Tactical Performance Profiling in Elite Level Senior Squash

Stafford Murray and Mike Hughes:
Centre for Performance Analysis, University Wales Institute Wales Cardiff

Presented at World Conferecnce of Performance Analysis and Computers in Sport. Cardiff 2001.

1 Introduction:
In order to compete at the world's highest level, athletes need to base their tactical plans on some form of objective data. If game plans are based on subjective feelings, which can sometimes be misleading for numerous reasons (Hughes and Franks, 1997), then the likelihood of success is minimised. Notational Analysis of sport provides us with the ideal tool to collect this objective performance data. When this data is supported by the corresponding video images it provides a very informative and powerful tool. Coaching is a deliberate act of intervention with the intention of improving sports performance. The feasibility of using data from computerised notation systems as feedback to performers has been well documented (Brown & Hughes, 1995; Murray, Maylor, & Hughes, 1998). In general the findings of this research have indicated that if the data is given to the performers correctly, i.e. correct amount, time and terminology, then the effect on performance is positive.

In conjunction with analysis of your own performance it is also essential to have an understanding of your opposition's tactical strengths and weaknesses. By modelling opposition’s performance it is possible to predict certain outcomes and patterns, and therefore intervene or change tactics before the critical incident has occurred. The modelling of competitive sport is an informative analytic technique because it directs the attention of the modeller to the critical aspects of data that delineate successful performance (McGarry and Franks, 1996). By using tactical performance profiles to pull out and visualise these critical aspects of performance, players can build justified and sophisticated tactical plans.

Recent research in squash (Hughes, Wells and Matthews, 2000) outlined the need to calculate the number of matches required to reach a normative profile of data. This research suggested that the optimum number of matches or performances required relied heavily upon the nature of the data and, in particular, the nature of the performers. By using statistical methods and comparing sets of data from differing amounts of matches it was shown that unless a normative profile had been reached (with supporting statistical data) then the subsequent analyses could be subject to significant flaws. However, the research did suggest that 5-6 elite level squash matches provide sufficient data for the normative profile to be reached.

The aim of this paper is to outline and review the development, methodology and application of tactical performance profiles used with elite level male and female English squash players. The aim of the methodology (tactical performance profiles) outlined in this paper is to provide the performers with quantitative analyses, highlighting their own, or an opponent’s, comparative strengths and weaknesses. By modelling performance in this way, tactical plans can be based upon empirical evidence as well as the usual subjective observations of the coaches. These profiles are used in conjunction with high-quality edited video tapes, providing the elite performers with both statistical and visual feed forward, the tapes lend visual evidence of these sets of data and heightens the user friendliness of the analyses. By describing the development of these processes it is hoped that there may be practical ideas that can be used in other sports by analysts and other sports science support operators. Defining some of the experiences and processes could start the process of delineating the generic functions of a performance analyst.

2 Method:

2.1 Subjects:
Five matches, for each profile, were selected from the top opposition (non English) male players (N=3[JP, PN, & AB]) and the top opposition female players in the world (N=3[SF, LJ, & CO). Only matches of a competitive nature (i.e. international governing body ranking events) were selected for the purpose of this study. The score outcome of the matches was deemed irrelevant as research has shown that elite squash players retain the same patterns regardless of winning or losing the match (Hughes, 1985, & Murray et al, 1998), although heavily one-sided matches were avoided.

2.2 Data Collection:
Two computerised notational analysis systems (Brown and Hughes, 1995 & Murray et al, 1998), one real time (in event) and the other lapsed time (post event) were used to collect the data from the 5 matches. Due to amount of data that was being collected the data capture was done from digital video recordings. This allowed the analyst to rest when needed during data collection, therefore minimising any user error. The systems were validated (Brown and Hughes, 1995 & Murray et al, 1998) and user (JW) deemed reliable (Murray et al, 1998 & Hughes et al, 2000). The court was divided into 16 cells for the purpose of data collection.

Fig. 1. Example of 16 cell division of squash court

The data position of the shot played was placed in the cell where the racquet face struck the ball. In order to increase the accuracy of this figure, frame-by-frame motion analysis, with freeze frame and jog shuttle was used.

2.3 Real-Time Analysis System:
The real-time system (SWEAT – Simple Winner and Error Analysis Technology) was based on hand notation system designed by Hughes and Robertson (1995) and later computerised to speed up both data collection and processing. This system analyses the distribution, frequency, and type of the rally ending shot. If more than the rally ending shot is analysed then the lapse-time system is needed. The data is entered using a QWERTY keyboard and mouse. The data is gathered into a Microsoft Access Database using a visual basic interface (Fig. 2). The data is processed real time and available for viewing between game sets of data. The data is processed into 3-D graphs, with the ability to filter by game, players, outcome and shot (Fig. 2).

Fig. 2. Squash SWEAT data gathering interface

Fig. 3. Squash Lapsed-Time data gathering interface

2.4 Lapse-Time Analysis System:
The lapse-time analysis system (Brown and Hughes, 1995) collects much more complex performance data than the real time system. The position on the court, the time and type of shot is entered for every shot in the match. The data collection for this system has around a 1:4, real:notation time ratio for a trained user (i.e. 30 minute match = 2 hours data collection).

3 Results and Discussion:

Fig. 4. Example of initial winner/error data produced from the computerised full analysis system

The data from these matches are processed and analysed. Analyses ranging from simple winner and error ratios to complex rally ending patterns were produced from the computerised systems. The full analysis system firstly provides simple winner and error ratios and average rally length data (fig. 4). The next set of data produced (fig.5) analyses the shot options taken (therefore patterns used) by both players. These two sets of data provide the coach and the player with the patterns used in a specific match, a simple thumbprint of the players’ patterns - a combination of five or more matches was used. The final four lines of the above data further processes the shot type into straight, cross court, short and long, this simplifies the data further for easier understanding by the players. Figure 6 shows the distribution graphs produced by the full system. The three examples below demonstrate the possible combinations (around 300) that the system produces.

Fig. 5. Example of shot frequency summary data

Fig. 6. Examples of various screens of data available

As a result of initial feedback from the coaches and players, and years of experience in interpreting these figures at the top level of squash, these sets of data were then further analysed and summarised into a briefer format (about 16 sides of A4) outlining the major strengths and weaknesses of the player. The coaches and players asked us to condense these data into bullet point form in order to simplify the information, and therefore avoid overload and confusion. Further, these bullet points were then used as a storyboard for the accompanying edited videos assembled to support the performance data. Below (Fig. 7) is an example of the bullet point data initially given.

Fig. 7. Summary of data used as feedback and storyboard for edited video

The feedback from the players and coaches was positive about this summary form of data, and coupled with the edited video made a very powerful tool. As a result of some more feedback from the players, it was decided to combine the data from the five matches into one figure, again reducing the complexity level of the data. Also the data was normalised, and put into percentage form (Fig. 8). Areas of the court that had unusual data in the analysis were further examined with respect to the shot types. The court was also split into forehand / backhand, front / back and the four quarters of the court. These more simple sets of data can more easily be put into tactical plans.

Fig. 8. An example of normalised distribution data

Fig. 9. Distribution of shots that were 2 shots before a winner by Player A [N-2)W].

Additional depth was given to the profiles by analysing the distribution across the court of not only the winners and errors, but also the distribution of shots that preceded the end shot – (N-1)W and (N-1)E. The full analysis system (see fig. 6. – right hand screen) also enables the analyses of the shot that preceded these shots – (N-2)W and (N-2)E. Using these we could then present the positive profiles of shot distributions, from winners (W), (N-1)E, (N-2)W (see Fig. 9 as an example); and negative profiles from errors (E), (N-1)W and (N-2)E. These overall distributions were also further analysed to examine which shot types were contributing most to the frequencies in the important areas of the court.
These profiles were then given to the players at a national squad, again the feedback was positive and ideas from the players were often very perceptive and always very practical. The British champion at the time suggested that we go one layer deeper in the analysis and analyse the shot selection of the top players from the four corners of the court (Fig. 10). This form of analysis assists the players in building a constructive rally and anticipating the opposition’s next shot.

Fig. 10. Example of shot option analysis

A further analysis produced by the real time system is the rally length analysis (fig. 11). This provides information on varying winner / error ratios over varying rally lengths. Players were especially receptive to this data as it often provided a focus for the mental approach to the tactics, for example a major very highly ranked opposition player showed a 1/3 W/E ratio in rallies over 15 shots. The next English player to play this opponent counted to 16 shots in his head before playing the ball to the front of the court. He won 3/1 beating this player for the first time in his career.

Fig. 11. W/E frequencies from the SWEAT system with respect to length of rally

Fig. 12. Example of ‘momentum analysis’ graph

A discussion with the SRA psychologist highlighted her interest in extremes of body language and the resultant outcomes of the next 3 or 4 rallies. We realised that we had the outcome data in the computer from the SWEAT analyses. By writing another analysis program we calculated a running score (momentum) for a player during a game. We gave a winning shot by a player a ‘+1’ score, an error a ‘-1’ score, and if the opponent hit the rally end shot, or it was a let, the score stayed the same (Fig. 12). This would also show any swings in momentum during the match, then the video could be used to analyse the body language and try to understand the reason for these swings. When the physiologist saw these graphs he asked if we could incorporate the rally length into the graphs. He was interested to see if these swings may be related to fitness aspects. By analysing the rally in relation to the momentum this could be observed (Fig. 13).

Fig. 13. Example of Momentum Analysis with rally length included

We are still discovering the potential of the ‘momentum graphs’ as they have only recently been developed. But they do seem such a strong indicator of the mental strength of a player during the different stages of a match and it is felt that there could be more that can be explored with these analyses. The exciting part is how the analyses are pulling together all the different parts of the sports science support team.

4 Conclusions:
It is very difficult to quantify the effect that these profiles may have upon the performance of the athletes, and to attribute transition of performance to the implementation of these profiles may be somewhat naive. However, considering that the world of elite sport (especially when on the playing field or court) is such a multivariate situation then any singular attribution would be very difficult to achieve. The verbal feedback from the players and coaches was both constructive and positive and has raised several issues around how we, as sports scientists, give our information to elite performers. Nevertheless, the process alone made the players more analytical and focused in their approach to matches and tournaments, which, arguably, is a singular positive effect in itself.
The process itself is one of analysis and, more importantly, of self-analysis and change. These experiences are presented as an exemplar of performance analysis, not because we think that they should be imitated, but that there are aspects of the processes that can be analysed and improved. Perhaps from these we can then define some generic indicators of process for the performance analyst.

5 References:

• Brown, D. and Hughes, M. (1995) The effectiveness of quantitative and qualitative feedback in improving performance in squash. In T. Reilly, M.D. Hughes and A. Lees (Eds), Science and racquet Sports. E & FN Spon: London, pp. 232-237.
• McGarry, T. and Franks, I.M. (1996) Analysing championship squash match play: In search of a system description. In S. Haake (ed.) The Engineering of Sport. Rotterdam: Balkema, pp. 263-269.
• Franks, I.M. (1996) The science of match analysis. In T. Reilly (ed) Science and Soccer, London: E. and F.N. Spon.
• Franks, I.M., Goodman, D., & Miller, G. (1983). Analysis of performance: Qualitative or Quantitative. SPORTS, March.
• Hughes, M.D. (1985) A comparison of the patterns of play of squash. In I.D. Brown, R. Goldsmith, K. Coombes & M.A. Sinclair (Eds), International Ergonomics ‘85, Taylor & Francis: London, pp. 139-141.
• Hughes, M.D. (1986) A review of patterns of play in squash. In J. Watkins, T. Reilly and L. Burwitz (Eds), Sports Science. E & FN Spon: London, pp. 363-368.
• Hughes, M.D. and Franks, I.M. (1994) Dynamic patterns of movement of squash players of different standards in winning and losing rallies. Ergonomics, 37 (1), 23-29.
• Hughes, M. and Knight, P. (1995) Playing patterns of elite squash players, using English and point-per-rally scoring. In T. Reilly, M. Hughes & A. Lees (Eds), Science and racquet Sports. E & FN Spon: London, pp. 257-259.
• Hughes, M. and Robertson, C. (1998) Using computerised notational analyisis to create a template for elite squash and its subsequent use in designing hand notation systems for player development. In A. Lees, I. Maynard, M. Hughes and T. Reilly (Eds), Science and racquet Sports II. E & FN Spon: London, pp. 227-234.
• Hughes, M. D., Wells, J. and Matthews, C. (2000) Performance profiles at recreational, county and elite levels of women’s squash. Journal of Human Movement Studies, 39, 85-104.
• Murray, S., Maylor, D. & Hughes, M. (1998) The effect of computerised analysis as feedback on the performance of elite squash players. In A. Lees, I. Maynard, M. Hughes and T. Reilly (Eds), Science and racquet Sports II. E & FN Spon: London, pp. 235 – 246.