Predicting the success of nations at the Summer Olympics using neural networks

doi:10.1016/S0305-0548(99)00003-9

Computers & Operations Research

Volume 26, Issue 13, November 1999, Pages 1243-1265

https://doi.org/10.1016/S0305-0548(99)00003-9 Get rights and content

Abstract

In this paper, we construct several models that try to predict a country’s success at the Summer Olympic Games. Our data set consists of total scores for over 271 sporting events for 195 countries that were represented at the 1996 Summer Games and information we gathered on 17 independent variables. We build linear regression models and neural network models and compare the predictions of both types of models. Overall, the best neural network model outperformed the best regression model.

Scope and purpose

Every four years, an enormous amount of attention is focused on the Summer Olympic Games. Sports fans and analysts do their best to predict the outcomes with respect to many sporting events and the overall performance of nations competing at the Olympics. In this paper, we develop and compare socio-economic-based models for predicting the success of nations at the Summer Olympics.

References (0)

Cited by (54)

Forecasting the Olympic medal distribution – A socioeconomic machine learning model
2022, Technological Forecasting and Social Change
Citation Excerpt :
First, by increasing the level of granularity beyond country-specifics; second, by including more years; and third, by exploring additional independent variables. As a common way to incorporate more granular data, and thus to increase the forecast accuracy, some authors considered predicting the Olympic success by focussing on different sports (e.g. Tcha and Pershin, 2003; Noland and Stahler, 2016a; Vagenas and Palaiothodorou, 2019), sometimes even exploring data on the level of the individual athlete (Condon et al., 1999; Johnson and Ali, 2004). Due to the increasing relevance of gender studies, other authors have begun differentiating their data sets by gender (Leeds and Leeds, 2012; Lowen et al., 2016; Noland and Stahler, 2016b).
Forecasting the number of Olympic medals for each nation is highly relevant for different stakeholders: Ex ante, sports betting companies can determine the odds while sponsors and media companies can allocate their resources to promising teams. Ex post, sports politicians and managers can benchmark the performance of their teams and evaluate the drivers of success. We apply machine learning, more specifically a two-staged Random Forest, to a dataset containing socioeconomic variables of 206 countries (1991–2020). For the first time, we outperform the more traditional naïve forecast for four consecutive Olympics between 2008 and 2020.
Multicultural managers and competitive advantage: Evidence from elite football teams
2019, International Business Review
Citation Excerpt :
Country population influences the size of the pool of available athletes for a team. Population is an important determinant of success in the Olympic Summer Games (Condon, Golden, & Wasil, 1999; Szymanski, 2000) and association football (Hoffmann, Ging, & Ramasamy, 2002). Gross domestic product per capita is an indicator of the economic potential of the country, which is necessary for developing elite sport talent.
Successful organizations have dynamic capabilities that are aligned with their competitive environments. As competitive environments globalize, the dynamic capabilities driven by managers with multicultural backgrounds are increasingly relevant. These managers are more likely to seize cognitively distant opportunities than monocultural managers. Using panel analysis of multilevel data from 355 elite football (soccer) teams, we find support for a moderated relationship among the degree of globalization in the competitive environment, the multicultural background of team managers, and team performance. When the competitive environment is highly global, teams with multicultural managers outperform teams with monocultural managers. When the competitive environment is less global, the reverse is true. The same effect holds when tested both cross-sectionally and longitudinally.
Quantifying individual performance in Cricket - A network analysis of batsmen and bowlers
2014, Physica A: Statistical Mechanics and its Applications
Quantifying individual performance in the game of Cricket is critical for team selection in International matches. The number of runs scored by batsmen and wickets taken by bowlers serves as a natural way of quantifying the performance of a cricketer. Traditionally the batsmen and bowlers are rated on their batting or bowling average respectively. However, in a game like Cricket it is always important the manner in which one scores the runs or claims a wicket. Scoring runs against a strong bowling line-up or delivering a brilliant performance against a team with a strong batting line-up deserves more credit. A player’s average is not able to capture this aspect of the game. In this paper we present a refined method to quantify the ‘quality’ of runs scored by a batsman or wickets taken by a bowler. We explore the application of Social Network Analysis (SNA) to rate the players in a team performance. We generate a directed and weighted network of batsmen–bowlers using the player-vs-player information available for Test cricket and ODI cricket. Additionally we generate a network of batsmen and bowlers based on the dismissal record of batsmen in the history of cricket—Test (1877–2011) and ODI (1971–2011). Our results show that M. Muralitharan is the most successful bowler in the history of Cricket. Our approach could potentially be applied in domestic matches to judge a player’s performance which in turn paves the way for a balanced team selection for International matches.
Olympic medals and demo-economic factors: Novel predictors, the ex-host effect, the exact role of team size, and the "population-GDP" model revisited
2012, Sport Management Review
Citation Excerpt :
As too complex regression models tend to increase variances in the coefficients and the predictions, while oversimplified ones suffer from biased coefficients and predictions (Myers, 1990, pp. 178–180), the main methodological problem in this area has been the determination of optimal specifications. When a large number of candidate correlates of Olympic success are explored, some of them end-up to simply be distal covariates of minor or no relevance, as for example correlates such as expected life span, death rate, number of airports, and total railway length found in study (Condon et al., 1999). On the other hand, there are cases of studies in which oversimplified models of Olympic success are proposed with only two predictors (population & GDP), as for example the multiplicative function of Morton (2002) and the ordered-logit model of Andreff (2001), or even with solely one predictor (GNP), as the simplest ever published model proposed by Nevill and Stead (2003).
The present study revisited the problem of estimating Olympic success by critical demo-economic indicators. The sample consisted of the 75 winner countries at the Athens 2004 Olympic Games (not previously analyzed). Medal totals were log-linearly regressed on land, population, GDP, urban population, inflation, growth rate, unemployment, labor force, health expenditures, ex-host, and team size. Multiple regression assumptions were tested with proper diagnostics including collinearity. Olympic team size was the best single predictors of Olympic medals (R² = 0.690, p < 0.001), and as an alternative criterion variable was significantly regressed on population, growth rate, health expenditure, and unemployment (R² = 0.563, p < 0.001). Medal totals were significantly regressed on population, ex-host, health expenditure, growth rate, and unemployment (R² = 0.541, p < 0.001). The classical population-GDP model extracted only 28% of the variance in total medals (R² = 0.277, p < 0.001), and this was slightly improved when combined with unemployment (R² = 0.365, p < 0.001). It appears that the size of the Olympic team plays the role of transmitting the composite impact of a country's size and economy to the end-phase of Olympic success. Winning Olympic medals depends on the combined potential of population, wealth, growth rate, unemployment, ex-host, and social-sport expenditures. Larger and wealthier countries win more medals by “producing” larger Olympic teams as a result of possessing more athletic talents and better support for social and sport related activities.
Socioeconomic variables and historical performance, the influences in the Olympic success at a sport-specific level, the case of judo
2023, Ido Movement for Culture
Assessment of Olympic performance in relation to economic, demographic, geographic, and social factors: quantile and Tobit approaches
2023, Economic Research-Ekonomska Istrazivanja

View all citing articles on Scopus

View full text