Observational research
The main type of research used in performance analysis of sport is observa- tional research. In particular, notational analysis studies involve empirical observation with events being counted and timed. Observational research is distinguished from experimental research in terms of the trade-off between control and ecological validity. Experimental research typically involves participants engaging in activity under controlled conditions, sometimes in a laboratory setting, so that the effect of manipulated independent variables on hypothesised dependent variables can be investigated. Observational research does not offer this level of experimental control but has a higher level of ecological validity. Ecological validity represents how well the con- ditions of a study refl ect the real world. In observational analysis techniques used in performance analysis of sport, it is common for sports performance in real competitions to be studied. Other quantitative data that can be recorded during observational research include distances covered, speeds of movement and locations of events.
This book considers types of research differently to other textbooks such as that of Thomas et al. (2005). One difference is that Thomas et al. (2005:
19) considered job analysis and observational research to be different types of descriptive research. This book views job analysis as an application of other types of research including observational research. Another difference is that in Thomas et al.’s (2005) textbook, qualitative research is considered to be a separate type of research to observational research. In the current book, qualitative methods are seen as methods of gathering and analysing data that can be used within different types of research including observa- tional research. Participant observation is very valuable as the researcher learns from experiencing the situations of interest. However, in perform- ance analysis research, non-participant observation is more appropriate as the researcher may not qualify as a participant for the investigation and it is important to remove any sources of bias from the data used. The commonly held view that notational analysis is a highly quantitative type of research is not entirely true. When one considers the activity done during notational analysis, human behaviour is observed and subjectively classifi ed according to a set of behaviour categories being used. For example, in time-motion analysis, there is a subjective judgement as to whether a player is jogging or running. Therefore, the counts and timings of movements are simply quan- titative counts and timings of the qualitative judgements that have been made by the observer. Some research in performance analysis has very little quantitative analysis. For instance, research into tennis player gaze when playing shots can be done through expert analysis of photographs without using any numerical measurements (Lafont, 2007, 2008).
Experimental research
As has already been mentioned, experimental research involves participants engaging in activity under controlled conditions for the purpose of investi- gating the effect of the experimental condition on some hypothesised dependent variables of interest. Performance analysis of sport essentially involves the analysis of actual performance in competition or training. Even when laboratory-based studies of key technical skills are undertaken in per- formance analysis, the investigations are highly quantitative descriptive studies based on detailed measurements rather than experiments into the effect of some independent factor. Therefore, performance analysis of sport does not include experimental research. However, there are many examples of research investigations that are comprised of descriptive observational parts and experimental parts. For example, Huey et al. (2001) used time- motion analysis to investigate the demands of fi eld hockey competition before developing a specifi c intermittent high intensity training programme and testing this programme using a quasi-experimental study. When explain- ing performance analysis results, it is often necessary to draw on evidence from experimental studies in other disciplines such as physiology, psychol- ogy and biomechanics. As a consequence, it is important for performance analysts to have a good understanding of experimental research and other types of research in sports science.
There are many different experimental designs that are distinguished by whether there are control as well as experimental treatments, whether there are pre-tests and mid-tests as well as post-tests, placebos, blind testing, random assignment of participants to groups, whether participants are assigned to groups before or after testing and whether participants are exposed to one or more experimental condition. Thomas et al. (2005:
330–43) listed no fewer than 13 different experimental designs within three broad classes of experiment; pre-experimental, true-experimental or quasi- experimental studies. This list is by no means exhaustive and variations on the different designs are possible.
The main true-experimental design described by Thomas et al. (2005:
333–5) is the ‘pre-test – post-test randomised-groups design’, where two inde- pendent groups of participants are tested before and after some experimental period; one group will have been exposed to an experimental treatment condi- tion while the other will be under control conditions. The advantage of using a control group is that if there is a change in the experimental group’s test per- formances over the experimental period, this can be compared to what happens over that period for participants who are not exposed to the experimental treatment. The use of pre-testing as well as post-testing is important because without pre-testing we cannot be sure whether differences between the groups’
tests performances were developed during the experimental period or whether they already existed before the study. Scientifi c evidence of an effect of the treatment on the hypothesised dependent variables being tested would require
control of all aspects of participants’ lives (eating, drinking, sleeping, activity and so on) with the exception of the experimental treatment, which would be manipulated by the researcher during the experimental period. However, it is neither ethical nor feasible to control the lives of participants to this extent.
Therefore, true-experimental designs use random assignment of participants to the control and experimental groups. If there is a large enough number of par- ticipants in the experiment and they are randomly allocated to groups, it is hoped that the variation in sleep patterns, diets, physical activity level and other lifestyle variables that may infl uence test performance will be similar between the two groups of participants.
Pre-experimental designs do not have random assignment of participants to groups and do not involve a control group. One pre-experimental design described by Thomas et al. (2005: 331) is the ‘static group comparison’, where an experimental group is tested after an experimental period. The post-test results are compared with post-tests results from participants who were not exposed to the experimental treatment. However, in this design participants are not grouped for the purpose of the experiment: the groups already existed prior to the experiment. For example, a group of partici- pants who engage in a certain type of training may be compared with a group formed from those who do not do such training.
Ex-post-facto designs are quasi-experimental designs that involve par- ticipants being tested but then grouped after the experimental period has been completed. A good example of this is McLaughlin’s (2003) PhD study, which involved time-motion analysis of 32 primary school children on seven occasions over a two-year period. At the end of the period, the chil- dren’s absenteeism and body mass indexes were used to divide them into two groups: the healthiest 16 children and the least healthy 16 children.
These groups were compared in terms of the activity observed in the play- ground at break times and lunch times to determine if there was an associa- tion between voluntary involvement in moderate to high intensity activity and health. This study did not qualify as a true experiment for two reasons.
First, the groups were not assigned randomly and there was no experimen- tal treatment. Secondly, observations of movement behaviour were used, although standard tests and measurements were also used within other studies that were included in the PhD.
Another type of quasi-experimental design is the ‘pre-test – post-test groups design’, which is the same as the ‘pre-test – post-test randomised groups design’ except that participants are not randomly assigned to the experimental and control groups. It is often not feasible to randomly assign participants to groups as some may be prepared to perform the pre- and post-tests, but are not willing to undergo the experimental treatment. For example, O’Donoghue and Ormsby (2002) devised a test of free taking per- formance in Gaelic football that they used in an experiment into the effect of mental imagery training. Some participants were able to perform the pre- test and post-test but did not wish to do the relaxation training two days for
six weeks. An issue with this study is that there was a possibility that those participating in the experimental group had a different attitude towards psychological skills training in general than those who were members of the control group. Therefore, there may have been psychological effects infl u- encing the results of the post-tests. Pre-experimental designs and quasi- experimental designs can be used but their disadvantages compared to those of true-experimental deigns must be recognised.
A further type of experimental design is the cross-over design where all participants are tested following both experimental and control conditions.
Because participants may perform the test better the more they do it, the participants are divided into two groups that are exposed to both the exper- imental and control conditions. One group is exposed to the experimental condition fi rst and control condition second, while the other group is exposed to these conditions in the reverse order. One of the main advan- tages of the cross-over design is that inter-individual differences are elimi- nated from the study as the effect of both experimental and control conditions are tested using the same sample of participants.
There are two main types of validity of experimental studies: internal and external validity. Internal validity is the extent to which any changes in the test performance can be attributed to the experimental treatment. Observed changes may be due to participants maturing over the experimental treat- ment, familiarisation with the testing process, biased allocation of partici- pants to groups and psychological expectancy factors. External validity is the extent to which the results of the experiment can be generalised. External validity requires participants who are representative of the population of interest to be used as well as tests that have ecological validity.
Historical research
Historical research is important as solutions to contemporary problems can be based on past lessons. The importance of historical research is recognised in sports science with several journals publishing sports history research, including Sport History Review, Journal of Sport History, International Journal of the History of Sport, Canadian Journal of the History of Sport, Western Historical Quarterly and British Journal of Sports History. The sources of data used in historical research are classifi ed as primary and sec- ondary sources. Primary sources include biographies, newspapers, minutes of meetings, memoirs, other documents and video recordings that are directly related to the events of interest. It is sometimes possible to use par- ticipants who were involved in the historical events of interest; these indi- viduals are primary sources of data. Secondary sources of data are indirectly related to the events of interest and include quoted material, reproduced material and published research that has investigated the events of interest.
People who lived during the time of the events of interest but who were not
directly involved in the events can also participate in historical research as secondary sources of data.
An issue in historical research is the need to establish the authenticity of the data used. Documents and artefacts may be genuine or they may be hoaxes or other misinformation. Therefore, in published papers in sports history, the accompanying reference list and notes state whether documents such as minutes of meetings are signed or not. Controversial social and political issues have an increased likelihood of being portrayed by biased viewers. Therefore, the context of any data used needs to be understood when establishing the reliability of the data. Performance analysis research techniques can be applied within historical research. For example, changes in Grand Slam tennis since the 1960s can be studied using video recordings of sets of matches from different decades. Such a study may address a need by the media. For example, one might wish to produce a television docu- mentary programme about changes in the game over a period of 50 years to be broadcast in the build up to a Grand Slam tournament.
Developmental research
Developmental research is concerned with changes in variables over a long period of time such as the human lifespan or a career in sport. There are two types of developmental research: longitudinal studies and cross-sectional studies.
Longitudinal studies investigate changes in the same participants over a period of time. For example, motor abilities could be tested over various points of human development from childhood to adulthood. Performance variables within fi tness tests or during competition could also be moni- tored over a long period of time. Talent identifi cation is a broad multi- disciplinary area of sports science research where performance analysis has a role to play within longitudinal studies to try to identify areas of performance that are associated with successful adult performance and the times within athletes’ careers where these can be identifi ed. Prospective longitudinal studies monitor participants who can be grouped according to some hypothesised talent factors at the beginning of the research. An example of this was a survey of young Belgian soccer players’ practice over a period of 18 years (Helsen et al., 2000). Retrospective longitudinal studies monitor participants, but do not group them for the purpose of analysis until later years of the study where they can be distinguished according to some status or criteria of interest. Cross-sectional studies, on the other hand, use different participants to monitor changing trends in variables between cohorts. For example, ProZone® (2007) has reported trends in movement during FA Premier League soccer over successive seasons where there will have been some changes in players participating in the competition.
Survey research
Survey research describes the reportable characteristics, beliefs, attitudes, opinions and intentions of populations of interest using samples. Occasionally censuses are done to survey the whole population, but usually a sample is used in research. Surveys can either be done in an exploratory way, with few theoretical assumptions, or can be used to test some pre-existing theory.
Survey research often involves self-reports such as interviews and question- naires, but occasionally surveys can be conducted by researchers gathering information without asking participants to be interviewed or complete questionnaires. Such surveys include comparing prices between retail outlets or surveying road traffi c. What distinguishes these types of surveys from observational research is that observational research involves detailed anal- ysis of behaviour while price or traffi c surveys would record data other than behaviour. Fitness surveys do not involve observation and analysis of behav- iour in detail, but instead record key results of tests and measurements made on participants. An example of a fi tness survey is the 1989 Northern Ireland Fitness Survey, which produced norms for Northern Ireland children for anthropometric measurements, fi tness test results and reported lifestyle var- iables (NIFS, 1989).
Undertaking a survey requires a lot of planning and pilot work to ensure that the surveying technique being used is clear, complete and understand- able. If interviewers are being used, they must be trained to interview par- ticipants for the purpose of the study. If self-completion questionnaires are being used, the questionnaires and accompanying instructions must be thor- oughly checked for clarity and completeness. If a large batch of question- naires containing mistakes is sent out to participants, the exercise will have to be repeated at cost, and some participants will not respond if they see the research as being done incompetently.
Surveys are also being done on the internet with specifi cally designed data capture forms and individual responses being stored in databases for effi cient processing and analysis. An important issue with internet surveys is that there is some means of ensuring the data are authentic. This could be done using e-mail addresses of potential participants that may be publicly shown on organisation internet sites. Telephone surveys are also possible, and have the additional advantage over questionnaires that responses can be clarifi ed and some probing into answers can be done.
Performance analysis can sometimes be used in conjunction with survey research, where the investigation requires both observational and self-report data. A good example of this was a study of scoreline effects on mood and work-rate of semi-professional soccer players (McStravick and O’Donoghue, 2001). This is a particularly good example as the data were collected and analysed within a level 3 research project. The student had been critical of previous research into the effect of scoreline on work-rate because the performance analysis methods used left a large
pathway of mechanisms that may be involved as a ‘black box’. The pre- vious study by O’Donoghue and Tenga (2001) showed a signifi cant scoreline effect on work-rate, but had to speculate on the processes that may have been responsible. Therefore, McStravick combined the use of performance analysis with survey techniques in order to gather work- rate data and mood data respectively. The work-rate data were gathered during observation of player performances during Northern Ireland Premier League soccer matches. After the match was completed, McStravick met with the player and recorded mood ratings reported by the player during parts of the match where the player’s team was in dif- ferent scoreline states.
Action research
Research improves our knowledge of phenomena of interest and provides evidence that can be used by policy makers, coaches or other people. The decisions that are taken often relate to practice and the benefi ts of changing practice should be supported by evidence. Therefore, research and action are often done separately by different groups of individuals: researchers and policy makers. Some research fi ndings may never be used to inform deci- sions, while other research studies may be commissioned by policy makers to obtain independent evidence to inform decisions. The term ‘action research’ represents a cycle of action and research that links these two activ- ities more closely. In action research, an act of intervention is taken within a real-world situation and the effect is closely monitored. Action research can be done on a range of scales from the refl ective practice of individual professionals (Schön, 1983) to a very large scale for entire organisation change (Zuber-Skerritt, 1996).
Action research is appropriate whenever specifi c knowledge is required about a practical problem. The role of performance analysis within a coach- ing context is essentially an action research role that follows four key stages of action research: observing, refl ecting, planning and action. Figure 2.1 shows a modifi ed version of a model of performance analysis activity within the coaching process (Mayes et al., 2009). When a match is played, it is observed with match events being entered into a match analysis system, live.
Match statistics are produced once the match has completed, although some live analysis systems also have the capability of producing match statistics during the match. Performance indicators are compared to norms for the level of opposition faced in the match (O’Donoghue et al., 2008) and areas of play requiring attention are identifi ed. Commercial match analysis pack- ages such as SportsCode (SportsCode, Warriewood, NSW, Australia), Focus X2 (Elite sports Analysis, Delgaty Bay, Scotland) and Dartfi sh (Fribourg, Switzerland) have password-protected internet sites for relevant video sequences to be uploaded and viewed by the squad. This helps the squad refl ect on the performance prior to the next training session. Coaches can