by Dr Adrian Worton
We have created a World Cup simulator. This uses bookmaker odds to generate a result for the upcoming World Cup. However, no two simulations are likely to be the same (there are three outcomes for 64 matches, making this a wheat and chessboard scenario).
Therefore, in order to get an idea of which teams are the strongest according to our model, we need to run the simulator multiple times and count the results. That's what we've done, and here we'll give the results.
We have created a World Cup simulator. This uses bookmaker odds to generate a result for the upcoming World Cup. However, no two simulations are likely to be the same (there are three outcomes for 64 matches, making this a wheat and chessboard scenario).
Therefore, in order to get an idea of which teams are the strongest according to our model, we need to run the simulator multiple times and count the results. That's what we've done, and here we'll give the results.
To provide our results, we ran the simulator 3000 times (three times the amount we did four years ago!) and totalled each team's final result, which could be one of eight possibilities:
The brackets give the code with which we'll refer to these outcomes later on.
To win the tournament
Let's start by seeing the breakdown of who is likely to win the tournament:
- Winner (1st)
- Runner-up (2nd)
- Third place (3rd)
- Fourth place (4th)
- Quarter-finalist (QF)
- Second round (SR)
- Third in group (X3)
- Fourth in group (X4)
The brackets give the code with which we'll refer to these outcomes later on.
To win the tournament
Let's start by seeing the breakdown of who is likely to win the tournament:
You would probably expect this to follow the coefficients we assigned each team. However, there are a couple of differences - Belgium have leapfrogged Argentina, whilst Uruguay are ahead of Portugal. It is also interesting to see Brazil sneak ahead of Germany despite identical coefficients.
Anyway, the pattern of the teams is largely what we would have expected. I would say that I'm surprised the 'big four' aren't as far away from the pack as they are. Really, only Brazil, Germany, France and Spain have squads full of big tournament winners. This said, the carnage of Euro 2016 showed what can happen in the knockout rounds.
Overall performance
Now let's show the full results for all 32 teams:
Anyway, the pattern of the teams is largely what we would have expected. I would say that I'm surprised the 'big four' aren't as far away from the pack as they are. Really, only Brazil, Germany, France and Spain have squads full of big tournament winners. This said, the carnage of Euro 2016 showed what can happen in the knockout rounds.
Overall performance
Now let's show the full results for all 32 teams:
There will be some unexpected kinks in the graph. For example, why are Uruguay much more likely than anyone near them to go out in the second round? This is because of the way the draw works - they are highly likely to play Spain or Portugal in that round.
Anyway, feel free to pore through the table to help you understand each team's chances. If you have taken part in a World Cup sweepstake you may wish to work out your chances of getting any money. I have Russia, so I'm not making plans for any winnings just yet...
Summary
Our simulator is a fun tool on its own, but when aggregated it can be very interesting as it allows us to see the long-term predictions of the model.
Please note, this is a list of probabilities. So we are simply saying that it is unlikely Costa Rica will win the whole tournament. If they do manage it, this just gives you a sense of how huge the achievement is.
It is interesting that Gracenote published their probability-based predictions recently, with wildly different suggestions to our's. Personally, I think their predictions are a bit wacky. This is because they are using the official FIFA rankings - which sides have been able to game (see Poland becoming a ranked side this year), and throws up weird results like Peru being above England.
If you are going to use rankings to produce predictions, at least use the Elo Ratings, which have been shown to be the strongest predictor of team performance.
Anyway, rant over... in summary, our predictions are much stronger than Gracenote's.
Anyway, feel free to pore through the table to help you understand each team's chances. If you have taken part in a World Cup sweepstake you may wish to work out your chances of getting any money. I have Russia, so I'm not making plans for any winnings just yet...
Summary
Our simulator is a fun tool on its own, but when aggregated it can be very interesting as it allows us to see the long-term predictions of the model.
Please note, this is a list of probabilities. So we are simply saying that it is unlikely Costa Rica will win the whole tournament. If they do manage it, this just gives you a sense of how huge the achievement is.
It is interesting that Gracenote published their probability-based predictions recently, with wildly different suggestions to our's. Personally, I think their predictions are a bit wacky. This is because they are using the official FIFA rankings - which sides have been able to game (see Poland becoming a ranked side this year), and throws up weird results like Peru being above England.
If you are going to use rankings to produce predictions, at least use the Elo Ratings, which have been shown to be the strongest predictor of team performance.
Anyway, rant over... in summary, our predictions are much stronger than Gracenote's.
RSS Feed