How Predictable is Major League Soccer?

Photo by Chris Liverani on Unsplash

One of the greatest aspects of soccer is its unpredictability. Any club can beat any club on any given day and that is part of why we watch week in and week out. We cling on to watch an unpredictable outcome such as Senegal beating the defending champions at the time France in the opening game of the 2002 FIFA World Cup, North Macedonia beating Italy in the 2022 FIFA World Cup qualifiers as well as Sheriff Tiraspol beating Real Madrid in the group stage of the 2021 Champions League. This unpredictability is what we will measure today in Major League Soccer.

How predictable is the MLS? Well lets ask Pinnacle Sports Betting. It operates as an online gambling website founded in 1998 in which people can place bets on soccer games. The odds published by Pinnacle can be tracked in the website Football Data.co.uk for leagues in 27 different countries and easily downloadable in CSV files. Below is an example of the data provided by Football Data.co.uk. In the April 18th, 2021 game between Columbus Crew and Philadelphia Union Pinnacle projected that the Home Team would win 1 out of 1.87 times, the Away Team would win 1 out of 4.25 times and they would draw 1 out of 3.86 times. Therefore since the numerical value is lowest for the Home Team, 1.87, then Pinnacle expected the Home Team Columbus Crew to win the game. Ultimately this particular game ended in a 0–0 draw therefore Pinnacle prediction was not correct.

Given that Football Data.co.uk. provides betting odds data for various leagues around the world we can aggregate the number of correct Pinnacle prediction based on calendar year for the ten years between 2012 and 2021. Pinnacle was able to correctly predict 56.1% of the games in the Italian Serie A while only being able to correctly predicts 47.4% of the games in Mexico’s Liga MX. For the MLS Pinnacle correctly predicted 52.2% of all games.

From here we can take the 52.2% correct predictions from Pinnacle for the MLS and separate it out by year. Pinnacle’s predictions were in the range of 55.1% correct in 2018 and their lowest was in 2021 at 49.2%.

Going a bit deeper at the individual club level we can split the number of correct games by Pinnacle whether the club was playing at Home or Away. However we do have to adjust for the total number of seasons that each club has played in the MLS. Reason being that a couple of clubs have played fewer seasons than others which provides Pinnacle with less historical data on which to create their predictions. Overall there are a couple of interesting observations:

1. Austin FC who had their inaugural MLS season in 2021 is in the bottom right quadrant by itself with 41.2% correct Home predictions and 76.4% Away predictions. This is quite peculiar because of the big difference between the number of correct predictions by Pinnacle for both Home and Away. However Pinnacle had no historical data on Austin FC given it was their inaugural season and prediction performance by Pinnacle outside of the normal range would be expected.
2. Nashville SC and Inter Miami are two clubs that joined the MLS in 2020 however their Pinnacle predictions are quite different. Nashville SC stands at Home 46.7% and Away 34.4% while Inter Miami by contrast stands at Home 55.6% and Away 48.4% These differences are striking given that Inter Miami stands within a normal range of correct predictions by Pinnacle while Nashville SC stands by itself in the bottom left quadrant.
3. New York City FC and Orlando City joined the MLS in 2015 and have quite different Pinnacle predictions. New York City FC stands at Home 58.3% and Away 43.4% while Orlando City by contrast stands at Home 44.1% and Away 54.3% These prediction are interesting given that both of these clubs have seven seasons of data available and are still quite different from each other.
4. Seattle Sounders has the highest correct Home predictions at 63.6% followed by Atlanta United at 62.5% and New York Red Bulls at 60.3%

Furthermore we can measure the average odds at the individual club level for all the games they played in. Similarly as the graph shown above the clubs are color coded based on the total number of seasons that each individual club played.

• Overall the club with the highest average odds is Chivas USA at 5.45 for which there is data from 2012 until their dissolution in 2014. In those three seasons Chivas USA placed near bottom of the league table.
• The next highest odds is FC Cincinnati at 4.7 which aligns with their performance in the MLS with being the bottom table club since their inaugural MLS season in 2019.
• At the other end of the spectrum is Los Angeles FC with an average odds of 2.11 who had a pretty stellar start in the MLS by finishing 3rd, 1st, 7th and 9th in the league table since their inaugural season in 2018.

Lastly I ran an OLS Regression Model on the MLS Overall League Table Position vs the Average Odds for a club. The Position variable is created by dividing the position in the Overall League Table by the number of clubs that competed in that season. As mentioned before the number of clubs in the MLS has changed across the years so I have to normalize for such. For example New England Revolution which ended in 1st place in the 2021 MLS Season is 1/27 clubs = 0.037 (3.7%). FC Cincinnati which ended in 27th place is 27/27 = 1 (100%) given that in the 2021 season there were 27 competing clubs. The OLS regression results indicate that the Average Odds variable is statistically significant at the 1% level with a 0.26 coefficient. This 0.26 coefficient means that if the Average Odds of a club were to change by 1 (i.e. going from a 3 to 4) they are expected to increase their Position by 26%, hence they would drop in the table by about 26%

Conclusion: So how predictable is the MLS? Overall I would say 50/50. Comparing the Pinnacle predictions for the MLS to other leagues around the world most of the leagues were very close to the 50% mark. The Italian Serie A was the most predictable at 56.1% while Mexico’s Liga MX was the least predictable at 47.4% When one looks at the MLS from 2012 to 2021 each season did fluctuate around an average of 50% correct predictions with 2020 and 2021 being the lowest correctly predicted seasons by Pinnacle for the MLS. And lastly using an OLS regression model on the League Table Position of a club vs the Average Odds can help to determine the position a club will be in the league table. However with an Adjusted R2 of 0.453 Average Odds only roughly helps to get you 50% of the way to an accurate prediction.