Another world cup and another opportunity to predict how the tournament will play out – we all want to know which teams will hang around and which will exit the competition, and, more importantly, who will win.

So, how should one go about making a prediction? You could use some high level historical trends like this, you could ask an expert, or you could utilise the prediction powers of a psychic octopus.

Alternatively, if you are Goldman Sachs, you could use machine learning to run 200,000 probability trees to mine data on team characteristics and individual players to predict match scores. You could then use this to simulate one million possible evolutions of the tournament to determine the probability of each team progressing through the rounds. Sounds like some serious leg work and fancy footwork, but hats off to Goldman Sachs for doing just this.

Goldman Sachs have used this methodology to predict which teams will make it to the round of 16 and how the tournament will play out from this point to the eventual winner. The model predicts the number of goals scored in each match, and the un-rounded score is used to determine the winner. For example, according to their results (as shown in the image below), Germany will meet England in a quarter final match and will narrowly beat England by 1.47 goals to 1.28 goals.

While this ridiculously complex analysis might appear a little counter-intuitive, especially to a girl (how exactly does one score a fraction of a goal?), it is undoubtedly very, very cool. But, it does beg the question, has Goldman Sachs scored a bit of an own goal here? Clearly, they have used very competent (and highly paid) resources with very good technique. But they seem to have applied them in the wrong direction. Consider for a moment that the average annual salary for a Goldman Sachs employee is $367,564 per annum [1]. Let us assume Goldman took three of their very average employees and assigned them 10 days each to prepare the data, run the analysis and compile the findings, that would equate to over USD 42,000 (~R570,000) just in resourcing costs, excluding any other costs, such as the cost of the data and systems required to run their models. This is an especially pricey exercise to go public with if they are wrong, which is highly likely given the random nature of soccer. Indeed, it is this very characteristic that makes the game so interesting to watch (for some people at least).

At 71point4 we too like to predict the future, but without spending USD 42,000, or spending anything preferably. With no psychic octopuses easily accessible, and because, like Goldman Sachs, we like to base our predictions in data, we turned to the betting markets.

Betting markets can be a useful source of data that is easily accessible (and free!). Prices (or odds) in these markets essentially reflect the combined predictions of many people who vote with their feet using their own hard-earned money. A particularly fine feature of this data is that it updates immediately as more information becomes available. So, for example if a historically mediocre team suddenly appears to playing beautifully mid-game, the betting markets will update to take this into account. Of course, there are obviously limitations to this data set, but nothing is perfect.

We have therefore used the odds of teams winning their initial group matches to predict who will get into the round of sixteen. Then using the odds of which teams will make it to the quarter-finals, semi-finals and finals and the odds of which team will win, we can see how teams will progress and who will win the tournament.

We pulled data from oddschecker.com on the 13th of June 2018 (one day before the start of the World Cup). Based on about 30 minutes of simple number crunching – mostly consisting of copying and pasting data out of an internet browser and into Excel – we came to the following results (in each game the winner is bolded).

Happily, (and unsurprisingly) the outcome is very similar to that of Goldman Sachs. We arrive at the same winner and see similar teams progressing to later rounds of the tournament. We get to the same prediction for the final, which bodes well for Brazil fans.

There are just two differences in our results. Firstly, Goldman Sachs predicts that Portugal and Spain will progress from Group B, with Portugal the group winner. Our analysis also shows them both progressing from Group B, but with Spain as the group winner. And secondly, Goldman Sachs predicts that Saudi Arabia will progress to the Round of 16, whereas our prediction sees the hosts, Russia, progressing. According to the betting markets (as at a day before the tournament kicked off) Saudi Arabia is going to convincingly lose all of their group games, and is the least likely team in Group A to proceed to the Round of 16.

Based on the results from the first game that kicked off last night, conveniently Russia vs Saudi Arabia, it seems Saudi Arabia is unlikely to come back from a whopping 5 – 0 loss. But who knows? They may shock all of us (except Goldman Sachs), and triumph over some of their other Group A competitors. This is, after all, a very beautiful game indeed.

At the end of the data, we will need to wait for the final whistles to see who was right. Almost certainly, neither of us will be. Perhaps the better question may be, who was less wrong? If history is anything to go by, we can expect the unexpected; a player will probably bite someone else and get sent off turning a game around, or maybe the hand of God will make an appearance. But at least we didn’t spend USD 42,000