Before we get into everything let’s just recap the project quickly
The 2018 FIFA World Cup was simulated 200 times using Football Manager 2018(Thank you to @Golden_FM for creating the world cup file). The official World Cup squads or any injuries were not used in the simulations as this was started before they were announced. Also Germany and Japan did not have their official players due to licensing.
The simulation results are being compared to the odds being provided Bet365. The win probability was taken from betting odds using (1/x)*100 where X are the decimal odds for that outcome. Then after each round of games the simulations and betting odds will be compared to real results to see how they compare and if Football Manager is as good at predicting match outcomes as a bookmaker.
So now we’ve had that refresher let’s look at how Football Manager got on,
Through the group stages neither Football Manager or the Bookmakers did amazingly well with their predictions with quite a few upsets. However, Football manager barely edged out the Bookmakers. Through all 48 group stage games the 2 prediction methods only differed on one game and that was Mexico vs Sweden. The Bookmakers took Mexico who lost which meant the bookmakers ended the group stages with 63.48% of games correctly predicted. Football Manager on the other hand took Sweden which gave it the edge over the group stages, correctly predicting 65.56% of group stage matches.
Now for the knockout stage the predictions weren’t quite as accurately recorded, because of the time restrictions of the world cup schedule, instead of running each individual match again 200 times I took the pre-tournament odds of a team progressing through to the next knock out round and compared that with how many times they progressed in the Football Manager simulations.
So just for example the below is the predictions for the round of 16 game Uruguay vs Portugal. The probabilities aren’t for each team to win that specific game just the probabilities they would make the quarter finals.
Using this method through all the knockout games Football Manager again edged out the Bookmakers in correctly predicting matches. Football Manager correctly predicted 84.38% of knockout games, while the Bookmakers had 65.63% correctly predicted.
This gives Football Manager 76.31% of games correctly predicted for the whole World Cup, while the Bookmakers had 71.85% correct.
Football Manager also correctly had France lifting the World Cup and Kane with the golden boot, while the Bookmakers had Brazil and Neymar respectively as favourites. However, neither had Modric anywhere near favourite for Golden Ball winner.
I think it’s also interesting to note that in all of Croatia’s group stage games and for every knockout round, Football Manager gave Croatia a higher probability of success than the Bookmakers.
I set out with this experiment to test if Football Manager simulations were accurate enough to predict the 2018 World Cup results as well as a big betting company like Bet365, not only did Football Manager do as well as the bookmakers but it actually performed the bookmaker predictions. Now this doesn’t mean Football Manager can out predict the betting companies when the 2018/19 seasons start up, but it does show it may be worth further investigation.
Anyone who wants to look at the raw data can check it out here
Before I leave this experiment behind just a quick note
I’m currently thinking if I will keep looking at some different football statistics experiments after the World Cup is all finished and posting the results on this blog or even maybe try this same method for the 2018/19 Premier league season. If you’ve enjoyed this blog and would like me to keep going or if you have any feed back let me know on twitter