As the start of the World Cup approaches, bettors will be looking at what factors will help inform their World Cup predictions. There is plenty of international soccer betting data available, but how can it be used to build a World Cup betting model? Read on to find out.
On June 14, 31 qualifying sides will join the World Cup hosts Russia in a rare meeting of the best teams from five of the six confederations that make up FIFA. The qualification process for this year’s tournament began back in September 2016 and once the groups were drawn in December 2017, bettors would have started their calculations for who might make it out of the groups and perhaps make it all the way to the final.
A closer look at qualification groups
The continuity of squad make up, regularity of matches and large amounts of data make ranking teams in domestic soccer a much easier task compared to international sides. Often the qualifying process for major tournaments can be over a prolonged timeframe (albeit consisting of relatively few matches) and team churn can be significant from the beginning of qualifying to the actual start of the finals.
The Oceania Football Confederation (OFC) final qualifying group matches amount to just four matches before a two-legged final and then a final inter-confederation playoff. The Asian Football Confederation (AFC), Confederation of North, Central American and Caribbean Association Football (CONCACAF) and Union of Europeans Football Association (UEFA) generally consist of ten games per automatic qualifier. While the South American Football Confederation (CONMEBOL) stretches to a usual 18 games per side, spread over two years.
The level of the opposition can also vary greatly and this is illustrated by the case of Australia, especially if we include games from the beginning of the qualifying process. The “Socceroos” have averaged 2.7 expected goals per game in qualifying, conceding just 1.0 xG per game for an expected goals difference per game of 1.7.
Australia’s 1.7 xGD per game is the fifth best xG differential of the 31 sides that progressed through a qualifying group, behind such sides as Germany, Belgium, Portugal and Spain. However, their opponents, as a group have accrued just 1.0 xG per game, while conceding 2.43 xG per game - an xG differential of -1.43.
By contrast, Germany has a positive xG differential of 2.88 in qualifying against a group of sides whose xG differential is -0.65. Not only does Germany have a superior expected goals differential to Australia’s, they have also achieved it against largely superior teams.
Equally, England’s xG differential was almost identical to Australia’s, but it was also gained against vastly superior opponents, who had a combined xG differential of -0.3 compared to the -1.43 of Australia’s opponents.
Can bettors use FIFA rankings to their advantage?
A more objective method of attempting to quantify a countries qualifying campaign is to take the average FIFA ranking points of their qualifying opponents.
By simulating the entire tournament on a match by match basis, we can also make estimates about secondary markets, such as a side’s most likely stage of elimination.
Uruguay faced the strongest batch of World Cup qualifying opponents in the traditionally strong CONMEBOLK conference. The average FIFA ratings of their opponents was 981, 60 ranking points higher than Uruguay’s own current rating, with double headers against the likes of Argentina, Brazil, Chile and Colombia.
Uruguay’s marginally negative expected goal difference from their qualifying campaign, good enough for just the 22nd best out of the 31 qualifiers, should be viewed in the light of the strength of their opposition.
Australia, despite their seemingly impressive qualification record, had the easiest group of qualifying opponents, with an average FIFA rating of only 362 points.
So before we can begin to rate the cosmopolitan line-up for Russia 2018, using the most up to date qualifying data, we need to account for the disparity in results and the different levels of opposition.
Analysing the strength of World Cup contenders
There are a variety of methods that can be used to estimate the abilities of sides who rarely meet in competition.
These range from using historical performances of the representative of each conference against the other conferences in previous World Cups or by ranking teams based on their current FIFA ranking and the likely difference in quality assumed from the ranking differential of each side.
When we apply these type of corrections to the qualifying records of the 31 visitors to Russia there are some big movers, both up and down.
The five CONMEBOL representatives, Brazil, Argentina, Colombia, Uruguay and Peru each rise at least ten spots in the tournament rankings, once the stiffness of their opposition is factored into their qualifying records. Brazil leap from 15th to 2nd, just behind Germany, who retain top spot.
Switzerland, Iceland, Iran, Japan, Morocco and Australia’s superficially impressive qualifying records appear less so under a strength of schedule adjustment and the latter fall from the 5th best raw rating to 30th out of 31 overall.
Predicting the outcome of the 2018 World Cup
Once we have a more objective assessment of the recent form of the finalists, we can begin to analyse the makeup of the eight groups.
Monte Carlo simulations, using expected goals ratings to give a probabilistic prediction for the outcome of each potential game at the finals, can readily produce 1,000’s of possible outcomes for the possible scenarios over the month-long tournament.
The draw is seeded, but some groups will inevitably be more competitive than others and the possible routes to the final will have already been whittled down to a limited number of combinations for all sides.
Group F, consisting Germany (ranked 1st by adjusted xG differential from qualifying), Mexico (12th), South Korea (20th) and Sweden (16th) appears to have most strength in depth while Group’s A and H appear to be the weakest, on average.
Brazil (27%) and Germany (24%) has perhaps a surprisingly high chance of being eliminated as early as the first knockout round.
Group’s A, G and H are the three most competitive, with the at least two sides in each group appearing to be evenly matched to win the group. Little separates Uruguay an hosts Russia, Belgium and England or Colombia and Poland at the head of the respective groups.
Unsurprisingly, the likes of Germany (19% chance of winning the tournament), Brazil (15%), Spain (11%) and France (11%) emerge at the head of the markets under such an analysis.
But by simulating the entire tournament on a match by match basis, we can also make estimates about secondary markets, such as a side’s most likely stage of elimination or even how many goals they are likely to concede in the group stages.
Australia, for example has around an 85% chance of being eliminated at the group stage, a 10.5% chance of falling in the round of the last 16, 3.5% chance of being eliminated in the last eight and just a 1% chance of suffering semi-final heartache.
The quirks of the now predetermined draw that pairs groups together in the knockout round can also be investigated by such methods.
Brazil (27%) and Germany (24%) has perhaps a surprisingly high chance of being eliminated as early as the first knockout round, largely because there is a possibility that they may meet each other at this early stage of the tournament.
Over the next few months, I’ll be revisiting this model as the markets and World Cup squad makeups become known.