The purpose of this literature review was to describe the state of rugby union performance analysis, highlight the various methods of analysis and explore variables used to assess performance. We have revealed that in the last two decades of rugby research, the approach to describing performance has remained largely unchanged. Investigations into successful performance typically continue to rely on univariate measures of performance, reducing performance to singular values (Table 3). In fact, 22 of the 41 studies retrieved focused on descriptive and comparative statistics and often lacked context. Confounding factors such as match venue, officials, weather and the nature of the opposing team have all been suggested to influence team performance, yet are rarely considered in the majority of the research [17]. This level of information details the origin of the data and arguably allows for more meaningful interpretations. Critical information may, therefore, be lost if performance-related variables are not contextualised and measured while considering these factors [44]. For instance, a major confounding factor is the opposition team yet only eight of the articles retrieved considered the opposing team in the analysis [10, 11, 14, 15, 28, 32, 33, 40]. More than half of the articles investigated successful and unsuccessful measures of performance by quantifying performance indicators over entire competitions. Although this approach is useful as a means to increase the number of data, this level of analysis ignores the variation in playing style over each match and typically lacks consideration of the influence of opposition. Ignoring data from the opposition will likely distort any relationships present [41], particularly when one considers that various studies included data over multiple competitions [3, 4, 6, 25, 38] as well as over several seasons [9, 21, 22, 25, 30, 34] potentially misrepresenting performance outcomes. One paper examined the efficacy of two methods of data analysis to predict match outcomes [41]; isolated performance indicators, considering only the isolated data from a single team, were compared to a descriptive conversion method by calculating the differences between each team’s data for each individual match. That study showed match outcomes were better predicted by relative data sets. Relative predictors of success included an effective kicking game, ball carrying abilities and not conceding penalties when the opposition are in possession.
Although the majority of the studies included contextualised results, it should be noted that some research included contextual information from multiple confounding factors such as pitch location, match period and team ranking. For example, a study of effective strategies at the ruck in the 2010 Six Nations Championship accounted for team ranking, pitch location and number of players involved [33]. The results indicated greater success in regaining possession with a higher ratio of defenders to attackers in ruck situations. Similarly, pitch location and the timing of ruck strategies influenced the outcome of ball possession in the 2011 Super Rugby competition [15]. Defending teams were more likely to turnover possession using an early counter ruck strategy in the wide attacking channels. Conversely, a jackal (a player on the defending team competing for the ball using his hands after a tackle was made but prior to the formation of a ruck) was the most effective strategy in the central field areas. Another study identified quick rucks within the first 20 min and within the 60–70 min time interval had the largest positive effect on match outcome [30], whereas slow rucks had the largest negative effect on winning a match, regardless of the time interval. These results highlight the importance of contextualising performance indicators, as game tactics may need to be adapted depending on the field location, time interval and ruck strategy employed.
Applying the outcome from research using simple, descriptive and isolated variables without consideration of confounding variables is problematic in tactical preparation. For example, set piece tries discriminated between successful and unsuccessful teams [28]; however, without contextual information such as score differential, weather conditions, pitch location or team ranking, little inference can be made regarding how or why behaviours occurred. One study [14] investigating defending strategies in tackle contact events which considered the playing situation, defensive characteristics and phase outcomes bore some insights into effective defensive processes such as defensive speed, field location and period within a match. This study demonstrated that the period of the match and the distance of the contact event in relation to the previous phase are key variables that predict the likelihood of a successful phase outcome. In a practical sense, teams execute different lineout plays depending on the field location (i.e. 5, 6, 7 man; they may play off the top or maul). They may also be more reluctant to throw the ball to the back of the lineout in poor weather conditions. On this basis, set piece selection is commonly dependent on context and, therefore, it is important to consider these factors when assessing performance indicators. Furthermore, analysing the performance of a team assumes that the behaviours in one game will provide insights into future performance in subsequent matches. The fundamental issue is that game behaviours may only specifically represent the performance of a team at the time the data were captured [45].
Performance Definitions and Indicators
Over 300 performance indicators were identified across 22 studies (Table 3). Interestingly, only 29 were identified as related to successful performance. International tests demonstrated 14 variables (Table 2) discriminating winning and losing teams including higher points scored, kicks, turnovers and penalties conceded between the opposition's 50- and 22-m line. In regional-level competitions, such as Super Rugby in the Southern Hemisphere, 25 variables were identified as successful indicators of performance including a greater number of metres gained, kicks out of hand, line breaks and percentage tackles made compared to losing teams. To illustrate differences in styles of play at different levels of competition, performance indicators that discriminated between winning and losing teams in international test matches and Super Rugby games were investigated [25]. Winners of Super Rugby games kicked more possessions, made more tackles, completed more passes and made less errors. No performance indicators were able to discriminate between winners and losers in international test matches played during 2003 and 2006 when only close matches were investigated (< 15 points difference) [22]. In contrast, another investigation of international games in the same time period showed that winning teams had higher points scoring-related statistics, turn overs and kicks and were more successful at set piece [22]. This discrepancy in outcomes may be a function of close games potentially being played by two opposing high-quality teams, demonstrating similar levels of performance behaviours. This continues to highlight the importance of contextualising performance indicators as vital information is likely to be lost when confounding factors are not considered.
There is typically a lack of transparency in the operational definitions used to describe and analyse rugby performance. Twenty-two retrieved articles quantified performance using performance indicators; however, only 7 actually defined the variables analysed. Furthermore, of the 22 articles, only 16 were explicit about the process of selecting the indicators used. The selection process included expert opinion and research group [1, 17, 21], commonly available statistics by a third-party company [22, 23, 25, 28, 38] and those selected solely by the research group [3, 18, 29, 39] (Table 3). The method used when selecting performance indicators in the remaining articles was undisclosed. Challenges may arise given a lack of clarity (i.e. lack of definitions or objectivity when selecting performance indicators) when comparing or replicating investigations, making it difficult to advance the body of research and for coaching staff to implement the suggested practices. However, a summary of the research and performance indicators relevant to successful performance can provide useful insights.
As mentioned earlier, performance indicators provide an overview of certain events that may contribute to and predict successful performance. However, isolated performance indicators do not consider the opposition, nor do they account for unpredictability and inherent match specificity. For example, game behaviours tend to be inconsistent and performance indicators will most likely be influenced by player-opponent interactions. It is therefore unlikely that a complex, dynamic game such as rugby can be represented by isolated measures of frequency data.
Evolution of Performance Assessment
Studies relating to attack are more common than investigations into defence (Table 1). Topics such as try scoring, possession duration and ball carries were investigated in relation to the attacking team, whereas tackle contest events and rucks were detailed as measures of defence. Most studies analysing performance indicators investigated both attack and defence situations. Specific investigations into defensive strategies only appeared from 2013 most likely related to rule changes [36] favouring the defensive team during breakdown situations.
To accommodate changing game styles, rule changes were introduced in rugby during 2007 and 2013 expediting the speed of play to increase appeal and competitiveness [36, 46]. The period prior to, during and thereafter should be considered and compared, understanding that successful performance indicators prior to 2007 may not be relevant thereafter. For example, amendments to laws surrounding the ruck led to a decrease in players involved in ruck situations [19]. Teams are instead favouring committing more players to the defensive line in preparation for subsequent phases. As a result, game actions have increased due to the added pressure on attacking teams to expedite the speed of play [36].
Between 2004 and 2007, winning teams won more lineouts on the opposition's throw, scored more tries, had greater metres gained, kicks out of hand, line breaks and percentage tackles made in international, Super Rugby and professional domestic competitions [17, 22, 23]. Successful teams also had higher points scored, conversions, successful drop goals, mauls won, line breaks, possession kicked, tackles completed and turnovers won. In contrast, losing teams lost more scrums and lineouts. Following this epoch, between 2007 and 2013, winning teams conceded more penalties between 50 m and opposition 22 m, and had more total kicks, including kicks out of hand, than losing teams. After 2013, variables likely to result in winning included higher average carry metres, clean breaks made and kicks made relative to the opposition in a professional domestic league. Negative outcomes were more likely when teams conceded penalties while the opposition was in possession. Data were considered in relation to the opposition rather than isolated data of each team considered discretely [41]. Isolated methods of analysis indicated winning teams missed less tackles in the Super Rugby competition [38]. Analysis of knockout stages of the Rugby World Cup, however, indicated that winning teams kicked a greater percentage of possession in the opposition 22–50 m and won more lineouts on the opposition ball [37], suggesting that successful test rugby may require a territory style of play. Performance indicators investigated were inconsistent across the studies, making it difficult to compare and assess the relevance and impact of key attacking and defensive variables. As such, although points scored were unrelated to match outcome post 2013 [41], it is problematic to suggest that point scoring is not important in rugby performance.
Factors such as competition location may rationalise the differing game styles observed. Approximately 20% of studies reported on Northern Hemisphere teams known to have a different style of play to [47] to Southern Hemisphere competitions. Southern Hemisphere teams tend to exhibit higher overall ball-in-play periods resulting in more game actions and injuries due to greater game continuity [47]. Additionally, ~ 40% of articles investigated teams competing in international competitions (Table 1) and 13% included data sets from multiple competitions, possibly decreasing their relevance as some information may be missed given the loss of contextual information [48]. Maintaining the integrity of each individual match when using the established descriptive conversion method of analysis, which considers all performance indicators in relation to the opposition, is preferred [41].
In summary, studies of performance analysis in rugby often show methodological shortcomings regarding the genesis of performance indicators and selection process, a lack of transparency and operational definitions with the investigated performance indicators and issues related to investigating performance indicators over entire competitions. The problems associated with investigating performance indicators without the consideration of contextual and situational factors limit the application of research outcomes into the rugby community.
Advancing Rugby Performance Analysis
There are some notable studies that have explored the performance processes in rugby union. Recently, researchers have used clustering approaches to identify important patterns in match data associated with certain game outcomes [35, 42]. These methods are useful for reducing large volumes of high-dimensional data to visualisable, low-dimensional output maps or identifying key playing patterns. One method identified that multiple game styles tended to result in success, such as a ball carrying, high-contact style of play. A low possession and strategic kicking style of play was observed to be just as effective. However, it is important to consider that data were not explored in relation to opposition game style for each specific match. This means that support for an ideal game style could not be established. Moreover, the level of competition analysed was low and restricted to a single nation. A K-modes cluster analysis was used to identify common playing patterns that preceded a try [42], suggesting plays following lineouts, scrums and kick receipts were common approaches to scoring tries in Super Rugby. A limitation to these approaches is the data related to collective team behaviour, such as player positioning and movements, were not collected in either of these studies.
Multiple studies have considered rugby union performance using a dynamical systems approach to analyse game characteristics [27, 32, 49,50,51,52,53,54,55]; however, to the authors’ knowledge, only three studies have used this approach in professional, male adult rugby union contexts [26, 27, 32]. In this approach, important characteristics of complexity are assessed by emergent patterns, due to the interactions between components in the system (i.e. players) over time [51]. This method has been found to successfully identify self-organising, emergent patterns from slight changes in interactions between players [56]. This suggests that players’ decisions and actions are governed not only by prior instruction provided by coaches, but by constraints in the player-environment interaction. In team sports, these behaviours emerge in space and continuously change over time, under the influence of constraints such as task (rules governing the game), environmental (weather) and individual constraints (physical capacity of the athlete) [57], resulting in the spontaneous reorganisations of intrapersonal and interpersonal coordination [58]. Some research has measured the constraining influences of one team on the opposing team’s playing system formation [32]. Attackers were observed to act as a coordinated sub-unit, measured through correlation values, accounting for distance and relative velocity values between each player within the sub-unit (two players from one team) [58]. When the sub-unit of the attacking team was able to disturb the coordination tendencies of the defending team’s sub-unit, this resulted in opportunities for the attacking team to cross the gain line (an imaginary line parallel to the score line, set between the attackers and defenders every time that attackers and defenders perform a ruck, maul, scrum or lineout [32]). However, when both sub-units remained equally coordinated, neither the attacking nor the defending team was successful in crossing the gain line or regaining possession of the ball, respectively. Small adjustments in players’ interpersonal distances and running line speed were considered useful tools to disturb the opponent’s coordination patterns. Using a similar approach, pass decisional behaviour was found to be predicted by the time-to-contact between the attacker and the defender [27]. The type of pass that emerged was significantly correlated (p < 0.001) with the variables available in the interaction between players and the environment, suggesting that intrateam coordination is necessary for crossing the gain line as well as effective passing in rugby union.
Capturing movements at the team level associated with successful attacking phases of play, such as advances in territory (achieving a more advanced position in the field of play), have additionally been explored in rugby union [26]. Investigating the multi-player sub-phases, ball displacement trajectory patterns were analysed, revealing the maximum distance the ball travelled backwards from a pass was lower in successful phases of attack. Greater advances in territory were additionally observed when lower backward movements of the ball were coupled with rapid ball delivery. Assessing the macroscopic order therefore suggests successful characteristics in collective behaviour patterns in attacking phases involve a fast ball delivery to a receiver within a close distance [26].
This constraint-led approach is commonly used in the field of skill acquisition and motor learning and proposes novel actions might emerge by manipulating key practice task constraints [51]. This approach has additionally been used to identify the interaction between the intrinsic dynamics and the external constraints within critical match events [27]. Examining the inter- and intrateam coordination patterns that influence successful performance may, therefore, yield critical insights into behaviours associated with successful match events, such as line breaks [22] and try scoring [42]. These methods have yet to be explored in international rugby union and should be addressed in future research.