By Thomas Negron
Later tonight, the United States and Colombia will kick off the Copa America Centenario, a month-long international tournament which will take place across the United States. While the Copa America is typically a South American tournament, this edition, which is a celebration of the 100th anniversary of the first Copa America tournament, will feature teams from North and Central America as well.
A simple metric that is often used to rank national teams or predict results to international matches is Elo ratings. While these ratings were originally developed for chess, they serve as a decent representation of the true strength of a team. They work especially well with national teams, as the player pool remains fairly constant from year-to-year, as opposed to club teams where transfers and manager changes are more frequent.
I used data from World Football Elo Ratings, which has Elo ratings for every national team in the world, compiled by Advance Satellite Consulting. They also have explanations on how to convert Elo ratings into game predictions. The important thing to note here is home-field advantage. They have found that home-field in international soccer is equivalent to 100 extra Elo points. For this tournament, I gave the United States full home-field advantage and Mexico half of the value of home-field advantage (an extra 50 Elo points). I decided to give Mexico this boost because the Mexican national team always draws a big crowd in the US, especially when they are playing close to the US-Mexico border, which is the case in the tournament, as their group stage games are in Phoenix, Pasadena and Houston.
With that in mind, we can move into the actual predictions. To do these predictions, I ran 10,000 simulations for each group, based on the match probabilities given by the Elo ratings. These simulations provided probabilities for a team’s chance of winning their group and their chance to come second in their group, which advances them to the knockout rounds. From here, I used win probabilities between all the teams and the structure of the tournament to show the chances a team would advance to a certain round. For instance, the quarterfinal matchups have the winner of Group A playing the runner-up of Group B and the winner of Group B playing the runner-up of Group A. USA is in Group A, so their probability of making the semifinals is based on their own probability of advancing, the probability of advancing for each team within Group B, and the USA’s probability of beating each team from Group B.
For USMNT fans, it is definitely going to be an interesting group stage, with this model giving them about a 56% chance to advance to the knockout rounds. Group A is the most even group across the board, as every team in the group has an Elo rating greater than 1700 (no other group has every team with an Elo rating above 1600) as well as being the group with the highest average Elo rating, narrowly edging out Group D. The Ecuador-Peru match in Group B looks like it could be the decider for who goes through to the quarterfinals, while Groups C and D each have two strong favorites for the top two spots, and the question is more which teams will end the group stage on top.
This model marks Argentina as the favorites to win the tournament. Argentina, who finished runners-up in last year’s Copa America, is looking to win the tournament for the first time since 1993. Brazil and Mexico should also be considered favorites to lift the trophy at the end of the tournament, with Colombia, Uruguay, and Chile being good dark horse picks. This model also gives the USA slim chances of winning it all, as the team has undergone a poor year as they attempt to transition to a new generation of players.
It is important to note that Elo ratings still miss a lot of what judges the true quality of the team. For instance, these predictions do not take into account that Neymar (Brazil) will sit out in order to play in the Olympics later this summer, that Messi (Argentina) may not be 100% after suffering a back injury, or that Keylor Navas (Costa Rica’s star goalkeeper) will miss the tournament entirely. Brazil’s and Costa Rica’s chances are probably overstated in the model because it assumed their stars would be present. It also is slow to adjust based on players joining the national team, so if you are on the Darlington Nagbe or Christian Pulisic bandwagon, you may think this underestimates the USA’s chances.
While these pre-tournament predictions are interesting to read, the best part of using Elo ratings for predictions is we can put a number on how a game affects a team’s probabilities. I will be updating the model after each round. In these articles I will be highlighting any big changes in a team’s Elo or probability of advancing to further rounds and pointing out the most important games of each round.
Finally, as a bit of a teaser, let’s take a look at the USA-Colombia game that will happen tonight. A win for the US, and things are looking good, as our model would put them at 64.15% to win the group and 88.43% to advance to the quarterfinals. A loss, on the other hand, would put them in an early hole. Winning the group becomes very uncertain, at 3.61%, while their odds of advancing also take a hit, dropping to 41.28%. In a group where second place seems to be set on a crash course with Brazil, winning the group could prove vital to a deeper run in the tournament. A win would improve the US’s odds of making the semifinals by about 8.5 percentage points, to around 31%. With many US National Team supporters growing weary of Jürgen Klinsmann, there is no question that there is a lot at stake tonight when the Copa America kicks off.