Stats Insider and the World Cup

The World Cup is perhaps the planet's favourite universal event, revered around the globe and watched by more than 3 billion people across a month of enthralling action.

Modelling thirty-two teams who play together approximately a dozen times a year is difficult and fraught with challenges. There is a far larger data set available when a team plays up to fifty games per season compared to a handful of qualifiers and/or friendlies, and the depth and quality of data available for major domestic leagues (EPL, La Liga etc) still significantly outweighs that kept on the majority of non-European international games.

A standalone tournament also can't simply be modelled off the last iteration - nation's managers and squads have generally undergone vast changes, and data from four years ago has very little relevance to the present day. However, all of that difficulty also presents an opportunity - bookmakers and other gamblers face the same challenge, so if you build the right databases and put in the work you can gain an advantage. 

Stats Insider's World Cup model incorporated data representing the quality of players (influenced also by their club and the standard of their club's league), performance history at the team-level, and the strength of opponents (commonly referred to as 'Strength of Schedule' in US-centric sports), amongst other variables, as the core of our machine-learning algorithms. The model also accounted for Russia's home advantage, with World Cup hosts historically holding a distinct advantage on their home soil.

All of this data combined to form our prediction model, ultimately allowing the model to identify games where it identified a "edge" against the market - creating a competitive edge against the bookies. In essence, our edge could be summarised by the percentage difference between the Stats Insider projection of a team winning, and the percentage indicated (or implied) by the bookmakers' odds.

We designed our own World Cup Simulator for the tournament, which simulated the outcomes of all 64 matches, 10,000 times while the user watched. You can read about how it worked here. 

The Stats Insider World Cup Simulator let users explore the 10,000 different ways the tournament could play out.

Stats Insider offered a wide array of coverage during the World Cup, with futures predictions for the entire tournament updated at the conclusion of each stage, match predictions updated in real-time - right until lineups were announced prior to kickoff - as well as a regular podcast and daily written content. Some of our best picks were also published on our Twitter account @StatsInsider.

So, let's take a look at the successes and failures of the model during the World Cup, across key markets such as Head-to-Head, Total Goals and Both Teams to Score, as well as looking at how favourites and underdogs fared throughout the tournament:

For example purposes, we have used a suggested outlay of $100 per bet on each of the markets that Stats Insider considered an "edge." In this example, results would have looked like this:

Across the Head-to-Head, Total and Both Teams to Score markets, the model would have registered a profit of $3055. This was thanks to a Head-to-Head profit of $2698 and a Total Goals profit of $781, despite a $424 loss on Both Teams To Score.

The "highest probability" does not reference an edge in the market, nor was it suggested to bet on. This is just a reflection of Stats Insider's "favourite". For reference, the bookmakers "highest probability" (favourite) won 36 times and lost 28 times. That's another win against the bookies - just! - for Stats Insider!

Here's how the profit/loss of these three markets trended throughout the course of the tournament:

The profit margin was helped by a couple of big successful plays on long odds underdogs, with both South Korea ($22.25) and Mexico ($8.00) registering victories over Germany (who we tipped pre-World Cup to finish bottom of the group on the Stats Insider Podcast) significantly boosting the coffers.

Despite popular opinion, this World Cup was not as unpredictable as previous versions of the tournament. There were fewer draws, fewer underdog victories and more favourites emerging victorious, than at the last two tournaments:

56.25% of the favourites won in Russia, higher than the 48.44% in 2014 and 51.56% in 2010. 23.44% of underdogs tasted victory in 2018, which was fewer than 2014 (26.56%) and 2010 (28.13%).

Had you been betting just on favourites, underdogs or draws here's how you would've fared:

• $327 profit on favourite betting
• $2171 profit on underdog betting
• $1278 loss on draw betting

So despite fewer underdogs winning than in the previous two tournaments, it was the most profitable approach to head-to-head betting outside of following our model's projections.

We're continuing to build on the success of the World Cup model with England's Premier League, Spain's La Liga and the Hyundai A-League on the horizon over the next couple of months for Stats Insider.

As part of our World Cup analysis, we teamed up with bookmaker TAB to deliver punters "The Edge" - a revolutionary new betting tool designed to help punters make smarter wagering decisions. This will extend to our Premier League, La Liga and A-League coverage, so you can expect even more detail and analysis of our projections. 

Make sure you've signed up to Stats Insider so you don't miss out on anything.

As a standalone event, a World Cup tournament is more difficult to model than many of the other sports we cover here at Stats Insider, but as we explained earlier, the lack of information also presents a wagering opportunity. Overall, we were extremely happy with the results from our tournament simulator, and of course, our projections. Based on the results outlined above, you'd be a very happy punter if you followed our projections! 

At Stats Insider we endeavour to be as transparent as possible, so in that vein, if you have any questions or comments feel free to get in touch with us and we'll be happy to answer them!

We are currently providing predictions for AFL and NRL, with more sports coming soon - you can get them free with a Stats Insider account. Sign up now to get free access to everything for a limited time.

Get more of this. Follow us on Facebook & Twitter.