Comparison of MLS is Back results with FiveThirtyEight predi

[NolanColton]

With much being said about FiveThirtyEight predictions going right out the window through the entirety of MLS is Back, I decided to actually crunch the numbers and see what the group stage standings and knockout bracket would actually look like if FiveThirtyEight was accurate.

I'm certainly not attempting to bash FiveThirtyEight's supposed inaccuracy in predictions. If anything, we should recognize how difficult it is to predict a tournament of completely unprecedented circumstances, derived from an already unpredictable league with forced parity and with newcomers almost every year. With so many wildcards being introduced into the data even every regular season, let alone in a one-off pandemic-response cup, it's a wonder there's any accuracy at all. That being said, here is my clumsy and simplistic methodology in my interpretation of the data, and the inherent limitations to my approach:

I used FiveThirtyEight's Projected Score values (not to be confused with Expected Goals or Adjusted Goals) for each match, rounded to the nearest whole value, to represent each match's predicted score - for example, the opener between Orlando and Miami had a projected score of 1.14 for Orlando and 1.6 for Miami; rounded to the nearest whole number this gives a predicted score of Orlando 1:2 Miami. The Projected Score values are quite conservative in scope, and so the majority of teams score between 1 or 2 goals each game, making for some individually believable but very repetitive results. In turn the rounding is a very reductive way of analyzing the already constrained data and does not take into account any nuance of the Projected Score values, i.e. 0.6 and 1.4 all end up as 1 goal. Ultimately the majority of matches having a predicted scoreline of 2:1 or 1:2 gives an okay representation of points but not enough variance in details such as goal difference to really separate teams any more than that.