Hartpury Student Research Journal

Home » A notational analysis of the scrum and its impact on team performance and end result in professional rugby union

A notational analysis of the scrum and its impact on team performance and end result in professional rugby union

Author Names: Cameron Avant (BSc (Hons) Sport Performance) and Dr Laurence Protheroe



Rugby Union stands out among other professional sports when understanding of the factors that affect successful performance are considered. Within most racket sports, and some team sports, there are numerous key performance indicators (KPI’s) available that can reliably provide insight into the winning team or athlete even within close competition. However, a similar successful model has not been created for Rugby Union.

This  lack of success is not due to an absence of research. Within this research some KPI’s are agreed whilst others are disputed. A common reoccurring argument is that Rugby Union is distinctively complex and has many variables influencing the KPI’s and impeding research. The purpose of this study was to aim to overcome and/or identify some of these obstructions to data collection through considering the scrum as a KPI and evaluating how the scrum influences the game of Rugby Union.

Twelve games were taken as a sample from the first two weeks of English Rugby Union Premiership. Notational analysis using sports coding software (Nacsport Scout Plus software) collected scrum KPI data considering the method of acquiring or losing possession, scrum duration, scrum quality and placement of scrum.

Frequency of using the ball won from the scrum as a kicking opportunity, quality of attacking scrum and scrum placement were found to be KPI’s with a significant effect on scrum and team performance, as well as the match’s concluding result. Through comparison of the results to other research it can be concluded that the scrum, and the specific factors therein, are influential upon performance and match result due to allowing teams to capitalise on the opportunity of space and structure to gain territory and field position.


1.0 Introduction

1.1 Background

Within competitive individual and team sports, preparation needs to be completed by the coaches, players and support staff. Part of this preparation is likely to be around tactics to guide decision making of players and coaches during performance. During games where scores remain similar and differences in performance between the competing teams are minor, coaches benefit from the most influential performance indicators (Csataljay et al., 2009). When within the practice and training environment, a clearer knowledge of position requirements can allow for specified and practical training for athlete development (Jones et al., 2015).


1.2 Research Justification

Alongside the increasing levels of professionalism within Rugby Union there is a rising demand for and value put on performance analysis. Coinciding with advancements in technologies and software there is increasing process efficiency. This provides the opportunity to perform research that was previously not possible to reliably conduct. With this there is the opportunity to locate and organise Key Performance Indicators (KPI’s) to construct profiles of individual and team performance. Whilst previous literature and research into Rugby Union analysis exists, a need for specific quantified study of the fundamental aspects of successful performance within rugby union remains (Hughes et al., 2012; Agnew, 2008).

Research that does not consider the scrum’s influence beyond its completion could be missing out on valuable data due to the unique qualities of the set piece to Rugby Union. The set piece offers the opportunity for predetermined structure and space which is not found as frequently with other methods of maintaining or losing possession such the ruck or maul. It is also an opportunity for defending teams to disrupt that structure (Correiaa et al., 2011; Westgate, 2007; Milburn, 1993).


1.3 Relevant Literature

Analysis within Basketball has identified many of the deciding KPI’s of the sport that remain consistent in close and high scoring matches. This has shown that factors contributing towards winning and losing are distinguishable and telling, even between equally paired teams. This has led to a progression in understanding of game determinants in basketball (Vaz, Rooyen and Sampaio, 2010). Within Rugby there have been similar efforts to catalogue contributing factors to successful performance in team and position formats. Position based KPI’s have been found to vary due to equally effective playing styles. Some factors that have been suggested as influential KPI’s included successful completion of scrums, turnover during opposition lineouts and rate of successful entry into the opposition 22 (Hughes et al., 2012; Vaz, Rooyen and Sampaio, 2010; Jones, Mellalieu and James 2004). Research focusing on team performance by Vaz, Van Rooyen and Sampaio (2010) has also found similar findings, especially concerning games of a final points difference of 15 or less. Variation in playing styles  between the northern and southern hemisphere teams was also suggested to contribute to the difficulties in collecting either significant or suggestive data. However, this study did discover the importance of kicking, tackle count and effective defending of territorial gains from effective kicks. It was also suggested that the location of the KPI’s occurrence could be important, not just the frequency.

Further KPI’s discovered to have a relationship with a winning performance include conceding minimal turnovers. In addition, suggestions by Vaz, Van Rooyen and Sampaio (2010) that losing possession in areas of the pitch where they are vulnerable to conceding impacts on the overall result were confirmed. Teams that invade the oppositions 22 frequently and regularly convert opposition 22 entry into points scored, are also more likely to have successful performances (Pavely et al., 2010; Van Ortega, Villarejo and Palao, 2009; Rooyen, Lambert and Noakes, 2006). Previous work endeavouring to produce a ranking of Rugby Union performance indicators placed tactical kicking above the set piece in importance (Pavely et al., 2010).

However, despite this past research, there has not been sufficient, definitive or consistent agreement in studies to allow formation of a comprehensive Rugby Union model of performance or outcome prediction of its own. This has been due to the game’s natural dynamic complexity in comparison to racket or other comprehensively researched team sports (Hughes et al., 2012, Reed and Hughes, 2006). For example; James, Mellalieu and Jones, (2011) reasoned that several profiles would be required for the data collection process to account for position, style, conditions and opposition proficiency.

Adaptation of game laws by governing bodies and the introduction of professionalism has also imposed or influenced changes in individual and team playing styles and capabilities. This may have reduced the impact of some performance factors while increasing the impact of others; possibly affecting the reliability of older pieces of research. The emendation of game laws has been found to increase safety and reduce cheating whilst increasing the time of the ball in play (Gianotti, et al., 2008; Williams, Hughes and O’Donoghue, 2005; Quarrie, Cantu and Chalmers, 2002). The introduction of professionalism has allowed for heavier, taller, and more experienced athletes, in addition to an increase in game speed and activity. Passes, rucks and lineouts have become more common since professionalism, whilst there has been a reduction in mauls, scrums and kicking (Sedeaud, et al., 2012; Quarrie and Hopkins, 2007; Eaves, Hughes and Lamb, 2005; Duthie, Pyne and Hooper, 2003; Eaves and Hughes, 2003).

It has been suggested by Hughes et al., (2012), Jones, Mellalieu and James (2004) and Vaz, Van Rooyen and Sampaio (2010) that further and more complex research is required. This research could potentially consider new variable combinations, definitive study of the issues collecting impactful data, and study of individual positions to form data for units’ analysis e.g. back-row.

Previous research by Ortega, Villarejo and Palao (2009) has demonstrated a correlation between a team performing effectively within the scrum and winning. Their research showed that winning teams maintained possession 90% of the time through scrummaging whilst losing teams lost scrum possession more often and forced scrum related turnovers less. Because of this, the Rugby Union scrum is a regularly assessed KPI by coaches, analysts and pundits. However, current data considering a relationship between successful team performance and scrum performance only has the scrum recorded as a won: lost ratio. This has resulted in no KPI’s being included of its immediate impact on the game. Any comparison made is likely taken from that ratio to the team’s overall success percentage. This has left an absence of data upon how the scrum immediately impacts the game and what scrum time behaviours are most effective (Ortega, Villarejo and Palao, 2009). For this reason, it is difficult and unreliable to judge its importance in comparison to other elements of the game that have also been shown to affect successful performance e.g. lineouts won or line breaks which will be important aspects of creating a prediction model of Rugby Union (Hughes and Franks, 2015, Fallowfield, Hale and Wilkinson, 2005). This lack of data is then often explained through study by subjective observations which has been dismissed as an unreliable and inaccurate form of review by several earlier studies such as Hughes and Franks (2015) and Agnew (2008).

An accurate ranking of the importance of the scrum to success has the potential to highlight areas of interest for stakeholders. Officials, governing bodies and team’ staff/players could be interested in such research. It also has the possibility of guiding stakeholders such as governing bodies in making informed judgements when evaluating against factors such as safety (Beck, 2016), entertainment or obstructions to precise officiating (The Telegraph 2016; Williams et al., 2005). The data could enable teams to restructure or amend their approach to training and games; either opting for specially trained and proficient players or focusing on the faster, generally more skilled players. The approach to attacking and defensive structure off the set piece could also evolve (Quarrie and Hopkins, 2007).


1.4 Project Aims

The overall aim of this study is to understand if, and how, the scrum influences the game of Rugby Union. To meet these aims, the following objectives of the study were considered:

  • To determine the effects of the KPI’s (Method of acquiring or losing possession, scrum duration, scrum quality, placement of scrum) on the scrums’ immediate impact and concluded play
  • To investigate how teams respond to the scrum from attacking and defending perspectives.
  • To determine if, and how, the scrum influences these responses.


1.5 Hypothesis

The hypothesis is that the scrum does influence immediate and overall team performance. Teams that are effective at maintaining their own ball while achieving possession of the oppositions’ through the scrum increase opportunities to attack with increased space and structure, thus increasing their scoring opportunities while reducing the oppositions. The null hypothesis is that the scumr does not influence immediate and overall team performance.


2.0 Methodology

2.1 Design

Nacsport Scout Plus software (AnalysisPro Ltd, 2014) was used for the post-event analysis of English Premiership Rugby Union matches during the 2016-2017 season. Scrum relative KPI’s were selected and authenticated with the advice of professional Rugby Union analysts. These KPI’s were identified as being elements of scrum performance, and possible influences on behaviour and resulting consequences. Initially the performance indicators were compiled using past performance analysis research of the Rugby Union scrum and the experience of external Rugby Union analysts. The list of KPI’s was then honed through the elimination of those considered unmeasurable or not relevant to the research aims. Once the list of KPI’s (Table 1) was finalised, a definition  and grading (if applicable) was assigned to each performance indicator.


2.2 Sample

Matches were selected from the first two weeks of domestic matches within the English Premiership from the 2016/2017 season (n = 12). Originally the study planned to use 24 games however a reduced sample was necessitated due to the unforeseen extension of the pilot study period. Only 1 team was analysed per fixture and only home games were selected (Nevill and Holder, 1999).  Before conducting the research, ethical approval was granted by University Centre Hartpury. Consent was lso granted by Gloucester RFC to use their database of match footage.


2.3 Procedure

A coding template/window was designed using Nacsport Scout Plus software (AnalysisPro Ltd, 2014) containing the criteria to collect data on every listed KPI. The template was used during the coding process to observe and collect data. Due to software limitations, scrum duration was calculated externally with a Sportline 240 Econsport Stopwatch (Sportline, 2007). Data was recorded immediately through predetermined categories within the Nacsport coding window. Upon completion of the data collection process data was directly exported from the Nacsport software matrix into Microsoft Excel (Microsoft Corporation, 2016) format to facilitate accuracy and minimisation of human error. Data was then compiled into SPSS Statistics v23.0 (IBM, 2015) software. SPSS was used to collectively analyse the data from the 12 games once all data had been collected.

For reliability purposes a pilot study was conducted before beginning the data collection (Lancaster, Dodd and Williamson, 2004; Van Teijlingen and Hundley, 2002). The pilot study consisted of a randomly selected match being viewed and data collected for three times over 2 days. The same analyst and coding template was used for each test. The pilot study indicated that there was some inaccuracy in data collection and also that data was produced that was not specific to the study aims. Alterations were made to the coding template and procedural guidelines which were retested using an identical pilot study. Percentage error calculations during the 4th pilot study demonstrated 100% consistency between the recorded data (O’Donoghue, 2012). The data collection process used the template and procedure from the final successful pilot study.


Table 1. Definitions of Key Performance Indicators.

Key Performance Indicators Definition
1 Zone Positioning


In relation to the length of the pitch, where the event is taking place
2 Attacking Scrum The team being analysed has the advantage of feeding the ball into the scrum
3 Defending Scrum The opposition of the team being analysed has the advantage of feeding the ball into the scrum
4 Scrum Won


Ball has been successfully turned over on opposition’s feed advantage
5 Scrum Lost


Ball possession has been lost on own feed advantage
6 Ball Maintained


Possession of ball has been retained via the scrum with own feed advantage
7 Opposition Maintains Ball Opposition has retained possession of the ball via the scrum with the feed advantage
8 Attacking Scrum Quality Rated 1-5, how well the attacking scrum was performed (not ranked same as defending scrum)
9 Defending Scrum Quality Rated 1-5, how well the defending scrum was performed (not ranked same as attacking scrum)
10 Duration of ball in scrum


How long the ball is kept in the scrum after the ball is released by the Scrum Half
11 Use of Ball Whether the ball is carried or kicked after it is played by the No 8 or the backline
12 Kicked Whether team in possession immediately kicks the ball before reaching game line or first breakdown
13 Carried Whether team in possession only utilises running or passing before reaching game line or first breakdown
14 Channel of Attack


Where the ball is carried to before first breakdown
15 Channel of Defence


Where the opposition carry the ball to before first contact
16 Immediate Metres Made from Scrum


Distance covered before first breakdown in relation to game line
17 Overall Metres Made from Scrum


In relation to game line where did play conclude
18 Result of Play How the play that originated from the scrum concluded


2.4 Data Analysis

The threshold for significance was set at p=.05 (Salkind, 2017). Data were presented in mode average format for the ranking of nominal data or represented as an average mean for scale and ordinal data (Salkind, 2017). Data normality was tested through the Shapiro-Wilk test and data was found to be non-parametric. Consequently, a non-parametric approach was applied. The Chi Square test was performed upon categorical variables; for example, the testing of statistical differences between channel of attack and scrum turnover. The Mann Whitney U test was carried out upon ordinal or scale data when the independent variable consisted of just two levels, for instance frequency of scrums awarded and the effect upon winning or losing. The Kruskal Wallis test was implemented for the analysis of scrum quality, metres made and other ordinal or scale data when the independent variable contained more the 2 levels (McKnight and Najab, 2010a; McKnight and Najab, 2010b). Data from when games were drawn (n = 1) were omitted from won/lost comparisons but included in data analysis not concerning the final score.


3.0 Results

3.1 General Scrum Results

No significant results were found for frequency of scrums per match (n = 12), possession turnover and maintaining scrum possession. Differences between maintaining scrum possession and end result were noticeably small at 2.17% and again no significant findings were obtained (Table 2). Significant effects were found when considering scrum being the most common method of maintaining possession, whilst forcing scrum related penalties is the most efficient method for turnover (< 0.05). However resulting play KPI’s, including metres made, are not significant (Table 3 and Table 4).


Table 2. Scrum Frequencies and End Result

Team Performance Indicators Mean P
1 Scrums Per Game


16.25 0.364
Scrums Per Game Without Resets


2 Winning Team Highest Awarded Scrum Frequency Difference (%)


+ 22.60


Losing Team Highest Awarded Scrum Frequency Difference (%)


+ 20.70
3 Winning Team Maintaining Scrum Possession (%)


86.79 0.699
Losing Team Maintaining Scrum Possession (%)


4 Winning Team Scrum Turnovers Won (%)


15.38 0.733
Losing Team Scrum Turnovers Won (%)




Table 3. Results of Maintained and Turnover Scrum Possession (n = 126).

Performance Indicator Mode Average P
1 Method of Possession Maintained (%)


1. Scrum

2. Penalty

3. Free Kick





< 0.05
Method of Possession Turnover (%)


1. Penalty

2. Free Kick

3. Scrum





2 Possession Maintained: Use of Ball (%)

(n = 82)


1. Carried

2. Kicked




Possession Turnover:  Use of Ball (%)

(n = 4)


1. Carried

2. Kicked




3 Possession Maintained: Channel of Attack (%)


1. 8

2. Centre






Possession Turnover: Channel of Attack (%)


1. Wide

2. 8

3. 10





4 Possession Maintained: Result of Play (%) 1. Ball in Touch

2. Opposition Turnover

3. Penalty Awarded








Possession Turnover: Result of Play (%)


1. Penalty Awarded

2. Penalty






Table 4. Immediate and Overall Metres Made from Maintained and Turnover Scrum Possession.

Immediate Metres Made p Overall Metres Made p
Maintained Scrum Possession



SD ± 14.66

0.851 14.72

SD ± 17.12

Turnover Scrum Possession 21.25

SD ± 30.65


SD ± 18.48



3.2 Defending Scrum

Significance effects were found for the use of ball (p = 0.028) and result of play (p = 0.032). Winning teams defend carries more on average at 91.30% than losing teams who must defend the kick with increased regularity. However, carrying the ball (73.91%) remains the mode used. Teams that are commonly awarded penalties (19.57%) whilst defending the scrum are more likely to be the successful side whilst losing defensive scrum performances result in conceding turnovers of possession (21.74%). No significnt findings were seen for metres made (Table 6) and channel of attack (Table 5).


Table 5. Defending Scrum Results (n = 92).


Key Performance Indicator

Mode Average
Winning Teams Losing Teams P
1 Use of Ball (%)


1. Carried (n = 42)

2. Kicked (n = 4)




1. Carried (n = 34)

2. Kicked (n = 12)



2 Channel of Attack (%) 1. 8

2. Centre

3. Wide




1. Wide

2. 8

3. 10

4. Centre






3 Result of Play (%) 1. Penalty Awarded

2. Ball in Touch

3.Scrum Penalty Awarded




1. Turnover Conceded

2. Ball in Touch

3. Turnover Won






Table 6. Metres Conceded from Defending Scrum.

Metres Made Winning Teams Losing Teams P
1 Immediate Metres 10.92

SD ± 14.18



SD ± 12.32

2 Overall Metres 15.13

SD ± 19.25


SD ± 17.80



3.3 Specific Scrum Results

Attacking scrum quality (Table 7) shows significant effects for the use of ball (p = 0.042) and result of play (p = 0.022). Attacking scrum quality did not significantly influence the channel of attack (p = 0.059) but does show a trend towards significance. No significant effects were found for distance gained, scrum quality (Table 9), and defending scrum quality (Table 8). Scrum Duration (Table 10) and distance gained (Table 11) produced no significant results.

Although alpha levels were not met scrum placement and influence on use of ball produced a trend towards significance at p = 0.051 (Table 12). 100% of ball used is carried in the opposition’s 22 and kicking frequency is at its highest in Zone “C” (36.67%). Winning and losing comparisons of scrum zone demonstrated significant effects for each zone excluding “A” that had no significant effects (P = 1.000). Teams that had the majority of their scrums in their opposition’s half (p = 0.017) or “B” Zone (p = 0.030) are more likely to win. Teams less likely to win have the majority of scrums in their own half (p = 0.017) or the largest percentage in their “C” Zone (Table 13).


Table 7 Cam Avant

Table 8 Cam Avant

Table 9 Cam Avant

Table 10 Cam Avant

Table 11 Cam Avant

Table 12 Cam Avant

Table 13 Cam Avant


4.0 Discussion

4.1 General Scrum Results

Broad scrum related data was collected despite acknowledging that aspects of this research already exists in previous studies (Pavely et al., 2010; Ortega, Villarejo and Palao, 2009; Treadwell, 1988). This was to provide baseline data and facilitate reliable comparison. If this data collected aligns with findings identified in other relevant studies this could suggest that other data sets within the study may also be valid (Reilly, Zbrozek and Dukes, 1993). In addition, comparison of this data with that of older studies could help substantiate suggestions made within the literature review that data preceding major rules change and professionalism could be less reliable (Sedeaud, et al., 2012; Gianotti, et al., 2008; Quarrie and Hopkins, 2007; Williams, Hughes and O’Donoghue, 2005; Eaves, Hughes and Lamb, 2005; Duthie, Pyne and Hooper, 2003, Eaves and Hughes, 2003; Quarrie, Cantu and Chalmers, 2002), further justifying the research undertaken.

These data showed no significant effects (Table 2) which is contradictory to those of Ortega, Villarejo and Palao (2009) and their findings regarding maintaining scrum possession and winning. Reasons for these differing findings could be due to changes in game structure in the previous 8 years or this study’s limited sample size (Button, et al., 2013). Although not significant, the mean of scrums per game at 16.25 is a considerable reduction on the 40 scrums per game mean recorded by Treadwell (1988), which predates the sanctioning of professional Rugby Union in 1995 (Eaves and Hughes, 2003). This could be suggested as further evidence that professionalism and the subsequent rule changes render older pieces of notational analysis obsolete in part or in their entirety.

Metres made produced results with no significant effects consistently throughout with large ranges of standard deviation (Table 4, 6, 9 and 11) regardless of the partnered variable. This proposes that the scrum is inconsistent and unreliable as a guarantor of successful or negative gain. A possible reason for this is the increase in defensive and attacking variables. Immediately as the ball is played from the scrum,decisions are influenced and manipulated by the structure of the attacking and defending backline demonstrated in Figure. 1, and 2. These original structures were likely to have been formed pre-game based upon how each team pre-emptively views their own and the opposition’s strengths and weaknesses (Evert, 2006). However, pre-determined structures have the possibility of adapting further during performance when more current information arises (O’Donoghue, 2010).

Figure 1 Cam Avant

Fig. 1 Examples of attacking indicators and options available to be considered by both the attacking and defending team (Evert, 2006).


Figure 2 Cam Avant

Fig. 2 Example of defensive structure that an attacking team may have to recognise and react to. The defending team drift across to defend their next opposing player outside them, freeing up the fullback and blindside winger to cover kicks or resulting defensive mismatches (Evert, 2006).


Once both teams have revealed features of what their attacking and defensive efforts will entail for the current play, both teams will make physical and cognitive efforts to manipulate, nullify and capitalise on the other team using physical force and athleticism while baiting opposition players to their lose shape, attacking options or intended ball carrier (Westgate, 2007; Evert, 2006).  This consequently produces four tiers of decision making independent from scrum behaviours thus reducing its direct influence over distances made. This is also with the exclusion of other external influencers such as officiating and technical accomplishments/failures of players (Wheeler and Sayers, 2009; Evert, 2006). This provides a further example of the complexities faced in the process of forming a performance model of Rugby Union (Hughes et al., 2012; Reed and Hughes, 2006).

Characteristics of scrum possession maintained and turned over were included within the collection of general scrum results to discover the best approaches. This was also used to consider whether there was an element of chaos to turnover scrum ball and whether teams could adjust to eliminate or capitalise upon the opportunities. A significant effect was found in the method of maintained and turned over possession (Table 3). Using the scrum to cleanly win possession as per design is the most common way to maintain possession from an own feed advantage (64.57%). Turnovers are more likely to be attained through being awarded penalties (52.17%). There were no other significant results or results of interest. This finding differs from that of Van Rooyen, Lambert and Noakes (2003) who found that maintaining scrum possession was a KPI to a winning result. There is no evidence to suggest that there is a suspected advantage of lacking defensive structure from turnover ball for the team now in possession.


4.2 The Defending Scrum

The defending scrum provides data collected from the perspective of the defending team. Defending turnover results were dismissed unless they were gained through a scrum related penalty or free kick. The analysis is not inclusive of turnover data due to the possibility of results being skewed by data occurring outside of the structure of maintained scrum possession. This was deemed important to this study as an accessory to the possession maintained and turnover KPI’s. Data collection techniques that isolated variables on the defensive side of the ball highlighted not only what methods may be successful in limiting the opposition’s capabilities but also what attacking trends may not be as efficient. This allows results to be obtained that might not have been available or as obvious from considering the attacking perspectives, as is commonly featured within general scrum results.

Significant results were found within the opposition use of ball for winning and losing teams (p = 0.028), in addition to the mode averages of the result of play (p = 0.028). Teams that won games were carried against more frequently (91.30%) than losing teams (73.91%) that had to defend an increase in kicking. These results coincide with those of Vaz, Van Rooyen and Sampaio (2010) and Pavely et al., (2010)’s findings of the importance of tactical kicking.

Winning defensive performances mostly deny possession to opposition. The top three occurrences for result of play all result in the opposition losing possession of the ball through penalties (scrum: 15.22% open play: 19.57%) or ball out of bounds. Losing teams also earn defensive turnovers (Ball in touch: 17.39% Turnover won: 13.04%). However, they also concede turnovers. For this to be possible they would have to have kicked possession, as other methods of exchanging ball possession would have indicated the coding process of the play in question. Therefore, losing teams are not regularly successfully counter attacking from kicks, often resulting in conceding turnovers themselves. Once again this fits within the findings of Vaz, Van Rooyen and Sampaio (2010) whose research states that winning teams are not just effective tactical kickers but also efficient in defending the territory gained. Through the subtraction of the percentage of turnovers conceded (21.74%) from the percentage of kicks defended (26.09%) this gives the resulting score of 4.35% which is the mode average that losing defending teams gain live possession without consequently losing possession shortly after.


4.3 Scrum Quality 

Scrum quality has been included in this study as a KPI due to literature referring to the scrum as an opportunity to play with pre-decided structure and stability which can only be found in Rugby Union from the set piece (Milburn, 1993). Scrum quality rates the standard and stability of the scrum from 1 – 5. One represents a poor scrum performance with either a fast retreating scrum or collapsing, whilst 5 represents a dominant scrum, demonstrating a capability to earn secure ball possession while forcing the opposition scrum backwards. Integers in-between offer ranked degrees of differentiation between the highest and lowest scores.

Attacking quality produced significant results throughout the tested variables. Kicking as a use of ball decreased as attacking scrum quality increased. Attacking down 8 (20.00%), 9 (60.00%) and 10 (20.00%) channels is increased in probability as a result of lower attacking quality. As scrum quality increases so does the tendency to use wider channels (p = 0.025). No results regarding ball in play are available for grade 1 as the poor attacking scrum quality would resolve in loss of possession. Result of play is shown to coincide with scrum quality. Increases in scrum quality results in increasing occurrence of positive results (p = 0.022). However, attacking scrum quality 5 features “opposition turnover” (14.29%) which would be categorised as a negative and undesired result of play to the attacking team. No significant results from defending scrum quality were found (Table 8).

The rationale for the increase in carry percentage and use of width is not clear and any suggested explanation would currently be hypothetical. Therefore, further research is needed within this area. It is also not clear how defending scrum quality is not as consistently instrumental as attacking scrum quality.

Due to attacking scrum quality producing significant results for “result of play” it can be concluded that structure and a stable base to attack from is influential upon the execution of skills. It also suggests that Milburn, (1993) and Westgate, (2007) are correct in referencing the scrum as the opportunity for said structure. By comparing these results to the research aims this indicates that the scrum does have an influence upon performance. It also indicates that any opportunity for “clean ball” and a stable base to play from increases the likelihood of a positive result.


4.4 Scrum Duration

Analysis of durations of any aspect of rugby union is very limited, minus that of ball in play or possession (Quarrie and Hopkins, 2007). Even less analysis exists concerning the actual impact and meaning of the duration. Despite this there are still infrequent claims that the process of slowing down the opposition ball is a worthwhile tactic and an opportunity to reorganise (Westgate, 2007). Rucks and mauls are typically the focus when duration is briefly discussed. Nevertheless, scrum duration could have been key in indicating influential elements previously not researched.

No significant results were found for any of the dependant variables (Table 10 and Table 11). Although no significant results or trends were seen among the average modes it was important for the thoroughness of the study to include even the areas with minimal understanding. There were also some approaches to the collection and presentation of the data that in hindsight could have been improved to develop a truer analytical portrait of the influence of scrum duration, such as the sample size.


4.5 Scrum Placement 

Arguably the most commonly agreed upon KPI for successful performance within the literature is field position (Pavely et al., 2010; Vaz, Van Rooyen and Sampaio, 2010; Ortega, Villarejo and Palao, 2009;  Van Rooyen, Lambert and Noakes, 2006). This is regularly presented in the varying formats of opposition 22 entry frequency or just time spent within the opposition’s half. Teams that played the majority of the game within the opposition’s half tended to have the more successful performance. Therefore, to truly answer the research question the scrum must be analysed against field position variables. Once again successful tactical kicking tactics predominately features as a common theme within territory literature.

During the data collection process the pitch was portioned off into 4 zones and labelled A, B, C and D. The “A Zone” is the opposition’s 22 while the “B Zone” signifies the area within the opposition’s half between the 22m and half way. The “C Zone” is within the analysed team half between halfway and the 22. “D Zone” is the analysed team 22.

Results from the analysed data concerning mean scrum placement and winning/losing (Table 13) found significant effects for zone B (p = 0.030), C (p = 0.017), D (p = 0.009) as well as the analysed and opposition’s halves (p = 0.017). The insignificant result from Zone A is likely due to the small amount of data (n = 13) making it vulnerable to skewed results (Button, et al., 2013). Other variables that have the capacity to make results fluctuate may also increase within these zones such as defensive tactics or urgency. Influence of scrum placement on use of ball (Table 12) showed a trend towards significance. This is above the set limit for significance but due to the minimal difference it will still be considered as a performance indicator of potential importance.

Influence of scrum placement on use of ball (Table 12) finds that the majority of kicking takes places within the attacking side’s half. There is no record to show any example of kicking off the scrum during the A zone (n = 0). There is a feasible rationale to explain the decrease in using the scrum as a kicking platform as progression towards the try line is achieved. As previously mentioned the scrum is often approached by an attacking back line as a unique attacking opportunity. When getting nearer to the opposition try line there is less ground to be made and increased opportunity to score, therefore the element of risk and unpredictability associated with kicking in Rugby Union at the cost of wasting the attacking structure outweighs territorial gain. Thus, the increased security and insurance of carrying the ball is more suitable for this portion of the pitch.

Scrum zone frequency (Table 13) as a KPI for rationalising team performance and result is feasible because of the scrum acting as a method for possession retention. Teams with the “put in” or feed advantage are a lot more likely to win possession back than the opposing team. Therefore, scrum zone frequency is telling of where winning teams play  the majority of their possession. Again, this coincides with the conclusions of Pavely et al., (2010) and Van Rooyen, Lambert and Noakes, (2006).


4.6 Limitations of Study

This study was largely successful in identifying areas of importance and interest concerning the scrum. This supplies conclusions that verify and build upon previous literature. There were also new conclusions within areas of limited previous research which could be advantageous to teams, governing bodies or even fans of the sport.

Despite this there are however limitations of this study. Increased sample size would have benefited the validity of the research (Button, et al., 2013) as for some specific variables the population count was very small for comparison, especially when statistical tests such as chi-squared that are vulnerable to small sample sizes are utilized (Lewis and Burke, 1949). Comparisons of scrum duration would have had increased efficacy if a p value was produced for each individual category of duration. Instead a p value was shared so the impact of each duration was not fully determined, perhaps missing out on an opportunity to discover the superior period of scrum duration for corresponding performance.

Global Positioning Systems (GPS) would have increased accuracy during the collection of data concerning metres made. Instead pitch markings were used which raises the possibility of human error impacting upon the results. If this process was to be repeated the use of GPS would be useful.


5.0 Conclusion

This study did find significant results and and findings of interest within scrum placement and quality. This was despite issues evident within previous research due to Rugby Unions’ complexity from a notational analysis perspective. These issues were also experienced during the study but were largely overcome due to the specificity of the project aims and data collection. Metres made was consistently an area where standard deviations and p values suggested it was not consistent enough to be of value to the aims of the study.

Discovering the most common methods of successfully maintaining and turning over possession does not have great use as an indicator of successful performance. However, it could be used by coaches and players to review their scrummaging tactics whilst additionally predicting their opposition’s tactics. Another result worthy of consideration but of no direct use to the conclusion of the study is the reduction in kicking from scrum possession as teams progress towards the latter zones of the pitch (p = .051). The result shows a trend towards significance and whilst not significant could be of use to players and coaches in structuring their defensive approaches dependent upon the location of the scrum.

Effects between using the scrum as a platform to kick and winning is the first result that could be applied to a model of performance predication. Teams that were found to kick immediately from the scrum at an increased frequency than their opposition were more likely to win. Other factors such as defending the counter attack from the kick are also likely to be important.

Attacking scrum quality is another KPI that showed its potential to be a worthwhile factor of consideration when judging performance and predicting result. The ability to play the ball wide, and produce positive results to play (e.g. try scored or penalty awarded) increased with the quality of the attacking scrum. This has been hypothesised to be due to structure granted by a stable or progressing scrum that simultaneously limits or reduces the defending opposition.

A majority of scrum placement within the opposition’s half and limiting frequency of scrums within their own half were observed to be directly indicative of winning performances (p = .017). This is theorised to be due to this limiting opportunity of conceding turnover in areas more likely to consequently concede. Scrum frequency and placement also explains possession and territory, both gained by winning scrums within the opposition’s half of the pitch. Possession and territory have been shown in previous research to be a KPI consistent with winning teams.

Using results found in this study alongside the relevant literature, this largely supports the hypothesis stated at the conception of the study that the scrum does influence immediate and overall team performance. However, the idea that teams that are effective at maintaining their own ball while achieving possession of the oppositions’ through the scrum increase opportunities to attack with increased space and structure, thus increasing their scoring opportunities while reducing the opposition’s is not accepted. Most of this statement was found to be correct, however the scrum as an instrument of turnover is not what makes it a tool for performance. The difference in scrum turnover frequency between winning and losing sides was identified as small (2.17%) and not of significance (p = .699). Although the scrum has demonstrated its influence on both immediate and overall team performance throughout the study, its influence lies in the opportunity of territorial gains, space and structure.

The implications of the study are that teams can use the KPI’s and results of interest recognise the possibility of scrum related events occurring, as well as specifying tactics to capitalise upon the influential factors discovered, such as the scrum as platform for tactical kicking. With the evidence that the scrum is a dynamic and influential method to reintroduce the ball into play, governing bodies can consider this information against arguments for the adaptation or replacement of the scrum. Areas for future research involve building off from this study as a foundation. Future research into scrum duration with more suitable data analysis could find areas of importance missed during this study. Assessing other elements of rugby union such as the lineout, with the same specific and detail driven approach, could highlight other areas of importance not currently identified while also creating a format to compare the set pieces and impacts and ranking their importance.



Agnew, M. (2008) Game analysis in Rugby Union. MSc, Auckland University of Technology.

Beck, J, (2016) Imaging the cervical spine following rugby related injury. Radiography [Online]. 22 (2), pp. 199-202. [Accessed 10 October 2016].

Lewis, D. & Burke, C. J. The use and misuse of the chi-square test. Psychological Bulletin, Vol 46(6), Nov 1949, 433-489.

Button, K., Loannidis, J., Mokrysz, c., Nosek, B., Flint, J., Robinson, E. and Munafo, M. (2013) Power failure: why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience [Online]. 14, pp. 365-376. [Accessed 5th February 2017].

Correiaa, V., Araujoa, D., Craigb, C. and Passosa, P. (2011) Prospective Information for Pass Decisional Behaviour in Rugby Union. Human Movement Science [Online]. 30 (5) pp. 984-997. [Accessed 01 March 2017]

Csataljay, G., O’Donoghue, P., Hughes, M. and Dancs, H. (2009) Performance indicators that distinguish winning and losing teams in basketball. International Journal of Performance Analysis in Sport [online]. 9 (1), pp. 60-66. [Accessed 03 February 2017].

Duthie, G., Pyne, D. and Hooper, S. (2003) Applied Physiology and Game Analysis of Rugby Union. Journal of Sports Medicine [Online]. 33 (13), pp 973–991. [Accessed 03 February 2017]

Eaves, S. and Hughes, M. (2003) Patterns of play of international rugby union teams before and after the introduction of professional status. International Journal of Performance Analysis in Sport [Online]. 3 (2), pp. 103-111. [Accessed 03 February 2017].

Eaves, S., Hughes, M. and Lamb, K. (2005) The Consequences of the Introduction of Professional Playing Status on Game Action Variables in International Northern Hemisphere Rugby Union Football. International Journal of Performance Analysis in Sport. [Online]. 5 (2), pp. 58-86. [Accessed 03 February 2017]

Evert, A. (2006) The Significance of Dominant Ball Carrying Collisions as an Indicator of Success in Rugby Union and the Biomechanical Analysis Thereof. Doctorate. University of Pretoria.

Fallowfield, J., Hale, B. and Wilkinson, D. (2005) Using Statistics in Sport and Exercise Science Research. Sussex: Lotus Publishing.

Foster, L., Diamond, I. and Jefferies, J. (2015) Beginning Statistics: An Introduction for Social Scientists. 2nd Edition. London: SAGE.

Gianotti, S., Hume, P., Hopkins, W., Harawira, J. and Truman R. (2008) Interim evaluation of the effect of a new scrum law on neck and back injuries in rugby union. British Journal of Sports Medicine [Online]. 42 (6). [Accessed 03 March 2017]

Hughes, M. and Franks, I. (2015) Essentials of Performance Analysis in Sport. Second Edition. Oxon: Routledge.

Hughes, T., Hughes, M., Williams, J., James, N., Vuckovic, G. and Locke, D. (2012) Performance indicators in rugby union. Journal of Human Sport & Exercise. 7 (2), pp. 383-401 [ Accessed 2 March 2017]

James, N., Mellalieu, S. and Jones, N. (2011) The Development of Position Specific Performance Indicators in Professional Rugby Union. Journal of Sports Sciences [Online].  23 (1), pp. 63-72. [Accessed 02 February 2017]

Jones, M., West, D., Crewther, B., Cook, J. and Kilduff, L. (2015) Quantifying Positions and Temporal Movement Patterns in Professional Rugby Union Using Global Positions System. European Journal of Sport Science [Online]. 15 (6), pp. 488-496 [Accessed 03 February 2017].  

Jones, N., Mellalieu, S. and James, N. (2004) Team performance indicators as a function of winning and losing in rugby union. International Journal of Performance Analysis in Sport. 4 (1), pp. 61-71 [Accessed 25 February 2017]

Lancaster, G., Dodd, S. and Williamson, P. (2004) Design and analysis of pilot studies: recommendations for good practice. Journal of Evaluation in Clinical Practise [Online]. 10 (2), pp. 307-312. [Accessed: 11 March 2017]

McKight, P. and Najab, J. (2010a) Kruskal-Wallis Test. In: Weiner, I. and Craighead, E. (2010) The Corsini Encyclopaedia of Psychology: Volume 2. New Jersey: John Wiley and Sons, pp 904.

McKight, P. and Najab, J. (2010b) Mann-Whitney U Test. In: Weiner, I. and Craighead, E. (2010) The Corsini Encyclopaedia of Psychology: Volume 2. New Jersey: John Wiley and Sons, pp 960.

Milburn, D. (1993) Biomechanics of Rugby Union Scrummaging. Sports Medicine [Online]. 16 (3), pp. 168-179. [Accessed 30 January 2017].

Moore, B. (2016) Collapsing Scrums Bore Fans – and Referees Must Now Get Tough on the Culprits. The Telegraph [Online]. 26 September. Available from: http://www.telegraph.co.uk/rugby-union/2016/09/26/collapsing-scrums-bore-fans—and-referees-must-now-get-tough-on/ [Accessed 25 October 2016].

Nevill, N. and Holder, R. (1999) Home Advantage in Sport: An Overview of Studies on the Advantage of Playing at Home. Sports Medicine [Online]. 28 (4), pp. 221-236. [11 March 2017].

O’Donoghue, P. (2012) Statistics for Sport and Exercise Studies: An Introduction. Oxon: Routledge,.

O’Donoghue, P. (2010). Research methods for sports performance analysis. Oxon: Routledge.

Ortega, E., Villarejo, D. and Palao, J.M. (2009) Differences in Game Statistics Between Winning and Losing Rugby Teams in the Six Nations Tournament. Journal of Sports Science and Medicine [online]. 8, pp. 523-527. [Accessed 04 February 2017].

Pavely, S., Adams, R., Di Francesco, T., Larkham, S. and Maher, C. (2010) Bilateral clearance punt kicking in rugby union: effects of hand used for ball delivery. International Journal of Performance Analysis in Sport [Online]. 10 (1) pp. 187-196. [Accessed 03 February 2017].

Quarrie, K., Cantu, R. and Chalmers, J.  (2002) Interim evaluation of the effect of a new scrum law on neck and back injuries in rugby union. Journal of Sports Medicine [Online]. 32 (10), pp. 633–653. [Accessed 04 February 2017]

Quarrie, K. and Hopkins, W. (2007) Changes in Player Characteristics and Match Activities in Bledisloe Cup Rugby Union from 1972 to 2004. Journal of Sports Sciences [Online]. 25 (8), pp. 895-903 [Accessed 02 February 2017].   

Reed, D. and Hughes, M. (2006). An Exploration of Team Sport as a Dynamical System. International Journal of Performance Analysis in Sport [online]. 6 (2), pp. 114-125. [Accessed 07 February 2017].

Reilly, M., Zbrozek, A. and Dukes, E. (1993) The Validity and Reproducibility of a Work Productivity and Activity Impairment Instrument. PharmacoEconomics [Online]. 4 (5), pp. 353–365 [Accessed 20 January 2017].

Salkind, N. (2017) Statistics for People Who (Think They) Hate Statistics: Using Microsoft Excel 2016. 4th Edition. Los Angeles: SAGE.

Sedeaud, A., Marc, A., Schipman, J., Tafflet, M., Hager, J. and Toussaint, J. (2012) How they won Rugby World Cup through height, mass and collective experience. British Journal of Sports Medicine [Online]. 46 (8), pp. 580-584. [Accessed 04 February 2017].

Treadwell, P.J., (1988) Computer aided match analysis of selected ball games (soccer and rugby union). In: Reilly, T., Lees, A., Davids, K. and Murphy, W.J. (1987) Science and Football: Proceedings of the First World Congress of Science and Football. Oxon: Routledge

Van Rooyen, M., Lambert, M. and Noakes, T. (2006) A Retrospective analysis of the IRB statistics and video analysis of match play to explain the performance of four teams in the 2003 Rugby World Cup.  International Journal of Performance Analysis in Sport [Online].  6 (1), pp. 57-72. [Accessed 03 February 2017].

Van Teijlingen, E. and Hundley, V. (2002) The Importance of Pilot Studies. Nursing Standard [Online]. 16 (40), pp. 33-36. [Accessed 22 February 2017].

Vaz, L., Van Rooyen, M. and Sampaio, J. (2010) Rugby Game-Related Statistics that Discriminate Between Winning and Losing Teams in Irb and Super Twelve Close Games. Journal of Sports Science and Medicine. 1 (09), pp. 51-55. [Accessed 07 February 2017].

Westgate, P. (2007). The Principles and Techniques of Defence in Rugby Union. Applied Sports Science [Online]. Available from: http://www.agard.rugby.hu/letolt/EDZOI/071030AVedekezesAlapelveiesTechnikai.pdf [Accessed 9th March 2017].

Wheeler, K. and Sayers, M. (2009) Contact Skills Predicting Tackle-Breaks in Rugby Union. International Journal of Sports Science & Coaching [Online]. 4 (4). [Accessed 3rd January 2017]

Williams, J., Hughes, M. and O’Donoghue, P. (2005) The Effect of Rule Changes on Match and Ball in Play Time in Rugby Union. International Journal of Performance Analysis in Sport [Online]. 5 (3), pp. 1-11. [Accessed 03 February 2017].

Williams, J., Thomas, C., Brown, R. and Jones, N. The Effect of the Wheeled Scrum in Rugby Union. In: Reilly, T., Cabri, J. and Araújo, D. (2005) Science and Football V: The Proceedings of the Fifth World Congress on science and football. Oxon: Routledge, pp. 274-279.