Macro: Sandcastle economics
Invest wisely in Q3 2024: Discover SaxoStrats' insights on navigating a stable yet fragile global economy.
Chief Investment Strategist
Summary: Like the galloping inflation in the world today another inflation has gripped football. Everywhere we look, the Internet is filled with predictions of which team will win the 2022 FIFA World Cup from universities, bookmakers, hobbyists and banks. Many of these predictions have one common thread. They are all way too close to consensus. This is neither a surprise nor interesting given they all use Elo ratings. Our team likes to be contrarian and as such we have jumped into the quantitative game of monte carlo simulations using Elo ratings but we adjust them based on five factors leading to surprising non-consensus predictions, namely that the Netherlands has the highest probability of winning the tournament.
Netherlands is the dark horse and France has zero chance
The 2022 FIFA World Cup begins on Sunday and the bookmakers have Brazil as the favourite to win the tournament with Argentina and France with the second and third lowest odds. Based on our monte carlo simulation on the tournament tree, Elo ratings and adjustments to the Elo ratings we come up with non-consensus views that the Netherlands will win the tournament with Argentina and Spain as our second and third most likely candidate. But even more controversial, we have Brazil as only the seventh more likely team to win the World Cup 2022 compared to being the bookmakers favourite and we France with almost zero probability of winning the tournament. We will update our predictions as the tournament progresses.
Elo rating and consensus
The Internet is overflowing with monte carlo simulations of the World Cup 2022 using Elo ratings which is a method for calculating the relative skill levels of players in zero-sum games such as chess or football. As the Elo rating system has shown to have good predictive capabilities it is a key input for many football bookmakers and as a result a monte carlo simulations using Elo ratings will yield results very close to the current average odds pointing towards Brazil and Argentina being the clear favourities to lift the World Cup trophy on 18 December 2022.
This observation is neither surprising nor interesting. Like in financial markets the alpha is not in going with consensus but going against consensus when expectations are set wrong, which is often the case given we have imperfect information. Take the World Cup 2018 odds. Here, the final match between France and Croatia was between two teams which prior to the tournament had the fourth and tenth lowest odds. In other words the bookmakers’ favourites did not reach the final.
Adjusting the Elo rating
In the Saxo Strats team we also like to be contrarian as we have been since late 2020 on our call that inflation was not going to be transitory. In order to move away from consensus we are adjusting the current Elo ratings using five factors: Elo momentum, the misery index, recent European cup win bias, shrinkage, and recent squad dynamics.
Elo momentum
Elo momentum describes the 1-year change in the Elo rating for each team and indicates which teams have done well or terrible going into the tournament. Momentum effects exist in financial markets and they do also exist in sports as a winning streak builds self-esteem in the team and perseverance. The adjustment reduces the chance of winning the most for France and England and increases it the most for Netherlands and Costa Rica.
Misery Index
The misery index is our smallest adjustment but is wild card that plays on an assumption that a country that is undergoing the biggest misery (the sum of the unemployment rate and the inflation rate) maybe has a bigger desire to make their nation proud. Argentina and Iran benefit the most from this adjustment while Switzerland and Japan benefit the least.
European cup win bias
European teams have done better than non-European teams competing in the recent world cups, which is why we have included a European bias for our world cup predictions. European teams have won the last four world cups. The World Cup Final in 2018 was played by two European teams, while the semifinals were played by four European teams. The favoritism has shifted from south America to Europe. In 2014, Germany won the world cup after defeating Brazil 7-1 in the semifinal and Argentina 1-0 in the championship game. As in Bayesian statistics one has to update the prior with the most recent information and in the case of football that means the posterior probability leans in favour of European teams.
Shrinkage
In statistic we observe sampling variation and that is why we a concept such as shrinkage is applied which “shrinks” values closer toward the mean. This means that teams such as Brazil and Argentina with the highest Elo ratings will be “punished” more than an average team. In layman’s terms this could be translated as a high Elo ratings lead to low odds and very high expectations which can weigh on a team – they can only fail against expectations.
Squad dynamics
We have included a factor called squad dynamics which takes the actual world cup squad into consideration and not just previous results such as injuries, players performance in their clubs etc. For instance, we have given France a big readjustment for the worse, because of the squad they are bringing to the World Cup. France is the current world champions, but coming off a Nations League tournament where they fought for survival in League A, group 1. After playing six games during 2022 and only winning one game, they are coming into the world cup with a squad without profiles like Paul Pogba, N’golo Kante, Boubacar Kamara, Presnel Kimpembe and the talent Christopher Nkunku.
When we add all the adjustments to the current Elo rating and run 10,000 simulations we get the following adjusted probabilities.
How does the simulations work?
All countries are ranked via their Elo rating, which is a relative way of measuring the skill level of teams/players within a certain field such as chess, baseball or soccer. The difference in the ratings has in soccer proven to be a fair predictor for the outcome of a match, with a higher probability of the high Elo rating to win.
We have mixed the Elo ratings of the participating world cup teams with betting odds in order to find a mapping from the Elo ratings to the probability outcomes for each world cup match. We artificially simulate that the world cup is played 10,000 times, using these probabilities to determine the outcome of every single match. The result is 10,000 different equally-probable outcomes for the world cup. And when averaging over all scenarios, we achieve the probability of each team winning the world cup.
As an example, Denmark (Elo 1971) will be facing Tunesia (Elo 1707) on Tuesday. According to our model gives Denmark a 59 % of winning, 25 % of a draw and only 16 % chance of Tunisia winning.
Note that the probabilities and the model behind the calculations should not be interpreted as any betting odds, but it should merely be considered as a fun exercise around the World Cup.