PLEASE SCROLL DOWN FOR ARTICLE

Transcription

1 This article was downloaded by: [Swansea University] On: 12 August 2009 Access details: Access Details: [subscription number ] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Journal of Sports Sciences Publication details, including instructions for authors and subscription information: An objective method for depicting team performance in elite professional rugby union Nicholas M. P. Jones a ; Nic James a ; Stephen D. Mellalieu a a Department of Sports Science, University of Wales Swansea, Swansea, UK First Published on: 15 March 2008 To cite this Article Jones, Nicholas M. P., James, Nic and Mellalieu, Stephen D.(2008)'An objective method for depicting team performance in elite professional rugby union',journal of Sports Sciences,26:7, To link to this Article: DOI: / URL: PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

2 Journal of Sports Sciences, May 2008; 26(7): An objective method for depicting team performance in elite professional rugby union NICHOLAS M. P. JONES, NIC JAMES, & STEPHEN D. MELLALIEU Department of Sports Science, University of Wales Swansea, Swansea, UK (Accepted 19 November 2007) Downloaded By: [Swansea University] At: 13:58 12 August 2009 Abstract Using a two-study approach, we examined a methodology for objectively depicting team performance and form in a professional rugby union side. Study 1 developed standardized indicators to examine a team s performance in a single match relative to their previous matches over a domestic season via a performance report and form chart. This resulted in standardized performance indicators that provided instant and coherent feedback on the team s performance relative to previous standards. Study 2 then utilized this methodology to compare a match between two professional sides, played the following season, to assess the extent to which the performance by one team affected that of the other. Comparison of the two teams performances, for the match they played against each other, identified a drop in relative performance (against previous standards) for both teams on the same performance indicators. This appeared to be due to the match being a close, hardfought contest particularly in the tackle and around the gain-line. The findings of the two studies suggest an accurate and viable methodology for depicting team performance that is superior to the global measures previously adopted. Presenting teams current standardized performances (i.e. form) on a single visual scale also has utility for coaches within an applied setting. Future research should further investigate the effect of one team s performance on another through both individual and team outcomes. Keywords: Performance indicators, form charts, rugby union, standardization Introduction Empirical research investigating performance analysis within rugby union has generally been limited to studies exploring specific aspects of the game, such as the patterns of play of teams or physiological estimates of positional work rates of individual players (e.g. Deutsch, Kearney, & Rehrer, 2002; Hughes & White, 1997; Hughes & Williams, 1988). More recently, several studies (e.g. Hunter & O Donoghue, 2001; James, Mellalieu, & Jones, 2005; Jones, Mellalieu, & James, 2004; Vivian, Mullen, & Hughes, 2001) have followed the recommendations of Hughes and Bartlett (2002) and focused upon the development and utilization of performance indicators. Here, performance indicators are referred to as a selection, or combination, of action variables that aims to define some or all aspects of a performance (Hughes & Bartlett, 2002, p. 739). The development of performance indicators subsequently leads to the creation of performance profiles, which are suggested to be a description of a pattern of performance from an analysed team or individual, typically created from collected frequencies of a combination of performance indicators that offer some prediction of future performance (Hughes, Evans, & Wells, 2001). To date, few studies have attempted to provide indicators of team performance through the comparison of winning and losing teams (Hughes & White, 1997; Hunter & O Donoghue, 2001; Jones et al., 2004; McCorry, Saunders, O Donoghue, & Murphy, 2001; Stanhope & Hughes, 1997). Hunter and O Donoghue (2001), for example, assessed positive and negative aspects of attacking and defensive play, changes in possession, and methods used to gain territory during the 1999 rugby union World Cup. Winning and losing sides were found to differ on the number of occasions that a team entered into the opposition s final third of the playing field and the frequency of attacks by which the team went around the opposition. One of the limitations of Hunter and O Donoghue s (2001) study was that they compared the aggregate data of two or more different teams Correspondence: N. M. P. Jones, Department of Sports Science, University of Wales Swansea, Vivian Tower, Singleton, Swansea SA2 8PP, UK. nicholas.jones@hotmail.co.uk ISSN print/issn X online Ó 2008 Taylor & Francis DOI: /

3 Downloaded By: [Swansea University] At: 13:58 12 August N. M. P. Jones et al. (the winning and losing sides) rather than analysing one team s success and failure outright. Comparing winning and losing sides may therefore have resulted in a potential loss of any meaningful information due to each team possessing different styles of play and, consequently, diverse performance profiles (see Taylor, Mellalieu, & James, 2005). To overcome this limitation, Jones et al. (2004) considered the winning and losing performances of a single team and found a number of significant statistical and practical differences (cf. Atkinson, 2003). For example, while lineout success on the opposition throw differed significantly between winning and losing performances, large observable (but non-significant) differences were apparent for a number of performance indicators such as turnovers won. Although providing a more fine-grained analysis of team performance in rugby union, Jones et al. (2004) used the distinction between winning and losing as the sole measure of performance, in line with previous research (e.g. Hughes & White, 1997; Hunter & O Donoghue, 2001; McCorry et al., 2001; Stanhope & Hughes, 1997). Although this can often indicate the level of team performance, given the multifaceted nature of the sport, which contains many different skills, it may be of more practical utility to coaches to adopt a performance measure that is independent of match outcome and accounts for potential confounding variables that may affect performance. These include factors such as the match venue, the weather conditions, and the quality of the opposition (James, Mellalieu, & Hollely, 2002). Although previous studies of rugby union profiling have successfully constructed methodologies to assess performance scores and relative form or consistency for individual players (e.g. Bracewell, 2003), no study has assessed team performance via the evaluation of team performance indicators. Indeed, to date, performance analysis in rugby union has not provided comprehensible, objective records of team performance and subsequent form that can be used by coaches in a practical setting. This is unsurprising given that rugby union is a complex sport whereby performance can be analysed and presented in a variety of ways (e.g. timed, frequency, and percentage data). Consequently, when the skills or areas of the game are combined to provide an analysis of team performance, different scales and frequencies prevent simple comparisons from being made. The aim of this study is to present a methodology for objectively depicting team performance and form in rugby union. Specifically, it is our intention to consider the validity of standardizing performance indicators and presenting them on a single visual scale to provide an accurate and viable technique for depicting team performance that is better than the global measures previously adopted and can be used by coaches within a practical setting. This is done in two studies. Study 1 develops a range of rigorous standardized team performance indicators and examines a professional rugby union side s performance in a single match relative to their current and previous form. Study 2 utilizes the methodology developed in Study 1 to compare the team performances of two professional sides playing against each other with reference to their respective current form. STUDY 1 Methods Study design and participants Comparative standardized scores for performance indicators were developed and used to assess the team performance of a rugby union team in a single match compared with their current and previous form over a domestic season. This was achieved in three stages. First, a list of team performance indicators was identified and standardized. Next, a standardized performance report was constructed to present the team s match data in relation to their existing form. The performance report was then subjected to validation by a panel of expert coaches. A 20-match data sample from the season of a European professional team was used during the study based upon the availability of match recordings. Team members ranged in age from 20 to 33 years (mean 25.7 years, s ¼ 3.3). Seventy-five percent of the squad had represented their country in international competition, with 330 appearances overall (mean 25.1, s ¼ 8.3). Before the study commenced, ethical approval was granted by the university s ethics committee and informed consent to use match recordings was obtained from the team s director of rugby. Identification of performance indicators for standardization In this study, we used a reliable and valid computerized data collection system based on the procedures adopted by James et al. (2005). Specifically, team performance indicators were identified by the authors and content-validated by the two teams professional coaches (combined playing and coaching experience ¼ 51 years). This resulted in 18 performance indicators, which were presented as percentages where possible (Table I). The rationale for the use of percentages was that this reduced the number of performance indicators (i.e. percentages combine positive and negative outcomes), thereby presenting a less complex depiction of performance that would still incorporate all aspects of team play on one visual scale (cf. Hughes & Bartlett, 2002).

4 Downloaded By: [Swansea University] At: 13:58 12 August 2009 Table I. Performance indicators used for standardizing a match against previous performances. 1 Analysed team scrum success (%) (þ) 2 Opposition scrum success (%) (7) 3 Analysed team lineout success (%) (þ) 4 Opposition lineout success (%) (7) 5 Analysed team ruck success (%) (þ) 6 Opposition ruck success (%) (7) 7 Analysed team maul success (%) (þ) 8 Opposition maul success (%) (7) 9 Analysed team successful tackles (%) (þ) 10 Offload percentage (%) (þ) 11 Breaks made (%) (þ) 12 Turnover percentage (%) (þ) 13 Analysed team open-play and restart kick success (%) (þ) 14 Analysed team goal kick success (%) (þ) 15 Penalty percentage given away (%) (7) 16 Analysed team total errors (frequency) (7) 17 Analysed team intrusions into attacking 1/3 (frequency) (þ) 18 Analysed team time in possession (seconds) (þ) Note: The þ and 7 signs represent whether an excellent performance for the specific behaviour should result in a high or low score. A þ equates to a high score for an excellent performance. Non-parametric standardization Standardization is a common linear transformation that typically rescales data and centres the mean of a distribution at zero and the standard deviation at one (Howell, 1992). Due to the non-parametric nature of rugby union data (typically skewed due to outliers), medians were used as a more accurate measure of central tendency to means when standardizing performance indicators (James et al., 2005; Jones et al., 2004; Zar, 1999). Unlike James et al. (2005), we did not use confidence limits alongside the median, as their purpose is to show how accurate a sample median is in relation to the true population median (Campbell & Gardner, 2000). Here, we used the inter-quartile range, as confidence limits (although similar in size to interquartile range) should not be used as a measure of the spread of the data and as an alternative to standard deviation (Hopkins, 2000). To create a recognizable performance scale for coaches, standardizations were calculated resulting in distributions with medians of 50, inter-quartile ranges of 15, and a range of scores typically between 0 and 100. Thus, the typical z score transformation was amended with the mean and standard deviation being replaced by the median and inter-quartile range, the subsequent standardized score being multiplied by 15, and 50 added to gain the desired average score and range. The resultant formula for the standardization of the data was: X Mdn transformed score ¼ ð15þ þ 50 IQR Depicting team performance in rugby union 693 where X ¼ the performance indicator value for a specific match, Mdn ¼ the sample median, and IQR ¼ the inter-quartile range of the sample. Initial testing of the data set had also indicated that when standardizing values, medians and interquartile ranges produced more sensible values than means and standard deviations for performance indicators where very high or low percentage scores were regularly achieved (e.g. opposition scrum success). For example, when standardizing the performance of the twentieth match against the previous 19 played by the team, the mean of opposition scrum success was 1.69% compared with the median of 0%. Typical of the skewed data collected, 15 of the 20 performances for this performance indicator were 0%, while a relatively small number of extreme values produced the mean value of 1.69%. Thus, if the mean had been used as the average for the standardization, then all performances of 0% (three-quarters of them) would be considered as lower than average (since they fall below the mean value). Using the median, however, provides a more realistic depiction of this variable whereby the average performance by the team (i.e. 0% of the scrums on the opposition ball being won) would be standardized to a value of 50. It should be noted that adoption of the median approach means that a poor performance for this variable is impossible, as a score of less than 0% cannot occur. However, a median of 0% can be an accurate portrayal of the situation for this variable, as not winning any of the opposition s scrum ball can be the norm within rugby union. Data samples used for standardization Two different sized data samples were used to investigate variations between each team s performances over the previous 19 matches of the season (previous form) and the last 5 matches (current form). First, the 18 performance indicators (Table I) were extracted from the analysis of the twentieth match of the sample. Data were then standardized relative to the previous 5 and 19 matches. In each case, the twentieth match was not included in the sample as this match was to be compared with the previous n matches. It was deemed necessary to use at least a 5-match sample based upon previous research of the size of sample needed to gain an accurate representation of typical performance (cf. Hughes et al., 2001; James et al., 2005; Vivian et al., 2001). Construction of the standardized performance report The standardized data were presented as a novel graphical form chart with tabulated descriptive

5 Downloaded By: [Swansea University] At: 13:58 12 August N. M. P. Jones et al. statistics beneath completing a performance report. The median line (y ¼ 50) was displayed on the form chart to visually illustrate those performance indicators that were above or below previously accomplished standards. The inter-quartile range was also plotted to provide further detail on the level of performance for each performance indicator. However, due to the variation in skewness values for the different performance indicators, it was impossible to determine a common standardized value where performance was above the 75th or below the 25th percentile. The absolute limits for these percentiles were therefore plotted (y ¼ 65 for an extremely positively skewed sample, i.e. performance indicators with a median of 0%, and y ¼ 35 for an extremely negatively skewed sample, i.e. performance indicators with a median of 100%). Thus, a standardized performance indicator value greater than 65 would always be above the 75th percentile, while a value less than 35 would always be below the 25th percentile. Additionally, for certain performance indicators a lower frequency or percentage would equate to a better performance (e.g. frequency of breaks made). In these instances, the data were inverted so that they could be compared and presented on the same chart with positive performances for performance indicators always appearing above the median line (450). Validation of the performance report To obtain a suitable content validity, a protocol was developed whereby the standardized performance report was examined by the head coach (elite coaching experience ¼ 12 years) of a European regional rugby union team. The report was compiled using the data collected and the performance indicators identified (n ¼ 13) by the team s performance analyst. Minor differences in the operational definitions of performance indicators were evident, with the majority being similar in nature to the existing indicators outlined (Table I). The coach was asked to complete a questionnaire rating his team s last match on a scale of 1 to 5 for each of the 13 performance indicators in comparison to their current form, where 1 ¼ very poor, 3 ¼ average, and 5 ¼ very good. He was then presented with the performance report of the match in question (using their previous five matches as the data sample) and asked to indicate the perceived utility of the chart. Differences between the subjective view of the coach and the objective view of the performance report were subsequently explored. Specifically, the chart was compared with the coach s scale using the following criteria: 1 ¼ in the lower quartile; 2 ¼ below the median but above the lower quartile; 3 ¼ median or 5% above or below; 4 ¼ above the median but less than the upper quartile; and 5 ¼ above the upper quartile. The coach suggested that the form chart provided a clear visual depiction of team performance compared with previously accomplished standards. Indeed, he described the form chart as a potential analysis tool that provides instant information on one scale that would normally be obtained from several pages of statistics. The importance of providing objective feedback on performance was further illustrated by the coach s subjective answers to the match questionnaire. Total agreement was achieved on only three performance indicators out of 13, while seven differed on the 5-point scale by a single point. However, more substantial variance was present for three of the answers where disagreement between the coach and the objective performance report was greater than 2 points. On seeing this disparity (i.e. while looking at the objective performance report), the coach expressed surprise and suggested inaccuracy in his recognition exclaiming: I didn t think my answers would be so different, but thinking about it, there is a great deal to remember in a match, especially when comparing it with previous matches. Coach error of this type is unsurprising, given that coaches have previously been shown to have difficulty in recalling and comparing past performances in multi-faceted sports (e.g. Franks & Miller, 1993). Results The standardized performance report The standardization of the performance indicators allowed the objective presentation of all the relevant data from the sampled rugby union match on one scale (Figure 1). The non-standardized statistics for each performance indicator provided further clarification and allowed fine-grained analysis by coaches through the use of actual match data. Figure 1 illustrates how a team s relative performance in a match may therefore be interpreted differently depending on the data sample it is compared with. Match 20 is depicted as a better performance when compared with the previous 19 matches (previous form) as opposed to the previous five (current from). This is indicated by the descriptive statistics, as 10 of the 18 performance indicators possessed higher median values in the five-match sample than in the 19-match sample. It is therefore apparent that the form of the analysed team was far better towards the end of the season, thereby reducing the positive effects of match 20 s performance that is, lower standardized scores resulted when using the 5-match (current form) sample.

6 Depicting team performance in rugby union 695 Downloaded By: [Swansea University] At: 13:58 12 August 2009 Performance Indicator Actual Match Value Median 19 Matches Median 5 Matches 1 Scrum Success Analysed Team Ball 100% 72.73% 84.62% 2 Scrum Success Opposition Ball 12.5% 0% 0% 3 Lineout Success Analysed Team Ball 70.59% 68.18% 71.43% 4 Lineout Success Opposition Ball 38.46% 7.14% 9.09% 5 Ruck Success Analysed Team Ball 88.24% 90.91% 89.29% 6 Ruck Success Opposition Ball 5% 3.85% 3.85% 7 Maul Success Analysed Team Ball 80% 100% 100% 8 Maul Success Opposition Ball 0% 0% 11.11% 9 Tackle Success 88.50% 88.89% 92.68% 10 Offload Percentage 43.75% 27.78% 21.43% 11 Freq of Breaks Made Turnover Percentage 33.33% 63.64% 72.22% 13 Kick Success 84.21% 70.59% 73.81% 14 Goal Kick Success 90% 66.67% 57.14% 15 Penalty Percentage (Given Away) 27.27% 48.39% 52% 16 Freq of Errors Made No. of Intrusions into Attacking Third Possession Time 694 secs 509 secs 485 secs Figure 1. Performance report of match 20 displaying standardized scores based on the previous 19 and 5 matches respectively, together with match 20 data and the median values of the previous 19 and 5 matches. STUDY 2 Methods Study design Study 2 used the methodology developed in Study 1 to compare the team performances of two professional sides playing against each other with reference to their respective current form. This was achieved in three stages. First, a list of team performance indicators was confirmed and standardized. Next, a standardized performance report was constructed for each team in relation to their existing form. Finally, the performance reports of the two teams were compared. Participants and data Data samples from the season of two elite European rugby union professional teams (Team 1

7 Downloaded By: [Swansea University] At: 13:58 12 August N. M. P. Jones et al. and Team 2) were utilized. Team members ages ranged from 19 to 35 years (mean 25.9 years, s ¼ 4.3) for Team 1 and 20 to 38 years (mean 26.9 years, s ¼ 5.1) for Team 2. In Team 1, 81% of the squad had represented their country at international standard, with 382 appearances in total (mean 28.2, s ¼ 9.4). For Team 2, 74% of the squad had represented their country in international competition, with 341 appearances in total (mean 29.1, s ¼ 7.3). Before the study commenced, ethical approval was granted by the university s ethics committee and informed consent to use match recordings was obtained from each of the team s director of rugby. Data were selected from the teams on the basis that they were closely matched in terms of ability. This was determined by their respective end-of-season league standings in their domestic competition (sixth and seventh). Data were gained from the two elite teams analysis databases for the five consecutive league matches prior to the teams playing each other during the season. Additionally, both teams databases were used to obtain the data for the match between them. This enabled a reliability check to ensure that both teams were recording the same information. Although the two teams did not have identical opponents preceding their match, there were a number of factors that suggested they were of comparable standard. All matches were from the same league, both teams played three away and two home matches, and matches were played on the same dates. Thus some reduction in the number of potential confounding variables was evident (Hughes & Bartlett, 2002; James et al., 2002). Identification of performance indicators To ensure acceptable reliability, performance indicators were carefully chosen through discussions with the elite teams respective performance analysts. Only those performance indicators (n ¼ 12) with the same operational definitions for each team were included in this study (see Table II). Although each of the teams analysis systems had reportedly undergone inter- and intra-observer reliability testing, a further check for their compatibility was made. The values for each performance indicator for the match between the two teams were recorded by each team s analyst and subsequently compared. The resulting error level of 55% was deemed acceptable and suggested that both analysts were using the same operational definitions for the variables used in this study. Results Eight of 12 of Team 1 s performance indicators were within the typical range that is, between their 25th Table II. Performance indicators used for standardizing the performances of Team 1 and Team 2 against their previous five performances. 1 Scrum success analysed team ball (%) (þ) 2 Scrum gain-line made (%) (þ) 3 Lineout success analysed team ball (%) (þ) 4 Lineout gain-line made (%) (þ) 5 Rucks/mauls won (%) (þ) 6 Phases 7 þ (frequency) (þ) 7 Passes made (frequency) (þ) 8 Offloads for (frequency) (þ) 9 Turnovers won (frequency) (þ) 10 Missed tackles (frequency) (7) 11 Penalties conceded (frequency) ( ) 12 Handling errors (frequency) (7) Note: The þ and 7 signs represent whether an excellent performance for the specific behaviour should result in a high or low score. A þ equates to a high score for an excellent performance. and 75th percentile performances for their previous 5 matches (dotted lines in Figure 2). Only the number of handling errors made was better than their previous median performance (5 versus a previous median of 12). Their performance was considerably worse for lineout success on their own throw in (standardized score ¼ 10.9, equating to a 60% success rate), gain-lines made at both scrums (33.65, 0%), and lineouts (33.11, 50%). Performance was relatively poor on scrum success on their own put-in (36.99, 50%), although this was better than their worst 25% of performances. Nine of 12 of Team 2 s performance indicators were between their 25th and 75th percentile performances for their previous 5 matches (dotted lines in Figure 3), although passes made, number of missed tackles, penalties conceded, and number of handling errors made were better than their previous median performance. Their performance was lower than the 25th percentile for gain-lines made at both scrums (14.74, 33%) and lineouts (32.11, 29%) and rucks/ mauls won (25.00, 80%). The form charts showed that both teams performed most performance indicators at similar levels to their previous five matches (i.e. between the 25th and 75th percentiles), with the only performances falling outside these values being poorer performances, by both teams. Both Teams 1 and 2 made less gain-lines than their previous median values at both scrums (0% and 33%) and lineouts (50% and 29% respectively), suggesting intense competition between the teams around the gain-line. Team 2 s ability to make the gain-line was poorer than that of Team 1 with respect to previous performances and thus may be partly indicative of the match result in favour of Team 1. One performance indicator related to gain-lines made is the number of tackles missed, as if one team did not miss many tackles the expectancy would be for the other team to have a

8 Depicting team performance in rugby union 697 Downloaded By: [Swansea University] At: 13:58 12 August 2009 Performance Indicator Actual Match Value Median 5 Matches 1 Scrum Success Analysed Team Ball 50% 88.2% 2 Scrum Gain-line Made 0% 54.5% 3 Lineout Success Analysed Team Ball 60% 87.5% 4 Lineout Gain-line Made 50% 80% 5 Rucks/Mauls Won 89.8% 90.3% 6 Phases 7 þ Passes Made Offloads For Turnovers Won Number of Missed Tackles Penalties Conceded Handling Errors 5 12 Figure 2. Performance report for Team 1 displaying standardized scores based on their previous 5 matches, together with match data against Team 2 and the median values of their previous 5 matches. low gain-line percentage. In this respect, Team 1 performed at their previous median performance level for number of tackles missed (n ¼ 8), while this was the highest rated performance indicator for team 2 (63.85, n ¼ 8). Thus, Team 2 seemingly raised their performance on this performance indicator during this match, accounting for the low gain-line percentage for Team 1, but this only equated to a similar level of tackling performance to that of Team 1, who consequently also restricted gain-line success. Team 2 were less successful than their previous performances at the rucks and mauls, suggesting that Team 1 were better able to turn the ball over in these contact situations than the previous five opponents of Team 2 had been able to achieve. Although Team 1 only performed at their median level for turnovers won, this would appear to be a strength of their game, at least relative to the previous opponents of Team 2. In summary, while it appears therefore that Team 1 did not perform at a higher level than their previous performances, the overall match result was in Team s 1 favour, which suggests that their performance indicators were sufficient to ensure a positive outcome to the match. Discussion The aim of this study was to examine a methodology for objectively depicting team performance and form in a professional rugby union side. Study 1 developed standardized indicators to examine a team s performance in a single match relative to their previous (19-match sample) and current (5-match sample) form over a domestic season via a performance report and form chart. The standardization of performance indicators presented via a form chart through the use of a combination of comparative scores satisfied the aim of creating an objective view

9 698 N. M. P. Jones et al. Downloaded By: [Swansea University] At: 13:58 12 August 2009 Performance Indicator Actual Match Value Median 5 Matches 1 Scrum Success Analysed Team Ball 100% 100% 2 Scrum Gain-line Made 33% 100% 3 Lineout Success Analysed Team Ball 82% 88% 4 Lineout Gain-line Made 29% 63% 5 Rucks/Mauls Won 80% 85% 6 Phases 7 þ Passes Made Offloads For Turnovers Won Number of Missed Tackles Penalties Conceded Handling Errors 6 8 Figure 3. Performance report for Team 2 displaying standardized scores based on their previous 5 matches, together with match data against Team 1 and the median values of their previous 5 matches. of rugby union performance that could be used in a practical setting. Indeed, the content validation of the system by the elite coach provided further clarification that the assessment of rugby union performance through standardization provides a versatile analysis tool with great practical potential. The form chart provides instant information concerning the performance of each indicator on a single scale. It also allows coaches to isolate areas where performances are poorer or better than previously and can be modified if so desired, to include different combinations of indicators to provide both team and individual feedback. This method of presenting and disseminating performance data has the potential to evolve. For example, the data sample can be modified to incorporate only those matches applicable to current form (most recent matches) or specific competitions (e.g. European matches as opposed to domestic matches). It would also appear to be easily transferable for use in other multi-faceted interactive sports such as soccer and hockey. In the second study, two teams playing against each other in the same match were subjected to the same form chart analysis. This also allowed a comparison of the two form charts to determine if each team s relative performances (to their respective previous 5 matches) were also indicative of the natural interaction that exists between teams in any match. By using the same methodology for both teams in a match, the form chart was therefore used to assess the extent to which a good or poor performance by one team affected the performance of the other team. The only evidence from this study was a drop in relative performance (against previous standards) for both teams on some performance indicators. This relationship is suggested to be a

10 Downloaded By: [Swansea University] At: 13:58 12 August 2009 consequence of a close, hard-fought contest, in this case in the tackle around the gain-line. Consequently, while it would appear that Team 1 did not perform at a higher standard than their previous performances, the match result in their favour shows this performance was good enough on the day. It appears therefore that Team 1 was marginally better than Team 2 and while Team 2 raised their game in some areas, this was not enough to overcome Team 1 who played closer to their normal standard. The expectation that a simple inverse relationship would exist between the form charts of the two teams such that a high-level performance by one team resulted in a consequent low-level performance by the other team was not supported. Indeed, the only relationship evident was where both teams underperformed on the same performance indicators relative to their current form. This was suggested to be a consequence of a hard-fought contest resulting in lower success for both teams compared with previously. Further studies are thus required to determine whether this is an unusual incident or is a typical effect evident in closely contested matches, as there is little evidence of this in the literature of the effect of one team s performance on another. Although the performance of any team in an interactive sport will be affected by the strength of the opposition, it is very difficult to ascertain whether a performance is a result of good or bad execution by a team (self-imposed) or is directly attributable to the performance of the other team. The form chart methodology thus has the potential to assist in making this relationship clearer. There could also be a case for using the form chart to assess the relative contribution of any performance indicator to overall team performance. A profile may be developed over time such that good performances on some performance indicators may be indicative of successful performance. This is likely to be specific to the particular strengths or weaknesses of a team but may also relate to the opposition s strengths and characteristics. For example, in rugby union, when playing against a team possessing a strong lineout, the performance within the lineout by the analysed team may have a greater contributory weighting to overall team performance. At present, this is pure speculation, but when larger data sets are available an answer to this question may be possible. In addition, while the current standardization methodology has only touched upon the comparison of two different teams, a further practical implication is its use in the analysis of the technical weaknesses or strengths of future opponents before a match or tournament. Based on these identified objective strengths or weaknesses within a future opponent s performance indicators, based on their previous or current form, coaches Depicting team performance in rugby union 699 can tailor training sessions to address specific areas and thus maximize a teams preparation for future competition. Finally, several methodological issues have been identified that need to be considered by researchers in performance analysis. First, the effect of the sample size used when standardizing data or indeed preparing any form of profile requires further investigation. While the use of an entire season s matches provides a relatively stable data sample, it also lessens the effect and potentially conceals the current form of the team or individual, thereby illustrating how the issue of sample size can markedly alter the way in which a match performance is perceived (cf. Hughes et al., 2001; James et al., 2005; Vivian et al., 2001). For example, although a coach may wish to compare his team s twentieth match of the season with the preceding 19 (i.e. previous form), it may be better to utilize only the most recent performances (i.e. current form). In the field of economics and business, using the most recent data and disregarding statistics from the start of the sample as more data are generated is known as a moving average analysis (Roberts, 2003). This means of giving more weight to the most recent data could be adapted, in conjunction with the form chart, to provide coaches with a more objective, consistent, and reliable means of defining their performance trend. It should also be noted that randomly occurring confounding variables may contribute to performance, such as an unusual bounce of the ball or an influential refereeing decision (James et al., 2002). In large data sets, these will tend to be balanced across variables but they may have large effects on smaller samples. The issue of current form and the use of moving averages (Bracewell, 2001, 2003; Roberts, 2003) therefore represent an interesting and important area for further exploration. In this study, the use of non-parametric statistical techniques was appropriate to the nature of the data and thereby reduced the effect of outliers that are a natural consequence of the sport of rugby union (cf. James et al., 2005; Jones et al., 2004). Further research is therefore also required to explore other potential analysis mechanisms to cater for these and other phenomena that are idiosyncratic to the sports themselves. References Atkinson, G. (2003). Does size matter for sports performance researchers? Journal of Sports Sciences, 21, Bracewell, P. J. (2001).Perception of individual rugby player performance and the impact of non-performance on statistical analyses. Research Letters in the Information and Mathematical Sciences, 2,

11 700 N. M. P. Jones et al. Downloaded By: [Swansea University] At: 13:58 12 August 2009 Bracewell, P. J. (2003). Monitoring meaningful rugby ratings. Journal of Sports Sciences, 21, Campbell, M. J., & Gardner, M. J. (2000). Medians and their differences. In D. G. Altman, D. Machin, T. N. Bryant, & M. J. Gardner (Eds.), Statistics with confidence (2nd edn., pp ). London: BMJ Books. Deutsch, M. U., Kearney, G. A., & Rehrer, N. J. (2002). A comparison of competition work rates in elite club and super 12 rugby. In W. Spinks, T. Reilly, & A. Murphy (Eds.), Science and football IV (pp ). London: Routledge. Franks, I. M., & Miller, G. (1993). Training coaches to observe and remember. Journal of Sports Sciences, 9, Hopkins, W. G. (2000). New view of statistics (available at: accessed 26 January 2005). Howell, D. C. (1992). Statistical methods for psychology (3rd edn.). Pacific Grove, CA: Duxbury Press. Hughes, M. D., & Bartlett, R. M. (2002). The use of performance indicators in performance analysis. Journal of Sports Sciences, 20, Hughes, M. D., Evans, S., & Wells, J. (2001). Establishing normative profiles in performance analysis. International Journal of Performance Analysis in Sport, 1, Hughes, M. D., & White, P. (1997). An analysis of forward play in the 1991 rugby union world cup for men. In M. D. Hughes (Ed.), Notational analysis of sport I & II (pp ). Cardiff: UWIC. Hughes, M. D., & Williams, D. (1988). The development and application of a computerized rugby union notation system. Journal of Sports Sciences, 6, Hunter, P., & O Donoghue, P. (2001). A match analysis of the 1999 rugby union world cup. In M. D. Hughes & I. Franks (Eds.), Pass.com: Fifth World Congress of Performance Analysis of Sport (pp ). Cardiff: UWIC. James, N., Mellalieu, S. D., & Hollely, C. (2002). Analysis of strategies in soccer as a function of European and domestic competition. International Journal of Performance Analysis in Sport, 2, James, N., Mellalieu, S. D., & Jones, N. M. P. (2005). The development of position-specific performance indicators in professional rugby union. Journal of Sports Sciences, 23, Jones, N. M. P., Mellalieu, S. D., & James, N. (2004). Team performance indicators in rugby union as a function of winning and losing. International Journal of Performance Analysis in Sport, 4, McCorry, M., Saunders, E. D., O Donoghue, P. G., & Murphy, M. H. (2001). A match analysis of the knockout stages of the 1995 rugby union World Cup. In M. D. Hughes (Ed.), Notational analysis of sport III (pp ). Cardiff: UWIC. Roberts, P. (2003).Moving averages the heart of trend analysis. The Alchemist, 33, Stanhope, J., & Hughes, M. D. (1997). An analysis of scoring in the 1991 rugby union World Cup. In M. D. Hughes (Ed.), Notational analysis of sport III (pp ). Cardiff: UWIC. Taylor, J., Mellalieu, S. D., & James, N. (2005). A comparison of team and individual behaviours in professional soccer. International Journal of Performance Analysis in Sport, 5, Vivian, R., Mullen, R., & Hughes M. D. (2001). Performance profiles at league, European Cup and International levels of male rugby union players, with specific reference to flankers, number 8s and number 9s. In M. D. Hughes & I. Franks (Eds.), Pass.com: Fifth World Congress of Performance Analysis of Sport (pp ). Cardiff: UWIC. Zar, J. H. (1999). Biostatistical analysis (4th edn.). Englewood Cliffs, NJ: Prentice-Hall.