is the World Cup of big data predictions a reliable match?

Ask：

The big data this year has replaced Paul, the Octopus, as a popular tool for predicting the World Cup. But is this a reliable predictor of how big data can really predict the World Cup?

Secretary：

At present, using http://www.aliyun.com/zixun/aggregation/13568.html "> Big Data Technology to predict the World Cup, both the Internet companies such as Baidu, as well as the well-known investment bank such as Deutsche Bank."

The prediction does use a lot of data

Through interviews with relevant people in Baidu, we found that Baidu forecast the World Cup main data sources include: Baidu search data, team base data, player base data, odds market data. Baidu large data through analysis of the past 5 years 987 teams of 37,000 game data, a total of 29,610 players, 112,285,543 related data, the establishment of the football game forecast model.

In order to verify the accuracy of the model, Baidu used the 2010 World Cup in South Africa to verify the accuracy of the knockout, the specific method is to input the model of the 2010 World Cup competition, team, players and other related data, from the forecast model to calculate the knockout results, compared with the results of the match, The results showed that the 16 knockout matches were 12, and the accuracy was 75%.

Deutsche Bank is based on the team's FIFA rankings, historical record, player composition and betting odds and other factors, the establishment of a quantitative analysis model, and based on a complex calculation to obtain a title probability table. Brazil ranked first, followed by Germany, Spain and France. And then according to some assumptions, the final winner.

Second, the extrapolation logic is not perfect

The final winner of Deutsche Bank's calculation is based on the probability table. Specifically, Deutsche Bank picked out the top 10 from the championship probability table, and the 10-strong rankings were Brazil, Germany, Spain, France, Argentina, Italy, the Netherlands, Portugal, Uruguay and England. Deutsche Bank believes that the final title can only be selected from these 10.

In the process of screening, Deutsche Bank put forward the concept of "rotation cycle", the bank believes that in the past 19 World Cup, there are 3 consecutive four World Cup by the four teams to win, and 2 are consecutive two World Cups from different two teams to win, The remaining 3 sessions from the three World Cup in 2002-2010 constituted Deutsche Bank's heart another four different winners of the rotation cycle of the top 3/4, Deutsche Bank thus excluding 3721.html ">2014 Brazil, Italy and Spain to win the possibility, Because they are the winners of the last three World Cups.

That leaves Germany, France, Argentina, the Netherlands, Portugal, Uruguay and England with seven teams, and Deutsche Bank has another assumption: the strong teams will come back, winning the World Cup and will certainly win the World Cup or at least one final. England won the championship in 1966 and has never been in the final since then, and Deutsche Bank believes England will have a big chance of winning.

Finally, the England team has 6 players from Liverpool, and it was the biggest player in Liverpool in 1966 and England won the only World Cup in history. At the same time, the Deutsche Bank report's author admits that he is a die-hard Liverpool fan, so the final confirmation that England will win the World Cup title.

It is not difficult to find that the logic is very far-fetched. Deutsche Bank seems to have set up a model or hypothesis only to produce a desired result, rather than to speculate on the final result based on a determined scientific model. There is a suspicion of "putting the cart before the horse."

Baidu is relatively more scientific, at least the causal order of speculation is not reversed. But Baidu's speculation is also clearly flawed, Baidu's model after its own verification, accuracy rate is only 75%. And the validation is an input to the past figures to speculate, but the past accurate future is not necessarily accurate.

Third, "drunken Weng's meaning" is not in the prediction itself

Predicting the World Cup with big data is more believable than using the octopus Paul (at least superficially). But this kind of forecasting activity is more like the use of the World Cup marketing campaign, not to draw a certain conclusion, but to attract the participation of netizens.

Baidu's World Cup forecast also with lottery betting activities, by predicting the title rate of the number, for users to buy football lottery for reference, more like a commercial activity.

However, on the eve of the World Cup, netizens have made predictions about which team can win the championship. There is no need to be too serious about how much credibility the forecast results are.