[ad_1]
Poor Mystic Marcus. The so-called prophetic pig who predicted that England would beat Croatia on Wednesday was vilified on social media following the loss of England, with fans calling for it to turn into bacon.
Although this is dark days for Marcus, the pig consolation in the fact that he is far from being the only one to be completely wrong about the World Cup. Before the tournament began, a number of investment banks have built sophisticated computer models to try to predict future world champions through the power of data badysis. It may seem odd that financial institutions compete to be football soothsayers, but they do so as a kind of swinging statistical model; The theory is that if they are good at predicting the World Cup, they should also be able to predict the direction of the markets. Unsurprisingly, however, just about all the predictions of the big banks have been embarrbading.
Let's start with Goldman Sachs. This year, the banking juggernaut continued its unsuccessful tradition of choosing Brazil as the winner (he has done so for the last three World Cups). Unlike Marcus the pig, Goldman Sachs did not reach this conclusion by instinct, but through "hours of calculations, 200,000 probability trees and 1 million simulations." Goldman Sachs 'artificial intelligence algorithms led them to conclude that' England meets Germany in quarters, where Germany wins; In addition, the banks said: "For the doubters, this final result has been rechecked with atrocious detail by our chief economist (German) Jan Hatzius!" [19659005AprèsquelaBelgiqueabattuleBrésilenquartsdefinaleilsontréitéré "AvecleBrésilhorsdelaCoupeduMondelaBelgiqueestausommetdenotretableaudesprobabilités" ontécritlesbadystesdeGoldmanlasemainedernièreLabanqueprévoitégalementquelaBelgiqueaffronteral & # 39; AngleterreenfinaleGoldmanSachsarévisésaprévisionle9juilletpourprédireunevictoirebelge "src =" https://i.guim.co.uk/img/media/042f67dc49f88978b46c42bde0417e5e0ea62e7c/0_52_923_417/master/923.jpg?w=300&q=55&auto=format&usm=12&fit max = & s = 2122fd4ec0f20449277b6558622fdf57 "/>
Goldman Sachs revised his prediction on July 9 to predict a Belgian victory. Photography: Goldman Sachs
UBS put so much unnecessary effort into trying to predict a winner. The Swiss multinational has used the results of the previous five tournaments, controlling factors such as the host nation's advantage and the Elo rankings (an objective measure of the team's strength that shows how a team is doing). Is behaved in the past). as Monte Carlo simulation. This gave them the results of 10,000 virtual tournaments that they then badyzed to see how many times each team won. Germany, Brazil and Spain were the teams most likely to win, Germany being the favorite
of each team, which was calculated from individual estimates of the game transfer value and previous performances. The idea being that the higher the value, the greater the chances of success. They chose Spain as the winner so clearly their strategy, although novel, was not entirely bulletproof.
The bank with the most successful prediction was Nomura. The Japanese firm used the portfolio theory, with badysts explaining that they looked at "the value of the players in each team, the momentum of the team's performance and the historical performance to arrive at three portfolios of teams to watch ". the 21st FIFA World Cup; Although they predict France would play Spain in the final
Unpredictable: Domagoj Vida from Croatia at a training session before the World Cup finals. Photography: Mikhail Japaridze / TASS
To be fair to all data experts who have not understood correctly, it is extremely difficult to try to predict the World Cup. Debs Balme, the badysis director for Merkle, a performance marketing agency ("and a super fan of England") told The Guardian that the accuracy of statistical modeling depends on the amount of data that you can provide. "For sports like baseball or basketball where there are a lot of games against the same opposition, it's an easier solution to predict because there is more data available." 162 games per season, for example, and the Mets and Yankees played 115 times, so there's more history of performance [than in the World Cup] and more data richness to make more accurate predictions. "Some abnormal results of this World Cup, as the elimination of Germany in the first group for the first time in 80 years obviously makes the predictions more difficult.
In recent years, the evangelists of Technology has touted the power of data to predict the future and make important decisions about everything from the economy to the performance of employees.However, time and time again – polls being wrong about Donald Trump to Brexit's forecast errors – we recall that data badysis has its limitations. Seeing how bad the big banks were about the World Cup is a serious reminder of the fallibility even the most sophisticated statistical models.
