Analysis of Predicted Medal Tables

We have listed the predicted medal tables for all the Sumer Olympic Games since 2000 (see medal predictions for 2000, 2004, 2008, 2012 and 2016). In order to compare the success of these predictions, we have come up with an analysis method that can be used on all the data so far, to compare the predicted to actual results. Basically, this method calculates the average percentage difference between their prediction and the actual results for the final top five results.

Our results demonstrate a large range of success in predicting the top medal winning countries.

We have analyzed the predictions from the 2000, 2004, 2008, 2012 and 2016 Olympic Games, and the results range from a percentage accuracy score of 56% up to 95%.

2016 Predictions

In 2016 there were 25 different predictions to compare. Medal prediction was particularly difficult for the Rio Olympics due to the unknown effect of the banning of many of the Russian athletes. The timing of the predictions was important, as the knowledge of how many Russian athletes would be competing was not known until just before the Games started. Using our percent accuracy score, the accuracy ranged from 65% to 90%.

Total Medal Predictions

Below is the top 10 predictions for total medals won, the best (90% accuracy) was Bredtmann et al., a prediction based on economic models, though adjustment was made just prior to the Games to reallocate 24 Russian medals. Just behind was Brian Cazeneuve of Sports Illustrated (88%), who used recent event results for his prediction. Good results were again achieved by the econometric model of Goldman Sachs (87%).

rank percent accuracy score prediction model model method published date
1 90% Bredtmann et al. economic model Aug 2016
2 88% Cazeneuve-Sports Illustrated based on recent results 29 July 2016
=3 87% Goldman Sachs based on previous Olympic performance and economic growth July 2016
=3 87% Olympic medals predictions recent sports results 4 Aug 2016
=3 87% Gonzales (Tuck School) model incorporates four factors: available resources, population and per capita income, medals in the most recent Summer Olympics and a host effect July 2016
6 86% ATASS Sports probabilities based on recent sports results 5 Aug 2016
=7 83% Nielsen results from recent world championships in Olympic sport disciplines and world rankings 1 Aug 2016
=7 83% WSJ part subjective reporting, part statistics and part computer simulation 28 July 2016
9 82% Gracenote recent sports results 3 Aug 2016
=10 79% AOC Benchmark recent sports results 17 Dec 2014
=10 79% PwC factored in the size of economies, performance in the previous two Olympics; and host nation effect. June 2016

Total Gold Medal Predictions

The best prediction for total GOLD medals won was by the website 'Olympic medals predictions' (89%), which updated their prediction data the day before the Games began.

rank percent accuracy score prediction model model method published date
1 89% Olympic medals predictions recent sports results 4 Aug 2016
=2 82% Gracenote recent sports results 3 Aug 2016
=2 82% Gonzales (Tuck School) model incorporates four factors: available resources, population and per capita income, medals in the most recent Summer Olympics and a host effect July 2016
4 81% Nielsen results from recent world championships in Olympic sport disciplines and world rankings 1 Aug 2016
=5 78% Goldman Sachs based on previous Olympic performance and economic growth July 2016
=5 78% WSJ part subjective reporting, part statistics and part computer simulation 28 July 2016
7 77% Barra expert opion Dec 2015
8 76% ATASS Sports probabilities based on recent sports results 5 Aug 2016
9 68% Cazeneuve (SI) based on recent results 29 July 2016
10 70% Kuper et al. based on recent World Championships results, the number of athletes per country, and a host effect. 14 July 2016

2012 Predictions

In 2012 we compared 19 different predictions. Using our percent accuracy score, the accuracy ranged from 54% to 95%.

The best predictions for total medals won were by Brian Cazeneuve of Sports Illustrated, who used predictions of each medal event based on recent results (94%). Goldman Sachs were also very accurate (94%) - they did not use any previous sports results, using instead previous Olympic performance and economic growth of each country. The highest accuracy score was by the Wall Street Journal (95%), which uses both athlete results and computer simulation, obviously quite successfully.

rank percent accuracy score prediction model model method
1 95% Wall Street Journal part subjective reporting, part statistics and part computer simulation
=2 94% Brian Cazeneuve based on recent results
=2 94% Goldman Sachs based on previous Olympic performance and economic growth
=2 92% Williams (Tuck School) model incorporates four factors: available resources, population and per capita income, medals in the most recent Summer Olympics and a host effect
5 90% PwC factored in the size of economies, performance in the previous two Olympics; and host nation effect.
6 89% Sports Myriad recent results
=7 87% Infostrada Sports (now Gracenote) recent sports results
=7 87% Bredtmann economic model
9 85% Luciano Barra expert opion
10 81% Johnson factors: income per capita, population, "nation-specific cultural effect", and a host nation advantage

The most accurate prediction for total gold medals won was also by Goldman Sachs (87%).

rank percent accuracy score prediction model model method
1 87% Goldman Sachs based on previous Olympic performance and economic growth
2 83% Sports Myriad recent results
3 82% Williams model incorporates four factors: available resources, population and per capita income, medals in the most recent Summer Olympics and a host effect
4 81% Brian Cazeneuve based on recent results
=5 79% Wall Street Journal part subjective reporting, part statistics and part computer simulation
=5 79% Johnson factors: income per capita, population, "nation-specific cultural effect", and a host nation advantage
7 78% Infostrada Sports (now Gracenote) recent results
8 70% Luciano Barra expert opion

Summary

Looking at the results over the years, and comparing the different prediction methods, there is still no clear better way. Now we wait for the predictions for 2020 to start coming in.

Related Pages