Mathematician who predicted past three FIFA Cup Winners on why he sees Netherlands winning this year

Joachim Klement isn’t a household name, but this German mathematician and economist- model had correctly predicted the previous three tournament winners and he now expects Netherlands to win the 2026 edition of the FIFA World Cup.

In an exclusive interaction with Fortune India, Klement, currently dabbling as a sell-side strategist (with a focus on European equities) and ESG strategy at the UK-based Panmure Liberum, a leading independent investment bank, had correctly picked the previous three tournament winners: Germany in 2014, France in 2018 and Argentina in 2022.

Klement, who holds a Masters in mathematics and economics & finance, had for the first time in 2014 set out to demonstrate why predicting winners was practically impossible and was pleasantly “shocked” that in his very first attempt, the model that he built got it right, not just once but thrice!

Klement tells Fortune India that the idea behind building the statistical model was to show that most of what economists predict is really hard and getting the forecast right relies on a lot of luck. Edited excerpts:

Why did you start this World Cup forecasting exercise in the first place? You’ve mentioned it was meant to be a critique of economists’ overconfidence in prediction. Can you walk us through the context that led you to create the model?

In 2014, I thought people took forecasts from economists and financial experts too literally and ignored the many things that can go wrong. Since the FIFA World Cup was around the corner and I am a big football fan, I thought I could use the World Cup as an obvious example where an economist predicts something that he clearly has no special knowledge to show that most of what we do is really hard and getting the forecast right relies on a lot of luck. And then - unfortunately - I got lucky and got my forecast came true.

When Germany won in 2014 as your model had predicted, what was your immediate reaction? Why did you continue with the model?

My immediate reaction was: ‘I got lucky, so let’s do it again next time, so I can show people that it was just luck.’ Unfortunately, I got lucky three times in a row, so I must continue until I get it wrong!

You've said systemic factors account for roughly 50% prediction power, with the other 50% being luck. Is the equal split empirical or just theoretical?

To be precise, the four factors explain 55% of the variation in international football matches, with luck being 45%. But to keep it simple, let’s just say it is 50/50.

You base your thesis on five systemic factors: population, national wealth, climate, and FIFA world rankings. Can you break down each of these five metrics, and how much predictive weight does each carry?

The most predictive factor is the current FIFA World Ranking, since this measures how good the current generation of players is. The second most important factor is GDP per capita because to get a really good team, you need to invest in players and coaches from a young age. And you can only do that systematically if a country has the money to do that. The population and the temperature are less important. Think about it this way: if population size were that important, India and China would dominate world football.

How do you quantify "national wealth" for a football context? Are you using GDP per capita, total GDP, investment in youth academies, or something else?

I use GDP per capita, but I also use GDP per capita squared. I do this because if a country is too poor, it doesn’t have the money to develop young talented players. But if a country becomes too rich, the kids tend to play video games or tennis (or many other, more expensive sports) rather than football.

What are the additional parameters that you have built into the model to make it robust and is it only modelled to predict football world cups; can these be extended to club matches?

This model only works for World Cups. It doesn’t even work for continental cups, let alone club competitions. It is a fun model based on an academic paper from 2002, but I wouldn’t bet my savings on its predictions.

You mention wanting to give people "a little distraction from all the bad stuff going on in the world," but does economics have a bearing on the model’s outcomes. For instance, could a recession-hit nation over-perform or under-perform relative to your predictions?

I have never seen any evidence that the current economic situation in a country makes any difference to the performance on the pitch. Indeed, the evidence of what determines success in the World Cup is extremely poor. It is mostly guesswork, which is what I wanted to show with my model in the first place.

Let’s talk about your predictions for the world cup this time. Which matches will throw up upsets, and which ones will be one-sided?

I don’t make predictions about individual matches in the group stage, but I do predict that we might get a match between Brazil and Japan in the last 32 and that Japan will win against Brazil. That sounds absurd to me at first, because Brazil is the world’s most successful team in history. Yet, the current generation of players is not at their best, while Japan has a very good team at the moment. Add to that that Japan beat Germany at the World Cup four years ago and Brazil in a friendly match in October 2025, and it doesn’t sound so outrageous anymore that Japan could beat Brazil if they meet in the knockout stages. Of course, whether they meet in the first place depends on how Brazil and Japan place in their group, but thus far they remain on a collision course.

Who according to you will be the four semifinalists in this year’s edition, and why?

If everything goes to plan (and it never does), then the top favourites France and Spain, together with Germany and the Netherlands, will be in one half of the knockout stages while England, Argentina, Portugal and Brazil will be in the other. That means that y the time we come to the semifinals, one of the two top favourites (France or Spain) will already be eliminated. My model expects that Spain and the Netherlands will make it to the semifinals, while the other semifinal will be played between England and Argentina. Arguably, if Japan really wins against Brazil, that would make life easy for England and open up a path to the semifinal for them, while Portugal has to beat Argentina in the quarterfinals.

But my experience from the last three World Cups is that I never get all four semifinalists right. On average, I get two or three semifinalists and five out of eight teams in the quarter finals right. After all, every tournament has some major upsets like Morocco getting into the semifinals four years ago and Croatia getting into the final in 2018.

Which teams will face off in the final, and who will lift the cup?

I expect that the Netherlands will beat Spain to reach the final and Portugal will beat England. The Netherlands beating Spain is a big call because Spain is the favourite to win against the Netherlands. But as I said above, 45% of the outcome of every match is luck, and my model simply predicts that the Netherlands - after being unlucky so many times - will be lucky enough to make it into the final against Spain.

In the final, the Netherlands are favourites to win against Portugal, which means the Dutch will become world champions for the first time.

Will the final go to the penalties like it happened in 2022 or an exciting finish?

No idea, my model cannot make predictions about the score of any match.

Which footballer will make history in the cup? Which nation will be a team to watch out for in the future?

Again, my model cannot make any of these predictions. But as I said, I think Japan is a team to reckon with, and what I have seen in this World Cup so far, they can go far in the tournament.

If your model is correct again in 2026, what will you do? Keep predicting, or retire while you're ahead to preserve the integrity of your original message about forecasting hubris?

I will keep doing this whether I am right or wrong. This is some good entertainment and given the many wars and problems in the world, I think we all can use a bit of entertainment and distraction sometimes.

Could you build a similar model to predict interest rate movements or economic growth outcomes? Has the advent of AI helping you build more robustness in the current model? Are you looking to commercialize the platform, or will it stay as a hobby project?

The football predictions will always be a hobby that I do once every four years. In my day job, I am an economist and equity trader at an investment bank. Hence, I have to build models about the economy and the stock market all the time. AI doesn’t really help there, to be honest. I find that AI is good at some things, but as I show again and again on my free Substack (https://klementoninvesting.substack.com/), AI is terrible at forecasting even such simple things like inflation, let alone interest rates. Maybe AI will get better in the future, but at the moment, it isn’t good enough.