Markets. It is now proven : Goldman Sachs says anything. There’s still a week to this football world Cup, but already the predictions of the investment bank are completely false. Proud of his algorithms, “machine learning,” which turned 200 000 models, and have simulated one million possibilities, the giant of Wall Street was released on June 11, a thick document of forty-five pages providing the results of the competition. At the dawn of the semi-finals, it’s been almost of the sentence to reread it.
According to the powerbrokers of the no less eminent organization, Germany should beat England in the quarter-finals then go through to the final. Portugal needed to beat Argentina and reach the semi-finals. France was due to meet Spain in the quarter-finals. Russia was not to move the hens. And Brazil was the favorite to win the world Cup.
For the knockout stages, the computer model had predicted correctly two of the eight meetings (Brazil, Mexico, Colombia – England), and the presence of ten teams out of sixteen. For the quarter, only one meeting had been well seen (Brazil – Belgium), and the presence of four teams. For the semi-finals, no meeting has been scheduled and only one team – France – was correctly considered. A draw, a red fish or a monkey scientist would they have done less well ?
UBS… Still missed !
Of course, the predictions were tongue-in-cheek, not taking itself too seriously. Still, the bank has done things well. For its calculations, it took into account four databases : the official ranking of teams made by FIFA, the recent results of the teams (the statisticians have discovered, who would have thought, that those who have recently won games have more chances to win other), the rankings of the individual players, and the number of goals…