Big data: Predicting the outcome of this World Cup tournament

From today onwards, another World Cup tournament officially kicked off, the world's fans and non-fans will also enjoy this passion burning, blood surging good times. The crisp beer has been filled, the shiny screen is ready, and everyone's love for the country and perhaps some unrealistic expectations are watching the pitch. The current World Cup statistics have been hot, Brazil is expected to usher in a total of 3.7 million people around the world, the resulting economic effect as high as 30.http://www.aliyun.com/zixun/aggregation/19145.html " >3 billion; pazopanib, which specializes in peripheral products such as star Cards, estimates that the revenues from stickers will be as high as £ 89.1 million in Brazil alone, while in the UK the company will expect to reach 84 million pounds in total sales during the World Cup.

But for some important topic, the statistical results and numbers seem scarce, and that is--who can be the last winner of this World Cup. We can figure out how many fans will go to Brazil in person, how many Britons will chew a good pizza in front of the TV, and how many avid fans collect star cards, but can we use the data to predict who will win the championship? To avoid bias, we will learn from the skeptics ' point of view, while feeling the firm determination of supporters such as Goldman Sachs that the data-driven model can successfully predict the World Cup title.

Skeptic: No, it's not possible.

Left player for Flamini

In short, football is a challenging sport, and it is difficult to use analytical methods to accurately predict its results. As The Economist reported last year, it is not easy to apply the "orb theory" to football. Compared to the more measurable discrete events in baseball, the 22 players on the football playground need to move and combine each other in endless ways. Football has a natural dynamic attribute, which makes it challenging to judge which factors need to be considered and to constantly get the results.

Despite the high degree of difficulty, but this is not impossible; we have recently reported on the same dynamic elements of basketball, the current camera system has been able to crack the complex data, tell the team's coaches basketball and the players on the pitch in the field-the accuracy of even every season each game for every second. Similar methods of analysis can be done in the field of football; Companies such as Prozone and Opta have started tracking a range of indicators on the playground-including the position of the players in the game, the way they pass and the chances of scoring. In general, there are about 2000 data events per game.

But the relative value of the data remains to be seen. Team managers have successful cases and failed situations when they rely on raw data for player selection. First look at the success story: Arsenal manager Arsene Wenger has noticed Flamini's excellent performance in the game and has decided to replace Vieira with him. But the failure is the same: Ferguson has determined that Stam's current number of steals is not as good as it was before, and decided to eliminate it. However, later data show that Stam in Italy with outstanding performance to prove their value.

The crux of the problem is that while the data can explain which players are running the fastest or running at the longest, and who steals the most, a good footballer is not just a simple summation of several values. The data reflects the past and does not prove that the player's future performance will continue to be the result of the current comprehensive indicator.

Scientist: Yes, we can.

In predicting the outcome of the World Cup, Goldman Sachs wisely avoided the tricky analysis of player-specific attributes and instead adopted a more general approach. They looked at the performance of each national team at the World Cup and their current ranking of ELO (that is, the level of competition) and made a forecast model. They explained their methods as follows:

The results of each event are based on a complete regression analysis and are used as reference information for all official international competitions since 1960-that is, not including friendlies. This brings us about 14,000 forecast indicators to evaluate the model calculation coefficients we use. In regression analysis, the variables are the number of goals for each participant in the weekly game. Based on the literature on the modelling of football matches, we assume that the number of goals that a particular team matches against a particular opponent follows the Poisson distribution (a common discrete probability distribution).

Their model finds Brazil has an astonishing--48.5% against winning rates. They expect Brazil to beat Argentina by 3:1 in the final, compared with 14.1% in Argentina. The Brazilian team has been able to get such a high rating, with a variety of factors, including the excellent ELO system rankings, the stronger actual performance of the World Cup relative to other events and the host's advantage this year-since 1930, In all World Cup competitions, the host Country team won the Hercules Cup by as much as 30%. According to the model, Brazil has a 65% chance of winning the championship in front of its home this year; The days of Europe's top teams are less than good, and history shows they have never won the World Cup in the Americas.

However, the model relies entirely on past reference indicators and clearly fails to reflect future uncertainties. Goldman Sachs used a similar analysis model to make predictions about Britain's performance in the 2012 London Olympics based on previous performance. They expect Britain to win 30 gold medals and a total of 65 medals, while in fact Britain eventually won 29 gold medals with a total of 65 medals.

Stephen Hawking, in a different, disparate way of analysing the predictions of Goldman Sachs, thinks the UK has the biggest potential this year, considering a lot of data. In Goldman's computing model, the UK's performance this year will be disappointing or even impossible to qualify for a group match, and it seems that the UK is best to believe Hawking's conclusions lest morale be low. In further proposals, Hawking argues that the UK is best able to take the 4-3-3 shape and that the time of the game should be at around three o'clock in the afternoon Greenwich Mean time, with as many bald or blond players as possible (because such players have higher scoring chances). In addition, he suggested that athletes who took a free-kick or a penalty kick would take a three-step run-up and play with a side kick (which would increase the chance of scoring by 10%) and try to put the ball on the top left or upper right corner of the goal (the area scored 84%). However, Hawking admits that this is already the data in the football game can give all the guidance. After all, England's performance has been pretty bad in the actual free throws.

With so much to say, can big data help us predict the outcome of the World Cup? The only way to get answers is to keep an eye on the next series of events and see if Brazil can make it to the top. Of course, you might as well pay attention to the England team's ability to make a breakthrough in its ever-catching-chicken penalty shootout.