Macro: Sandcastle economics
Invest wisely in Q3 2024: Discover SaxoStrats' insights on navigating a stable yet fragile global economy.
Chief Investment Strategist
Summary: The 2022 World Cup is well under way and as a result we have updated our tournament predictions based on adjusting our squad dynamic adjustment factor the Elo ratings we use in our monte carlo simulation. The new predictions show that the Netherlands is still a favourite to win the 2022 World Cup, but Portugal has climbed to the second place after two won games. In this update we also go through our thoughts behind adjusting the Elo ratings and why these adjustments have proved to be correct based on the results so far in the tournament.
Netherlands is still favourite with Portugal running second
On 18 November we published our initial predictions on the 2022 FIFA World Cup making the Netherlands the favourite to win the World Cup instead of the consensus bookmaker choice of Brazil and Argentina. As described in our initial note and in more depth here we based on our predictions on Elo ratings with five adjustment factors added and then simulated 10,000 times using monte carlo simulation. Our non-consensus predictions were met with ridicule, but as we pointed out on social media, football and also our predictions are fun and games with a grain of seriousness of course. A more important point as well, is that in order to have markets and people betting on events traders and gamblers must have predictions that diverge from consensus.
Four days later on 22 November we made our first updated predictions which had the Netherlands maintaining its role as lead favourite due to its win in the first round. Argentina’s defeat to Saudi Arabia immediately reduced their win probability to just 9% down from 22% consensus before the tournament and our 14% win probability before the tournament based on our Elo adjustments, which we elaborate on below. Our first update was short and only made available on Twitter, but in this second update we not only show the new updated probabilities, we also go into more detail on our Elo adjustments and what we have done to them since the start of the tournament.
Based on the results as of yesterday our new model simulation is still having the Netherlands as the top favourite to win the tournament, but Portugal has jumped to second place with Spain trailing at the third place. Besides Portugal seeing a big jump in its win probability, Denmark has seen a massive drop from 11% in our initial set of predictions to just 2.8% due to only getting a draw and losing one in the first two matches below well below the indicated Elo level ahead of the tournament.
Adjustments were justified
We make five adjustments to the Elo rating: Elo momentum (1-year), Misery Index, European cup win bias, shrinkage, and squad dynamics. In this second update we have only adjusted our squad dynamic factor to reflect what we have seen in the matches so far. When the group stage is done we will update the current Elo rating and the Elo momentum in addition to a new squad dynamic adjustment before we run the 10,000 simulations again. This simulation will lead to our third update and predictions for the knockout stage. Following that prediction we will update our predictions at each new round of the knockout stage until the day before the final.
As said we were criticised for our non-consensus predictions and especially our Danish bias. While the adjustment to the Danish team created a too optimistic prediction some of our other adjustments have proved to be validated. Our downward adjustments to Argentina have proved correct given the two matches and if we remove Messi, the Argentina team is quite mediocre. We also lifted Spain and Portugal in our adjustments which have proved correct. We were harsh on Belgium in our squad dynamic adjustment giving the team the largest reduction in Elo score based on this factor; this proved correct as the team has slipped from a strong outsider to a team nobody believes in. We were too tough on France due to their injuries ahead of the tournament and we have removed their penalty in the squad dynamic adjustment.
Financial Times brings today an article World Cup briefing: Ghana win shows flaws in Fifa rankings in which the authors discuss the flaws of ratings as Ghana has clearly been underestimated. This is exactly why we added a shrinkage factor to our adjustments. The shrinkage factor shrinks Elo ratings closer to the average making the teams more equal, which fit better with the high variance nature of the game. Also remember that the Elo rating convergences to the true skill level after around 30 games, but as a result of football being a high variance game the prediction error will be quite sizeable over just three matches and thus Elo ratings will over- and underestimated the true strength and weakness of teams.