What are the odds?

On 18 August 1913 at the famous Casino de Monte Carlo in Monaco, a roulette wheel came up black 26 times in a row. During the sequence some gamblers practically bankrupted themselves betting on red, feeling certain that a change of colour ‘must be due’.

The odds of 26 consecutive spins of a roulette wheel each being black is about 138 million to one. But if you think about it, the fact that it happened wasn’t remarkable.

The run of 26 consecutive blacks was no more or less likely to occur than any of the other possible combinations when you spin the wheel 26 times.

It wasn’t surprising that it happened; it was just incredibly unlikely that those people in the casino that night would be there to see it. And it certainly was unlucky for them if their misunderstanding of the nature of random numbers meant they bet their shirts on red.

The lesson of the 26 blacks that day is that anything that can happen, and will happen if you wait long enough. It’s a lesson that we take to heart at GIG where our motto is ‘we put numbers on everything’. By everything, we mean ‘everything’. Even the really, really unlikely things.

As generators of our own in-house prices (aka odds, or probabilities) for sports events, we never get tempted to think ‘that will never happen’. We don’t do vague, we don’t guess. We like being specific.

We take inspiration from the roulette wheel when we price up sports events such as this summer’s FIFA World Cup. A roulette wheel is effectively just a highly polished RNG, and it has inspired a style of modelling that we use extensively at GIG, a Monte Carlo simulation.

In the same way that spinning a roulette wheel an infinite number of times would show you that a run of 26 consecutive blacks happens about once every 138 million times, setting up a simulation of an event like the World Cup and running it many times on a computer tells you how likely different teams are to win it.

Picking winners

The simulation reveals how likely Germany and Brazil are to lift the trophy, which is interesting and useful to know if your business is setting odds.

But the real value of a Monte Carlo model is putting a number to things that are more complicated to compute, and propositions that are much rarer, such as outsiders Saudi Arabia winning the final on 15 July.

Tempted to think: ‘that’s not going to happen’? Well, you shouldn’t. Because it will. Or at least it would if you were to play the tournament enough times. Saudi Arabia winning the World Cup is a lot more likely than a roulette wheel coming up black 26 times in a row, that’s for sure.

GIG Sports’ in-house modelling team has created its own Monte Carlo simulation model for the 2018 World Cup. It’s built in F#, which is less prone to errors than object-orientated languages, hosted on Azure in a virtual machine with 16 cores running Ubuntu Linux, deployed with Docker.

There is a lot of data, so we serialise it using Protocol Buffers and then the odds feed calls the engine every time something changes in a match. It is accessible by the pricing, modelling and risk and ops teams via the company’s own web-based model platform.

This gives GIG Sports the ability to price up and offer outright markets on the World Cup quickly and accurately.

Will any of us live to see Saudi Arabia win the World Cup? Probably not. But we might.

Endre Nesst, interim sports director GIG Sports