World Cup scores and predictive failure - why things never go as planned

Dr Nicolas Scelles

Posted: August 1, 2014

Before the World Cup, I made a model to explain national football team performance over the period 2011-2013. Then I applied coefficients found for the different determinants to World Cup matches. Based on my model, Brazil should have won in final against Argentina after having eliminated Germany in semi-final, Spain should have been in semi-final whereas the Netherlands and Costa Rica should have been eliminated at the end of the first round. These predictions were globally wrong. My model based on 2011-2013 did not work for the World Cup. To understand why, I incorporated the same determinants to explain scores during the World Cup in adding two variables: home disadvantage for Brazil against Germany and Netherlands. The table below shows that the coefficients are not the same in the model based on the World Cup compared to the model 2011-2013. Of course, we have to be careful with the interpretation of the coefficients for the 2014 World Cup that are based on only 64 matches. Some comments can me made all the same.

Variables Coefficients 2011-2013 Coefficients 2014 WC
Population 1.7554 1.4358
GDP per capita 0.4465 3.3711
Climate 0.0000 0.0000
Experience 1.5897 1.6544
Percentage of players 0.5164 13.2475
Player quality 0.0299 -0.0958
Foreign managers 0.2158 -0.9834
Technology transfer 0.1949 -0.6428
Home advantage 0.2730 0.9938
Home disadvantage Brazil-Germany - -4.3766
Home disadvantage Brazil-Netherlands - -1.6943
Prize 0.8419 0.6596
Prize difference / favourite team 1.7274 0.9866
Prize difference / underdog -0.7509 -2.1892
No prize 0.5284 4.9091
AFC -1.6060 -0.6369
CAF -0.9203 0.9765
CONCACAF -1.0669 -0.3355
CONMEBOL -0.2993
OFC -1.3629 -
UEFA -0.9712 -0.3228
Constant 0.0127 -0.1059
Observations 2854 64
Adjusted R² 0.328 0.564


The main interesting point is the impact of the percentage of players in a country on performance during the World Cup. Take a look at the table below. The four countries with more than 10% of players – Costa Rica, Germany, Chile and the Netherlands – performed well in spite of a limited population for Costa Rica, Chile and the Netherlands. Among the twelve countries with less than 6% of players, only Algeria and Nigeria reached the round of 16 and Algeria was with Russia and South Korea in the group stage whereas Nigeria was with Bosnia and Herzegovina and Iran. The limited number of observations during the World Cup makes difficult the possibility to know whether the percentage of players was really determining for performance or it was a coincidence.

Countries Percentage of players Population (in K inhabitants)
Costa Rica 22.2% 4,894
Germany 19.9% 81,862
Chile 14.7% 17,751
Netherlands 10.4% 16,719
Croatia 8.5% 4,270
Italy 8.3% 60,044
England 7.9% 53,000
USA 7.7% 318,744
Belgium 7.4% 10,981
Switzerland 7.3% 7,866
Uruguay 7.0% 3,443
Greece 6.9% 11,041
Mexico 6.8% 125,247
France 6.6% 63,969
Brazil 6.5% 202,307
Ecuador 6.4% 16,058
Argentina 6.3% 41,903
Colombia 6.2% 49,177
Spain 6.1% 46,378
Bosnia and Herzegovina 5.3% 3,816
Portugal 5.2% 10,506
Honduras 5.0% 8,353
Algeria 4.5% 39,782
Australia 4.2% 22,990
Russia 4.1% 140,930
Côte d’Ivoire 3.9% 20,646
Japan 3.8% 126,153
Nigeria 3.7% 177,934
Cameroon 3.4% 22,979
Iran 2.3% 78,466
South Korea 2.2% 48,991


Another interesting point is the negative sign for player quality which is the number of players among the 10 most valuable clubs according to Forbes. These clubs played a lot of matches during the season and their main players could have been tired or even injured (Cristiano Ronaldo, Ribéry, Neymar at the end of the quarter-final but he seemed to be in trouble with his ankle from the beginning of the competition).

Brazil had an advantage of one goal for playing at home until its quarter-final. Nevertheless, it had a “home disadvantage” of more than four goals against Germany in semi-final that could be associated to the absence of Neymar and Thiago Silva and – more related to home disadvantage – too much pressure and the fact that players freaked out after having been led 2-0 in a World Cup semi-final at home whereas they represented Brazil (remember that they conceded four goals in six minutes). Another home disadvantage is identified for Brazil-Netherlands that we could interpret as the consequence of the psychological trauma due to the large defeat against Germany.

At the end, it is important to note that the World Cup is a very uncertain event since it welcomes the best national teams (competitive balance) with only three matches per team during the group stage then a knockout stage. Predictions based on previous matches are particularly difficult since the World Cup incorporates a lot of inter-confederations matches whereas such matches during the period between the previous and the current World Cup are essentially friendly and teams do not automatically play at their best (Germany did not perform very well in inter-confederations matches before the World Cup). If making predictions before the World Cup was a lot of fun, I could predict that my predictions will be globally wrong!

About Dr Nicolas Scelles

Dr Nicolas Scelles is lecturer at the School of Sport, Stirling University, Scotland. He holds a PhD in sports economics from the University of Caen Basse-Normandie, France. He has articles in international journals including Applied Economics, Economics Bulletin, International Journal of Sport Finance and International Journal of Sport Management and Marketing.