi be rooting for? We are staying true to the spirit of American football where it is said that:
“on any given Sunday, any team can beat any other team”
Jon JorgensonSenior Data Consultant
So, like you, we anxiously await gameday to put our model to the test and see if the Philadelphia Eagles will take home the Super Bowl LVII title as predicted on February 12th.
Ready to turn game day modelling into smarter forecasting at work?
Building an Elo model for the Super Bowl highlights how unpredictable things can be. Neutral-site matchups, last-minute injuries and changing performance trends all influence the outcome. Exploring these variables in Dataiku is more than a fun exercise. It builds the analytical skills you need for real-world forecasting.
When you work with open-source data, scenario adjustments and probability shifts, you are not only predicting a winner. You are learning how to test assumptions, understand uncertainty and model outcomes that support better decisions across your organisation.
If you are ready to apply that same curiosity and structure to your business data, we would be happy to help.
In the preceding chart, we see the relationship between the team’s seed (1 being a top seed) and the number of wins (or “rounds”) those seeds achieve on average in NCAA tournaments. We have a total of 960 team seasons to analyze, an adequate amount of data to base our analysis on. From this chart, we can observe some very interesting insights – among them:
- 1 seeds in the bracket tend to make it a little past the 3rd round (meaning they win more than three games).
- 16 seeds almost never win a game. 16 seeds are always matched against 1 seeds to start the tournament (called a seeded knockout tournament). Picking a 16 seed to win even a single game is a real shot in the dark!
- There’s a tremendous dropoff from 1 to 2 seeds – over a full round. Logic might lead you to believe that on a scale of 1-16, there is a very small skill difference between a 1 and 2 seed, but in this case, the historical success does not agree.
- Other significant cliffs also exist from the 3 to 4 and 4 to 5 seeds – nearly a half-round dropoff between each.
- If we followed an exponential decay curve (which is what the rest of the data looks like), we’d expect the 6 seeds to be slightly more successful than the 7, but they’re not. In the first round of the tournament, 6 seeds face off against 11s, and for some reason, they don’t fare well. 11s do nearly as well as 7 seeds.
- If you’re trying to find that Cinderella story, it’s pretty clear that, even rounding up, anything after the 12th seed doesn’t usually win a game. Of course, it seems that every year there’s a double-digit team or two that makes a deep run, but it certainly is a statistical anomaly when they do.
From this analysis, it’s clear that a team’s seed plays a significant role in determining its likelihood of success in the tournament. Higher seeds, particularly 1s and 2s, have historically advanced further, while lower seeds face an uphill battle, with only a few breaking through with unexpected success.
There are also some unexpected patterns—such as the strong performance of 11 seeds—that challenge conventional wisdom. While seeding provides a useful baseline for predicting tournament success, it doesn’t tell the whole story. To gain a deeper understanding, we need to explore additional factors, such as Elo rankings, to see if they offer further predictive value.
Is Elo the Dark Horse?
Let’s take a look at a similar chart to see if a team’s Elo ranking offers similar predictive value in determining tournament success. Elo is designed to assess a team’s relative strength based on past performance, and in theory, it should provide a useful measure of a team’s potential. In this chart, we observe a familiar trend—just like seeding, teams with a lower (better) Elo ranking tend to win more games in the tournament. This suggests that Elo, like seeding, captures important information about team performance.
Some additional key insights emerge from the data. We observe steep drop-offs in tournament success once teams fall outside the top 60 in Elo rankings, with those teams rarely advancing past the first round. This aligns with what we observed in seeding, where higher-seeded teams struggle to make deep tournament runs. But does this mean that Elo and seeding are interchangeable metrics? Or does Elo offer additional predictive value beyond what is already captured in the seeding process?
A natural question arises: Does the NCAA selection committee already factor Elo into their seeding decisions? If seeding and Elo rankings are perfectly correlated, then Elo may not add much unique insight. However, if discrepancies exist between the two, it could suggest that Elo captures something distinct—perhaps a team’s true strength beyond just its tournament seed. To explore this further, let’s look at a specific example where Elo might provide additional predictive power– the curious case of 11 seeds.
A Cinderella Story: The 11-Seed Anomaly
In the previous section, we uncovered a fascinating anomaly in tournament performance for 11 seeds. While most seed-based trends follow a predictable, seemingly exponential decline in success as seed numbers increase, 11 seeds consistently outperform expectations. This unusual pattern suggests that something beyond just seeding might be influencing their success. To dig deeper, we can isolate 11 seeds and examine their performance based on Elo rankings to see if this additional metric provides further insight.
The following chart focuses solely on 11 seeds and their relationship between Elo ranking and tournament wins. An interesting trend emerges—Elo ranking appears to be a strong predictor of success for 11 seeds. Teams with an Elo ranking higher than 20 perform about as expected, aligning with historical seed-based trends. However, the true Cinderella 11s—the ones that make deep runs—tend to have an Elo ranking inside the top 20. This suggests that not all 11 seeds are created equal. Rather, certain 11 seeds have a much higher chance of making a surprising tournament run, and their Elo ranking helps us distinguish them.
While this visual analysis alone doesn’t provide a definitive predictive model, it does offer a compelling case for further exploration. The strong relationship between Elo and tournament success among 11 seeds indicates that Elo could provide additional predictive value beyond just seeding. This finding encourages us to quantify these relationships more rigorously in the next step of our analysis.
Second Round: Quantifying the Relationships
The visual trends we’ve explored so far suggest a strong relationship between a team’s seed, Elo ranking, and tournament success. Both metrics appear to play a crucial role in determining how far a team is likely to advance. However, visual patterns alone can only tell us so much—we need a more precise way to measure the strength of these relationships.
To take our analysis a step further, we’ll quantify the correlation between seed, Elo ranking, and tournament wins. This will allow us to determine which factor typically has a stronger predictive relationship with success. By applying statistical correlation methods, we can assess whether these metrics are related to wins and how strongly they influence a team’s tournament performance. This will help us understand whether seeding alone is a good predictor of success, or if Elo provides unique and valuable insight.
Seeds, Standings, and Spearman Correlation
One effective way to measure this relationship is by using Spearman correlation, which evaluates how well the rankings of two variables align. Spearman correlation assesses monotonic relationships—meaning it identifies whether one variable tends to increase as the other increases, even if the relationship isn’t perfectly linear. This makes it particularly useful for tournament data, where performance metrics like seed and Elo rating may not have a strictly linear impact on wins but still follow a clear ranking-based trend. By calculating the Spearman correlation, we can determine which factor—seed or Elo rating—has a stronger and more consistent association with a team’s success.
Performing a Spearman correlation in Dataiku is an easy task and in the resulting matrix, we can see that both metrics have a highly inverse relationship with wins (as expected). As we may have gleaned from the previous two charts, the team’s seed is more strongly correlated with the number of wins they’ll see in the tournament than their Elo ranking is. This doesn’t mean that the Elo isn’t valuable for making predictions, just that it historically isn’t as individually correlated.
While both seed and Elo rating are strongly correlated with wins, it’s important to note that these two metrics are not perfectly correlated with each other, with a Spearman correlation of 0.841. This suggests that while they share a significant relationship, they are likely capturing different aspects of team strength. As a result, combining both metrics in predictive models could add valuable insights, as each brings unique information to the table. This mirrors what we observed in our analysis of the 11 seeds—where using multiple factors together provided a more complete picture of a team’s potential tournament performance.
The Sweet 16: Data Preparation
After performing this analysis on the Resumes data, we have a grasp on a couple of the interesting factors that influence the depth of the run a team might take in the tournament. Our goal with this analysis is to see if we can predict the outcome of individual games – so in the next step, we’ll join these historical rankings from the Resumes data with the larger dataset containing the actual results of each NCAA tournament game since 2008. Along with the seeds and Elo rankings, we’re also going to pull a couple of other potentially valuable team metrics from Resumes, “B Power” and “WAB Rank. We’ll dig into these more in our next article.
The details of this join are available in the downloadable Dataiku project, but at a high level, we’re joining the datasets, using the team name and tournament year as the join keys.
The result of this join gives us a record including the score of each game in tournament history, along with each team’s seeds, Elo, WAB, and R-Score rankings.
Creating the Features
This snapshot of our data is interesting, but to better understand and describe any given matchup, we may want to generate some more specific features involving the difference between the teams’ seeds, WABs, and Elos. Specifically, we’re going to utilize a formula in Dataiku that subtracts the away team’s value from the home team’s value to give us the differential between the two teams. For example, if the home team is a 1 seed and the away team is a 16, the differential would be -15, indicating a large discrepancy between the seeds in favor of the home team.
Also using a formula, we can create another one-hot encoding feature that indicates whether the home team (team 1) won each game
Now, with our data joined and prepared, we can look at the new relationships to see if the insights of seed / Elo / wins still hold true in individual games. Performing a Spearman correlation, we observe a significant inverse relationship between both the seed and Elo differential (as well as the WAB and B-Score which we’ll dive into more in the next article) with the winner of any individual game, just as we may have expected.
