What About the Elo Rankings?

 

While We Global Football has looked at the differences between our ranks and the FIFA ranks, we are going to tackle a new challenge. Widely accepted as a reasonable alternative, the Elo ratings system has a pretty decent following in the global football community.

There are some major benefits to the Elo system (which are also incorporated into our system) which FIFA does not take into account. They are margin of victory and a home advantage factor. We absolutely agree that they should both be included, as they were the two main points we made against FIFA’s ranking system. Similarly, the Elo system also does not apply continental bias, which is a huge plus. There is no need to give teams a bonus for where they are located on a map.

There are, however, two drastic differences between the Elo system and the WGF system.

1. The Elo system applies a different weighting to different competitions, while the WGF system does not.

While this is something we considered, we ultimately decided this was a factor we wanted to exclude. I get why Elo and FIFA would want to include this factor, but we don’t see enough justification to include it. Brazil gets punished in the FIFA rankings because they are only playing “friendlies”. We feel that you need to operate under the assumption that the best teams don’t always win. While a win certainly improves your standing more than a loss does, every game is independent.

Unfortunately, in both the FIFA rankings and Elo rankings, the effect of one loss can be compounded severely. A prime example of this right now is the United Arab Emirates. They had a poor qualifying campaign back in the fall of 2011, and thus were eliminated from World Cup qualifying. Because all of their matches currently get given far less weight, they are handicapped for YEARS in the rankings by a couple bad games. We don’t like this. Since the start of 2012, UAE is 15-1-1, with the only loss coming by 1 goal at Japan. This is an unbelievable run.

In reality, the only disadvantage of not advancing to further qualifying rounds is that you may not face the top competition, and we take strength of schedule into account. The only legitimate argument we see to including a competition factor is the fact that teams do not always field their best teams for friendlies. While this is the case in some instances, the small portion of games it affects, we do not believe warrants an additional adjustment. In a predictive model, you must take the good with the bad.

2. The Elo system starts with a 0 basis which begins in the early 1900s.

For every ranking system there is a starting point. There is a game #1 which gets factored into every rankings system. The first game factored into the Elo ratings took place more than a CENTURY ago. While yes, everyone that played in those games is no longer living, the fact remains that if you remove any of those games from the Elo calculations, your current rankings will not be the same.

Obviously over time the impact of such changes is very, very minimal, but it’s there. The reason for this is that every game’s Elo rating change is based off of the previous match’s standings. To get back to the current point, you need to account for over 36,000 matches. The Elo system is predicated on quality results after around 30 matches. While statistically this makes sense, teams play so infrequently that often times when a team converges on 30 games, the roster is entirely different. This is like carrying over results from a prior season in sports where the roster has a huge makeover.

We found that a statistically meaningful cutoff point was 18 months. There are enough data points to produce quality results, but it is not too long where the rosters have had a dramatic makeover. The strategy we use is a rolling 18 month schedule, so it is possible that the ratings could change tomorrow without a game being played. While this may seem unusual, the alternative is including every game ever played. FIFA uses a rolling 48 month schedule, which takes the worst of both systems.

One other small aspect of Elo is that it is a zero-sum game. That being, the number of points a team gains by winning, the other team loses an exact equal amount. This may be a more statistically explainable method, but it is not necessarily a benefit to a rankings system.

The WGF System
As we have explained previously, we take 5 factors into account:
– Where was the game played?
– What was the result?
– What was the margin of victory?
– What was the quality of your opponent?
– When was the match played?

Our system is what we call a dynamic, recursive system. We weight each of these elements differently, and recursively run calculations until the results converge on a ranking. Likewise, as the date changes, so does the timeframe of games we include.
In simple terms, what we say is: Here is a set of games in a specific time period. Here are the scores and teams from those games. Rank the teams.

A specific example of where our rankings differ from FIFA and Elo rankings is on Italy. Italy is consistently ranked a top 10 team by both systems. They reached the finals of the Euro tournament last summer and as a result played 6 matches that carried with them an extremely heavy weight. Additionally, they have accumulated a significant amount of points from historical matches which caused them to be near the top of the standings. In that EURO tournament, they won 2 games, had a draw in 3, and lost 1. Their goal differential in these games was -1. This does not take into account their 3-0 loss to Russia in a warm-up game.

From our perspective, in these 7 games Italy gained 9 of a possible 21 points. and they had an average goal differential of -0.57 despite playing very tough competition. While these results have been offset by stronger performance in other matches, the sum of their numbers leaves them at #32 in our rankings. In the past 18 months, Italy has played 16 games against teams in the top 100 of our rankings. They’ve only come away with a win in 4 of those games.

In conclusion, every system has their positives and negatives. The Elo System is a much more reliable method than the FIFA rankings, but it is also not without its flaws. We have aimed to create something that removes as much bias as possible. The only user inputs are the weighting of our factors. We hope that in the future, the WGF Rankings can become a more respected and accurate rankings system than any other out there.