Sep 22 / Tim Klein

It's not an accident... RWC Injury Model

Rugby World Cup 2023

Modelling injury burden in the Rugby World Cup

The Springbok’s World Title defence of the Webb Ellis Cup is officially underway. Following a pre-tournament thrashing of New Zealand, a first game win over the fierce Scottish, and a subsequent thumping of Romania; the Bok hopes are high. However, we have not made it through these games unscathed. Unfortunately, the key figure of Malcolm Marx has been ruled out of the tournament due to a “freak accident” in training – leaving him injured. However, while unfortunate, losing players due to injury during a Rugby World Cup (RWC) campaign may not be unexpected.

In 2017, researcher Colin Fuller developed an algorithm for modelling a rugby team’s injury burden across an RWC campaign (1). A team’s injury burden can be defined as the number of players unable to train/play matches at a given time. The algorithm considers factors based on observations during the 2015 RWC. These factors include the team’s game and training loads, injury rate, and the rate of recovery from injury. The injury rates were 90.1 injuries per 1000 hours and, once injured, players typically returned to play 8 days later.

Using this data, he modelled the injury burden of a team’s RWC campaign – assuming the team won their group and progressed all the way to the final (Fig. 1).
Figure 1: Predicted injury-burden profiles for the teams with the earliest and latest start dates during the 2015 Rugby World Cup (Fuller et al., 2017)
Empty space, drag to resize

The algorithm predicts that by the final, a team can expect to have at least 2 players unavailable due to injury (red). Interestingly, it also shows that after 2 games, at least 1 player is expected to be out with injury (orange) – which happens to be the case for the Boks.

It is important to note, however, that the algorithm is not without limitations. Injuries arise from a complex interaction between internal and external factors (2). The algorithm, naturally, makes assumptions about some and fails to account for others. One example of a factor the algorithm fails to account for is previous injuries. Previous injuries are a strong predicting factor for future injury (2). This would affect the predictions about a team’s injury burden. Looking at the Bok team, for example, Siya Kolisi has made an incredible return from a partial ACL tear. This previous injury likely puts him at a higher risk of being injured during this campaign (i.e., affecting his individual injury rate). Siya might not be the only player in the team with a previous injury affecting their injury risk. So, the Boks may have to deal with more than 2 injured players should they reach the final. Other assumptions the algorithm makes is around the teams conditioning going into the tournament, that injury rates remain constant across games, and that recovery rates are consistent when there is likely variation between teams and individuals. Also, if we wanted to accurately compare this algorithm to the 2023 RWC, we would need to input updated injury and recovery rates (e.g., from the 2019 RWC). 
Algorithms are forced to make assumptions but, while limited, they do present interesting findings. In this case, the algorithm may highlight the importance of medical and support staff. Teams will likely have injuries during the campaign. Having the necessary support staff to cope with injuries and ensure a prompt return to play will be important. Also, it seems likely that teams will have players unavailable for some games – highlighting the importance of squad depth for success.
Empty space, drag to resize
  1. Fuller CW. A Kinetic Model Describing Injury-Burden in Team Sports. Sport Med. 2017 Dec 1;47(12):2641–51.
  2. Bahr R, medicine TKB journal of sports, 2005 undefined. Understanding injury mechanisms: a key component of preventing injuries in sport. bjsm.bmj.comR Bahr, T KrosshaugBritish J Sport Med 2005• [Internet]. 2005 [cited 2023 Sep 20];39:324–9. Available from:
Created with