Thursday, June 3, 2010

World Cup Simulator

While I cannot recommend a particular approach to estimating probabilities in soccer, I can provide a simulation framework for the coming World Cup tournament. The schedules were compiled from the FIFA and Wikipedia sites, and the structure is there to input team ratings, churn through a hastily modified log5 approach, and generate simulations in a Monte Carlo method.

Email or comment with advice on improvements, including team ratings and a formulaic approach to assess win/lose/draw probabilities given those ratings.

World Cup Simulator
Original Author- Erich
Original Release Date- 20100603

5 comments:

ilike#s said...

Where did the rating that are included come from? For now, I would recommend Nate Silver's SPI's he made for ESPN.

xlssports said...

Ratings were a hack job using FIFA rankings.
Silver's SPI would be great, though he has not released many details of his game probability model. If you can generate a function that comes close to his two examples described here, that would be something we can try to put underneath this spreadsheet's hood.

ilike#s said...

Could we just create log5 numbers using his Offensive and Defensive Ratings?

During qualifying, I used his Offensive and Defensive Ratings between opponents (scaled against average) and got winning percentages that were pretty close to lines at Pinny so I felt good about them.

xlssports said...

Ties present a difficulty with log5.
While I am skilled in Excel, I'm not the best mathematical reverse engineer. If you have a formula which you would like to use within this engine, email me and I'll replace the current hackery with whatever has worked for you.

ilike#s said...

Just sent you a ratings file