I, like many have been deeply frustrated by the highly subjective (and often rude) assessments of player's performance by fans of the USMNT, so I set out to create a simple and objective way of evaluating how players have performed for the USMNT. I built from the theory that scoring more goals than your opponent is always the objective of the team on the pitch, regardless of game state, so the model simply summarizes who is on the field when we score (+1) and who is on the field when our opponents score (-1). Then I take the net scores for each player and normalize to 90 minutes (data from transfermarkt.com).
In graphical form we have Total Minutes Played across the horizontal and Net Goals per 90 minutes on the vertical. As always it is good to be in the upper right corner, but in this case you can think of it as, the player controls how they do on the vertical (net goals per 90), while the coach controls how much they are on the pitch (Total Minutes Played).
Here are the results so far through the first five World Cup Qualifying games for the US Men’s National Team.
As you can see, Cristian Roldan is leading the pack by a significant margin in terms of net goals per 90 produced by the team when he is on the pitch. Also in this neighborhood are Ricardo Pepi and Antonee Robinson. On the other side, Brooks, Sargent, Moore, Zardes, and Bello have been in when things haven’t gone as well.
It is worth noting there are also a lot of very big salaries playing for very big teams that have yet to be on the field when the team have outperformed their opponents. The more minutes they have without tangible results, the more valid the conclusion.
Which brings up the obvious point. Is this really a reasonable way to look at a player’s performance? For only a small number of games I think it is fair to recognize that the small sample size will cause wild swings in results. But, the more games we analyze the more valid the results are. In the end the way we evaluate success for the team is scoring more goals then the opponent, so why shouldn’t we use the same yardstick for evaluating players on that team?
With that point in mind, we can expand our dataset by looking at the other competitive games played by the USMNT this year. Here is the summary for all USMNT games from the Nations League Semi Finals/Final, the Gold Cup, and World Cup Qualifying.
Interestingly the results still remain the similar, with Roldan, Pepi and Williamson looking like players that are helping the team perform well, while great players, from big teams haven’t been present for much success yet.
You will notice that the average for the team is to outperform our opponents by about 1.1 goal per 90 minutes, which makes sense as we have largely been winning our matches during the games evaluated.
OK. There is your data to think about as you consider who the US should start in our game against Costa Rica. If you choose to share your thoughts – please be nice.