FoG Prediction
This is a player based rating adjusted for the market information.
The first step for each country was to list the squad and the clubs that their players play for. Once the full squads were announced, I averaged the top 16 players club's elo rating (from clubelo.com).
Assumption: The assumption here is that players find their level, a player that has made it to Man City is better than a player stuck in some middle eastern league.
Risks:
Not all teams are rated on clubelo.com and I have had a guess on the clubs outside Europe and the lower league European clubs. This will only affect some countries who aren't going to make much of a dent in the outrights anyway.
The big 5 leagues of England, Spain, Germany, France and Italy are packed with foreign talent filled in with local talent. As a result I feel its inflated the ratings for those countries.
It's possible also that teams can be so much more than the sum of their parts, this aggregation of the individual players clubs ratings does not take into account how the players fit together and compliment eachother.
A team with lots of weak players and one strong player could have the same average as a team of average players. I feel the strong player wouldn't get the service required and the more balanced team would perform better. The method only takes the overall average so this is not accounted for.
Not all players at each club are of the same level. Haaland is much better than Alvarez for example, but they will both carry the same weight in this analysis.
The second step was to use the ratings above to produce odds for the group matches. Comparing against the already released match odds we can see where the player based rating was too big on some countries, so we can adjust them as appropriate.
Assumption: Here we are relying on the betting markets to find the fair probabilities for each match, which we can infer ratings from.
Risks:
So using the group match odds is a bit weak as it wont take into account that the stronger teams may play to save themselves for the knockout matches, especially in the final match if they have already sealed qualification. In other words different teams will have different motivations.
Some clubs are starting the group matches with injured players (Depay, Walker, Mane etc) and they will be stronger when these stars return.
The third step was to simulate the entire tournement with the second set of ratings. Comparing against the outright market we notice that the match odds of the group games only gave the teams strengths in relation to the other teams in their group. So this step lets us adjust all groups by the same amount to try and better fit with the outright market.
For fourth step I started running and rerunning the full sim and adjusting the teams individual ratings to try and replicate the outright betting market.
Assumption: Again as with the group match odds, we are relying on the betting markets to find the fair probabilities for the outright winner. I assume this will paper over some of the risks and issues listed for the second step.
Risks:
The match odds markets will be more efficient/accurate than the outright market.
The Final Ratings and Predictions
Below is the final ratings, the player based ratings are all but completely over ruled by the market. I still like to include it. It might prompt you to consider looking deeper into some squads. In some cases you might decide that the market is adjusting too much and then reel it back a bit. Most of the adjustments can be justified this time by some of the risks listed above.
To be honest as well it serves as a mental picture to have a rating based on the clubs. You can quickly compare the aggregate club ratings of Brazil's players to see where they would sit in the clubelo table. Brazil are rated between Milan and Ajax. Qatar is roughly equivalent to Coventry. Australia are roughly around Spanish Liga 2 team Malaga based on the initial played based ratings.
So the summary above shows the initial player based ratings and the adjustments at each step to incorporate the wisdom of the betting markets. The final adjustment for the outrights does paper over some of the issues (player availability, motivation) of the match odds adjustment, however I do think in general the match odds markets are more accurate than outright markets. So I have reduced the final adjustment back a bit to get my final ratings. Running a sim with the ratings to produce the final probabilities.
The numbers I submit will differ just based on the randomness of the sim.
Comments
Post a Comment