The report states that all the predicting was based on past performance and common sense and the probability that the same nimble footwork, will be displayed by most of the top teams like Brazil, Argentina, Italy, France, Germany, Spain, England and Portugal. But anyone who has seen the matches till the very end has seen almost all of these big names fall by the wayside with the exception of Spain. Holland came out of the mist and took on Spain in the final round.
To forecast the winners at each step, they have used and, I quote from their report -
Simulation-based model with Poisson probabilities
Our model calculates the probabilities of each team reaching different stages of the world cup. The results are derived by simulating the world cup games by assuming that goals scored by teams follow a Poisson distribution. The following steps are involved in developing the model:
Step 1: Identifying the top 16 teams. We identify the top 16 teams based on our subjective judgments, current ranking and forms.
Step 2: Building lambdas. For each pair of team in the top 16, we assume that the goals scored in a match up between the pair will follow Poisson distribution. The mean (henceforth addressed as lambda) of the Poisson distribution is a function of historical performance with higher weight to recent matches, FIFA ranking and current form (captured by Elo ranking).
Step 3: Simulation. Once we have lambdas for each possible combination, we run a simulation of the world cup post the league stage. For a draw, we assume that the team with better ranking has a higher chance of moving ahead.
We believe that by taking goal scored as our simulation variable, we are closer to the actual match scenario and capture both attacking and defensive strengths of a team (vis-à-vis taking a win-loss as our simulation variable).
The report gave Holland (Netherlands) 74% chance of reaching the quarter finals and a mediocre 22% chance of reaching the semi-finals and a low 17% chance of making it to the final round. Yet, look what happened. The favourite was Brazil, which was expected to make it all the way to the final day.
The other top teams considered to be good bets to make it to each successive round were in the following order.
Quarter final Grads – Portugal-82%, Brazil-77%, Holland-74%, USA-74%, Argentina-70%, Paraguay and England-68%, South Korea-33% and surprisingly Germany was at the bottom of this list at 32%. Spain, Italy and France were not even considered.
Semi-final Grads – Portugal-66%, Brazil-57%, Uruguay-40%, USA-37%, Holland-22%, Argentina-26%, Holland-22% and Germany-18%.
Final Grads – Brazil-38%, England-33%, Portugal-22%, Argentina-20%, Holland-17%, Germany-14%, Spain-12%, USA-11%.
The final winner was supposed to be Brazil at 27% followed by England at 14%, Holland at 12%, Portugal at 10% and both Argentina and Spain at 8%.
Anyone placing their bets based on these numbers should instead have just gone the Russian roulette route, or taken a chance with Paul - the Octopus - who got it right with plain old guesswork and probably what passes for his gut instinct. I know I have the benefit of hindsight, which is why the report makes for great reading.
The icing on this cake is that the finals was supposed to be played between two of these four teams – Brazil, Argentina, South Korea and Portugal.
Brazil and Argentina are previous champions, so they will always have a reputation to maintain and defend and Portugal has the flashy Christiano Ronaldo (though I prefer the tall, dark and handsome Luis Figo) but South Korea?
See the report here - http://www.kotak.com/kotak_gamechanger_crystalballkicking_FIFA_2010.pdf