Predict the 2014 World Cup Final with a Poisson analysis

Predict the 2014 World Cup Final with a Poisson analysis

By Mark Taylor Jul 10, 2014

Tweet

Bettors should read this 2014 World Cup Final betting Poisson analysis to help them predict the winner between Germany and Argentina. It explains HFA for the Americas, how not to overvalue the 7-1 outlier and the Poisson percentages for the World Cup final.

No assessment of Germany’s chances of lifting the World Cup on Sunday can begin without reference to their extraordinary 7-1 victory over the hosts, Brazil in Tuesday’s Semi- final. The pre-game odds to qualify for the final favoured neither side. Therefore, Germany were perceived as the stronger side, with the home and continental advantage that had been enjoyed by teams from the Americas bringing Brazil on par prior to the game.

The impact of HFA on the Americas

Following the Quarter-final matches, South American teams had an overall success rate over 90 minutes – where draws are counted as half wins – of 0.78 in matches against the rest of the world. The average FIFA ranking of the South American sides had been 1095 and that of their opponents was 898. By contrast European teams had recorded a success rate of just 0.57 in very similar matchups. The European teams, on average had been rated slightly lower at 1068, but so had their opponents at 856.

So given comparable tasks, South American teams over the tournament as a whole had performed better as a group and it was reasonable to assume that was at least partly down to a continental advantage and an additional home advantage in the case of Brazil.

Concacaf sides had recorded as success rate of 0.5, inferior to that of European teams, but achieved in much more difficult matchups, perhaps indicating that the continental premium had advantaged Central and North American teams, as well.

Don’t overvalue the Germany performance

These factors, along with the absence of key Brazilian players combined to make choosing a favourite difficult and in defeating an evenly matched opponent on the pitch by six clear goals, Germany delivered a result that might occur or be bettered once in 1400 attempts.

Further evidence of the unlikely nature of Germany’s progression to the final can be found in the shot location data for their matches prior to Tuesday. Goal scoring depends on the ability to generate shots and the ability to convert these chances. Success in this field is prone to luck driven factors, so the more extreme a team’s figures become, the more luck should be considered as a possible non repeatable cause.

Inclusive of blocked efforts and dependent upon source, Germany had scored 10 goals from 75 shots, including penalties, up to their semi-final date with Brazil. Using the shot location and type of attempt for all 75 efforts, based on goal expectations derived from competitive international games, a typical international team would have expected to score just over 8 goals from the 75 goal efforts made by Germany.

An average international side would expect to equal or better the German total of 10 goals around 30% of the time. So Germany, on this evidence had possibly been slightly lucky to score around 2 extra goals, or they had better finishers than was general for internationals, or most probably a combination of the two.

However, once we include the 14 shots from Tuesday night that resulted in seven goals, the figures change dramatically. Germany now has a goal expectation of just over 10 goals from 89 attempts, but an actual total of 17 goals. So their scoring rate over the tournament is 70% higher than expected if an average team had created and attempted the German’s 89 shots. Also, an average team would expect to equal or better 17 goals from such attempts only 1.6% of the time.

These figures describe perfectly the level of performance Germany had achieved during the World Cup, but are likely to over rate their level of future performance if they form a sizeable part of that evaluation.

A similar effect was to be seen in the Netherlands’ scoring rates prior to the semi-final game with Argentina. Their shots, headers and regular time penalties would usually yield on average seven goals, but Spain’s cavalier disregard for their own goal difference in the opening group game helped to contribute to an actual Dutch goal tally of 12.

Results such as these are therefore both rare and partly driven by factors that are unlikely to repeat with any regularity. Brazil’s meltdown, especially in the period following Muller’s opening strike from a corner is virtually unprecedented in top class international football. This presents problems when attempting to calculate credible scoring rates that may reflect future performance that could be used to project future match odds, for example by way of a Poisson process.

Poisson analysis

Larger data sets can dilute the effect of atypical score lines, such as Germany’s 7-1 win or indeed their 4-0 win against a depleted Portugal in the group stages, but such data may require multiple years of results, during which the team make up may change greatly, especially in competitive international football.

In all competitive matches from the start of Euro 2012 Germany scored an average of 2.9 goals per match and conceded 0.9 goals. Argentina over a similar number of competitive matches comprising the current World Cup and their qualifying process, scored an average of 1.95 and allowed 0.8.These raw numbers, when applied to a Poisson give Germany a 48% chance of winning on Sunday in 90 minutes, compared to a 28% chance for Argentina.

However, smoothing the data, to dampen the effect of atypical results does bring the two teams closer together. Smoothed data gives a 40% chance for Germany against 26% for their rivals, with the increased possibility of a draw reflecting the excellent recent defensive record of both teams.

Very similar win percentages are found if we use the goal expectations for each side based on the amount and location of the shots and headers they have created and faced. As we have seen, Germany has created attempts at the World Cup that would yield 10 goals. Argentina has created a similar amount of poorer quality chances, with a goal expectation of just over 6.

Defensively both teams have attempted to keep opponents shooting from distance and Germany’s four goals allowed tallies well with a location based expectation of just over five. Similarly Argentina has conceded shots that would on average expect to yield just over four goals compared to the three they have actually conceded.

Again translating these expected scoring and conceding rates to a Poisson based approach gives Germany a near 40% chance over 90 minutes, much more in keeping with the actual odds available.

Use of smoothed data or goal expectation based on shot volume and location, rather than luck based reality, often offers a better alternative to odds setting and both of these approaches generate probabilities that just fall short of finding value in any of the three 90 minute alternatives in the final.

However, the continental advantage that has been present in previous World Cups contested in the Americas does also seem to be present for South American teams in particular at Brazil 2014. This should give an additional boost to Argentina and while it won’t elevate them to favouritism to lift the trophy, it may well make them a value alternative in the result market at the current prices.

Click here to see the latest World Cup Final odds.

If you have feedback, comments or questions regarding this article, please email the author or send us a tweet on Twitter.