Ranking of the Football Rankings – 2016
International football rankings certainly have their detractors. The FIFA Rankings, in particular, take a lot of grief from many around the world. We Global Football started with a list of rankings and evolved. We wanted to create something that best represented teams’ strength at the current point in time. But different football rankings have different purposes. The FIFA Rankings are not necessarily representative of what you’d expect to see in a match. and therefore are not intended to be predictive. More simply, if Team A is ranked higher than Team B, the FIFA Rankings aren’t saying Team A would be expected to beat Team B. Only, based on past performance, they rate Team A higher than Team B.
As is commonly known, the FIFA Rankings can be gamed. Most notably, Romania reached Pot 1 of the World Cup Qualifying draw due to smartly navigating the rankings system. We’re not here to rail the rankings, but we just wanted to point out that in no way are the FIFA Rankings a good representation of predicting match outcomes. And even more, there is no tangible way to measure the expected result of a football match using those rankings.
Predictive Football Rankings
What we wanted to do is take ONLY the rankings which are clear in their motives to be predictive. There are two types of ranking systems that we’ve found. One, which is what WGF does, is have a predicted goal difference among two teams. The second, which is what the Elo system does, is produce win probabilities among two teams. The challenge here is to coordinate the two methods to get everything on the same basis.
Thankfully, we have a formula that converts a margin of victory to win/draw/loss probabilities. Using that formula, we can reverse engineer the systems that already have the probabilities to get an expected margin of victory. While it is confusing, we do have 5 ranking systems that can be compared against each other using the authors’ intent. The ranking systems are:
We Global Football – View Here
You’re here. These are our rankings. We use 18 months of matches and don’t weight competitions differently. As a result, we’re a bit off from the consensus. And that’s OK. As we stated above, you can take our rankings and get an expected margin of victory when comparing two teams.
Chance de Gol – View Here
This is a Brazilian ranking system. Marcelo from Chance de Gol confirmed to us that their rankings can be used similarly and result in an expected margin of victory when comparing two teams.
The Power Rank – View Here
Ed Feng also confirmed to us that his rankings can be used to compare goal difference for two teams. Located in the USA, Ed uses four years of matches and weights matches differently by their importance.
CTR – View Here
Located in Germany, Norbert Jäger uses the second method of rankings. His rankings use a points based system which can be converted into win/draw/loss probabilities using a formula he has provided. And as we indicated, we can use our formula to get an expected margin of victory. While this is not exact, we think it is the best representation of comparing systems fairly.
Elo – View Here
The Elo ratings are commonly used in football predictions. While there is no true “owner” of the ratings, Kirill currently maintains the site. The rankings take into account matches going back over 100 years. Their formula for win/draw/loss probabilities is different than CTR, but they use the same “points to probability” method.
And that’s it! There are other systems out there which were not included due to either being not predictive (FIFA), unsure about intent/calculation method (Mondfoot, IntMark, Massey), outdated (Maas), or just not enough information (SPI). We’ll be happy to include those if there is a direct way to get a prediction from the rankings, but for the time being we’re only going to look at these five.
Comparison
While we embark on this journey, we just want to say that we view all of these systems as colleagues. People have been generous in responding to us and making their data available. However this turns out, we respect everyone that has put their time into creating something people can use.
Also, while there are over 200 matches to be played between now and the end of EURO 2016, this is a small sample size. The results don’t necessarily mean that one system is better or worse. But it is a decent enough sample where hopefully we can generate some findings.
Here are the rules we are going to abide by:
- Results will be presented by competition type. We will use 3 different groupings: Friendlies, Qualifiers, Finals. EURO and Copa America will fall into the finals category. AFCON, World Cup, Asian Cup, Caribbean Cup qualifiers will be grouped together. All other matches will be grouped together.
- We will do our best to get the most accurate projection. We’re not going to take rankings as of this instant and use them as the prediction for EURO and Copa America. Starting on 5/28, with OFC World Cup Qualifiers, we will update our database of each system daily to get the latest rankings.
- For EURO and Copa America, we will also create a “tournament preview”. We will freeze the rankings at the beginning of those tournaments to generate a full prediction. This will not count towards the overall comparison, but it will be a separate preview.
We want to be as transparent as possible, and we think this the fairest way to do it. What we’ll ultimately do is take the average absolute difference between the projection and the result to get an “average error” by ranking system. The system with the lowest average will be the most accurate.
Projections
Without further ado, here is the first batch of May friendlies. When teams are at home, we applied a home advantage factor. We’ll keep the projections updated under the “Match Predictions” heading at the top of the page.
