THE WORLD FOOTBALL, BAD PREDICTIONS AND HOW TO MAKE PREDICTIONS LESS BAD

This is not necessarily a blog about the 2018 World Cup — Russia 2018, but how well does The World fit the topic of today, why do predictions of whatever seem to be so bad, and if they are bad then? Why do we make predictions and forecasts? and the best, why do we make decisions about our future based on our own predictions and on experts we trust blindly?

For the record, I have this issue “in the gut” from the disappointment of a polling company that was terribly wrong, week after week, in its oracle of the next president of Costa Rica (2018 elections). First the media made this company the diva of the polls and filled it with praise, in the end turning it into “meat of memes” and synonymous with bad divinations. The fault of the company? Not necessarily.

But let’s go back to the world. My hero of the predictions (and not for that reason exempt of blunders) Nate Silver and his boys of the company FiveThirtyEight (prize for the one that tells me where that name comes from …), I create a “modelazo” to predict the next champion world, or rather, an index of possible world champions.

Let’s notice several things. The first thing is that four of the first ten, Spain, Germany, Argentina, and Portugal are all out of the World Cup. The second, and much more important, observe in the last column the chance to become champion: Brazil 19%, Spain 17%, Germany 13%, etc. The message behind the message (before the competition started) was that everyone was even and nobody was an extreme favorite. If you want to study in detail the whole model you can search here .

One of the things that I love about FiveThirtyEight is that their studies are not static (precisely the problem the company had with ethical surveys was not to update the assumptions of its model). Let’s see the information updated to today (Sunday, July 1, 2018).

Even so, the level of uncertainty, the number of variables to be taken into account, and last-minute fortuitous situations during matches make it almost impossible to predict the results. Not even the Soccerbot, the favorite model of the punters, created by the mathematician David Sumpter, author of the book Soccermatics, can paste them all. These are the estimates of the Soccerbot before the games this morning:

  • Brazil 7–2
  • France 4–1
  • Spain 9–2
  • Belgium 6–1
  • England 8–1
  • Croatia 10–1
  • Uruguay 20–1
  • Colombia 20–1
  • Mexico 40–1
  • Switzerland 40–1
  • Russia 60–1
  • Sweden 60–1
  • Denmark 100–1
  • Japan 250–1

And we already know what happened: Spain (third in the estimates of the Soccerbot) is out and Russia (eleventh of the list) is still alive. And Croatia (sixth in the preferences of the model) saw sticks with Denmark (number thirteen).

By the way, bookmakers handle the following numbers for the Brazil-Mexico game: Brazil -200 to win (bet $ 200 to win $ 100), while Mexico is at +650 (risk $ 100 and win $ 650). A tie in regular time is paid at +315. By the time you read this blog we will know if the Soccerbot was right or if Mexico will have its famous “Fifth Party” that has not reached since 1994 (6 consecutive eliminations). In the afternoon on Monday we will know the real result, or as Winston Churchill would say:

“I always avoid prophesying in advance because it is much better to prophesy after the event has already taken place”

How to improve predictions and forecasts?

Create scenarios about the future is necessary for decision making, how much material will I need? How many clients will visit me next year? Who will win the 2018 Russian World Championship? Understanding the limitations of these scenarios is even more important. An article in the 2007 Harvard Business Review gives us the following six tips related to forecasts ( Six Rules for Effective Forecasting, Paul Saffo ).

  1. Define your cone of uncertainty

Saffo calls “cone of uncertainty” the possibilities that extend from a particular moment or event. The width of the cone is determined by the factors to be taken into account for making a decision. The quantity of factors gives a qualitative idea of ​​the uncertainty. For example, in order to predict how many electric cars, I am going to sell in a market, I need to consider factors such as the price of oil, political decisions regarding electric cars, environmental care culture, and not only the statistical-mathematical elements that are the product of a series of time.

  1. Observe the S curve

Changes usually do not follow a straight line. The change begins slowly and incrementally, and explodes suddenly. Understanding the inflection points of the S curve is important for forecasting.

  1. Take into account things that do not fit

Instead of eliminating the “outliers” or spurs of your model, study them at depth, they can be the signal of an anticipated change.

  1. Keep strong opinions weak

Similar to the previous point. Do not keep a lot of time clinging to historical information that confirms your conclusions (although historical information can give you more light about the future according to the following advice).

  1. Look back twice as much as you look forward

For historical information to be useful you should look as far back as possible, so you can find patterns that you would not otherwise have found.

  1. Know when not forecast

Sometimes forecasting will be easier, other times it will be impossible. If you do not have enough information to make an educated forecast, it is better to wait for better or more recent information.

In other words, add a lot of “indigenous malice” to your predictions.

To finish, let’s return to the theme of the Soccer Prediction World Cup. At the time of writing this blog we already know that at least one of the finalists of Russia 2018 will be a non-traditional team. The baseball player became a popular philosopher Yogi Berra when he said:

“The future is not what it used to be”