Measuring the effectiveness of offensive match-play in professional soccer

Abstract

The broader measures of offensive effectiveness, such as scoring opportunities and shots at goal, are commonly used as an alternative to goals scored due to the naturally low probability of scoring (about 1%) in soccer match-play. These measures may enable soccer practitioners to objectively see behind single match results, which are often influenced by chance. The purpose of this study was to examine the relationship between broader measures (scoring opportunities and score box possessions) and the ultimate measure (goals scored) of offensive effectiveness. We analysed data from videotapes of 163 of 182 (90%) matches played in the Norwegian men's professional league during the 2004 season. Multiple logistic regression analyses showed very similar results when comparing the effectiveness of different offensive tactics, regardless of which outcome was used. For example, counterattacks were more effective than elaborate attacks in producing goals (odds ratio OR=2.07, 95% confidence interval: 1.40 to 3.05), scoring opportunities (OR=2.30, 95% confidence interval: 1.28 to 4.15), and score box possessions (OR=2.12, 95% confidence interval: 1.39 to 3.25). The receiver-operating characteristic function statistical procedure was used to examine the association between each of the three measures of offensive effectiveness: scoring opportunities, score box possessions, and goals scored. No significant difference was observed between the area under the curve (AUC) for the broader measures (scoring opportunities and score box possessions) and the ultimate measure (goals scored) of offensive effectiveness. The 95% confidence interval of the AUC for both scoring opportunities (0.74–0.84) and score box possessions (0.68–0.76) includes the AUC for goals scored (0.74). Thus, the results are very similar regardless of which outcome measure for offensive effectiveness is used. This indicates that scoring opportunities and score box possessions (shooting opportunities) can be used as a proxy for goals scored when comparing the effectiveness of different playing tactics in soccer. Compared with goals scored, using scoring opportunities or score box possessions requires smaller match samples for meaningful analyses, and may therefore be more feasible alternatives.

Keywords:

• Match performance analysis in soccer
• logistic regression
• ROC curves
• goals scored
• scoring opportunities
• score box possessions