Does the Audi Player Index have statistical relevance? We will continue the experiment by sharing predictions for the Week 34 MLS matches.
In my earlier post, “Audi Player Index: Entertainment or Science?”, I questioned whether the API ratings could be used for more than just a fans enjoyment. What if this is the first ripple in a wave of artificial intelligence ready to flood Major League Soccer?
My first attempt to answer the question was an experiment using a team’s total overall points through Week 33. I wanted to see if the data could successfully predict the 2018 playoff teams.
Let’s review the results. While the index was mostly correct predicting the teams, it was mostly wrong when forecasting their place in the table. So, it could be valid or it could have been blind luck. This question will need more research.
More from MLS Multiplex
- Javier Milei Elected in Argentina: Potential Impacts on MLS and Signings of Argentine Players
- Orlando City and New York City FC in the Battle for Matías Arezo; Grêmio Enters Negotiations! Who Will Come Out on Top?
- USA, Honduras, Panama, and Canada Close in on a Spot in the 2024 Copa America
- De Gea Turns Down Al-Nassr’s Lucrative Offer: Speculation Points to Possible Reunion with Messi at Inter Miami
- Messi’s Magnetic Impact in the United States
In our next experiment, I am going to use the API ratings to predict the winner of each game in Week 34. Before I begin, a few points.
The first, I hate ties — drop them from the sport. As such, my predictions will be “Draw, No Bet”.
In other words, if the game in question ends in a tie, it will be as if it never happened. For example:
- LAFC vs. Houston, I pick LAFC to win
- If LAFC does win my record will be 1-0
- If LAFC and Houston tie, my record will be 0-0, since the game technically never occurred
- If Houston were to win, my record will be 0-1.
The second point. I have attempted to interpret the API ratings using various formulas in Google Sheets, or as I refer to them, complex algorithms. I would like to think this will improve accuracy. Undoubtedly, others could look at the same data and have a completely different perspective.
I’m going to adjust for this variance by calling my conclusions, “Soccer CPU”.
If someone comes along with a different approach they can name theirs too. In the meantime, I am taking suggestions on how to make CPU a legitimate acronym.
Now that we’ve got that out-of-the-way, onto the predictions:
- LAFC vs. Houston – LAFC
- Minnesota FC vs. Colorado – Minnesota FC
- D.C. United vs. FC Dallas – D.C. United
- New England vs. Orlando City SC – Orlando City SC
Four games is not going be enough to draw a final conclusion, obviously, but it will start to offer some insight.
Check back next week to see if we are any closer to answering this intriguing question.