TECHNIQUE TAXONOMY APPLIED TO GOLF PUTTING. Pat McLaughlin and ...
Biomechanical analysis of golf putting has focused almost exclusively on ...
IDENTIFYING DIFFERENT BIOMECHANICAL ‘TECHNIQUES’ TECHNIQUE TAXONOMY APPLIED TO GOLF PUTTING. Pat McLaughlin and Russell Best 1
Institute of Sport, Exercise and Active Living, Victoria University, Melbourne, Australia email:
[email protected]
INTRODUCTION Biomechanical analysis of golf putting has focused almost exclusively on separating players into groups based on handicap or putt result. Study designs are then based on assumptions on the relationship between handicap, putt result and putting technique. These studies then test for statistically significant differences between groups split on handicap or putting performance. In simple terms, the assumptions used in previous studies are: Handicap assumption: The putting technique of low handicap players is different to that of middle handicap players, and high handicap players. Accuracy assumption: The putting technique of accurate putters is different to less accurate putters. McLaughlin, Best and Carlson (2008) divided 38 players into handicap groups (low=0-9, middle=1018, high=18-27). ANOVAs showed no significant differences (F(2,107)=1.65, p=.2) in putt result data (low 33±33cm, middle 34±24cm, high 45±32cm) and effect size was small (η2=.03) for a 4m putting task. The question of technique was ignored. McCarty (2002) divided less accurate from accurate players based on ball finishing position. Groups were not significantly different on handicap (p=0.98 at 2m, p=0.06 at 4m). McCarty showed groups were very inconsistent; ie. a large variation in how an accurate (and less accurate) putt was achieved. Putt result data was not related to putting technique. Another method for finding different techniques within one sporting activity is cluster analysis. Cluster analysis is the grouping of like items based on defined parameters from the movement of interest. Like patterns are grouped together and, if more than one pattern/techniquue exists, each pattern is shown to be statistically distinct from the other. Cluster analysis has been used in many areas
(eg. biology/taxonomy). In biomechanics its use is less common but it has been used infrequently for over 25 years (eg. Wilson and Howard, 1983). The aim of this paper is to present cluster analysis as a more appropriate tool for the identification of different techniques in a sporting activity. The study will present data from one study in three different ways and the types of errors produced when not using cluster analysis will be discussed. METHODS Putting testing was conducted on the practice putting green at a private golf club in Melbourne. Thirty eight players (handicap 15.3 ±6.9; age 55.3 ±17.8) participated in the study. Each player completed five putts at a hole 4m away. Players putted while standing on a pliance pressure mat (capacitance sensor mat, 16 x 16 sensor matrix, 50Hz; novel gmbh, Germany) and moved off the mat after each putt. 2D video (50Hz) was recorded perpendicular to the putt direction and synchronized with the pliance mat. The pliance mat was validated against an AMTI platform for measurement of centre of pressure (COP). The result of each putt was noted (holed out OR left/right, long/short, distance from hole). The putting stroke was broken down into backswing (BS), downswing (DS), ball contact (BC) and follow through (FT). Data were analysed in three ways, each based on dividing players into groups: 1) handicap; 2) putt result (accuracy); 3) cluster analysis. Nonparametric tests were used to determine significant differences between groups within each analysis method. Putts that contained incomplete data were excluded, thus 108 putts were assessed. Each individual putt
was treated as a separate item in the analysis. A total of 62 parameters were available for assessment. For cluster analysis, the most influential parameters related to medio-lateral centre of pressure movement (COPx). These parameters form the basis of the results and discussion section.
both clusters; ie. use both techniques. The inconsistency of technique within accuracy groups that McCarty (2002) alluded to is further complicated when two techniques are used within the same player. This technique ‘change’ within an individual could arise as a response to the previous trial’s result, or for many other reasons.
RESULTS AND DISCUSSION CONCLUSIONS Table 1 reveals that two distinct techniques were established using cluster analysis. The three methods show inconsistent results with no parameter providing a significant difference between groups for all three methods. Of the three methods, only the cluster analysis groups have scientific validity. The data clearly shows that the assumption of different techniques based on handicap or accuracy leads to Type I (as depicted in Table 1) and Type II errors (most likely to occur in regressions if the two techniques are not dealt with separately). Cluster analysis data shows that although there is a significant difference between the two clusters/techniques for handicap, there is a wider range of handicaps in each group (meaning the split isn’t as simple as high vs low handicaps). These two clusters, or techniques, are defined by parameters related to execution of the skill, rather than the result or the players’ overall golf performance. This study also found that players can appear in
Method Handicap h d Low (n=30) Middle (n=53) High (n=25) Cluster 1 (n= 77) 2 (n=31) Accuracy More (n=54) Less (n=54) Total (n=108)
This study’s findings add weight to the argument that, for scientific studies, players should not be classified based on pre-existing (handicap) or resultbased (accuracy) measures. Cluster analysis is the best way to search for different ‘techniques’ within the same activity. Also, it should not be assumed that one player uses only one of the techniques. Searching for and choosing the correct method for distinguishing biomechanical ‘techniques’ is a critical aspect of biomechanical analysis (for all skills) but is rarely carried out. Type I and II errors can be avoided if no a priori assumptions are made. Cluster analysis is a robust technique that searches for techniques based on execution of the skill. REFERENCES 1. McCarty, J. D. (2002). MSc thesis, Purdue University, USA. 2. McLaughlin, P, Best, R. & Carlson, J. (2008) World Scientific Congress of Golf V,USA. 3. Wilson, B.D. & Howard, A. 1983 J Hum Mvt Stud 9: 71-80.
Table 1: Mean and SD for COPx parameters by group and by study design. Absolute DS BC BS Range DS Range Putt Result COPx Max. COPx Vel. COPx (mm) COPx (mm) (cm from vel. (cm/s) (cm/s) hole)
Handicap
4.6±2.9 6.5±4.8 7.8±6.3
4.5±4.1* 5.2±3.6 8.7±5.6
29.9±24.6* 36.3±26.0 55.0±34.3
7.7±24.5* 23.0±28.9 30.3±36.3
35.3±34.6 34.1±24.1 47.5±32.2
5.9±2.0* 13.7±2.3 22.3±3.4
4.9±2.7* 9.6±7.0
3.9±2.6* 10.6±4.8
25.5±15.7* 71.8±28.3
5.2±16.9* 58.4±22.9
36.8±28.5 39.5±32.3
12.4±5.9* 16.4±6.6
5.8±3.7 6.8±5.7 6.3±4.8
5.5±4.4 6.1±4.7 5.8±4.5
37.1±26.9 40.5±31.2 38.8±29.0
21.2±27.3 19.6±33.8 20.5±30.6
15.8±11.5* 59.3±25.6 37.6±29.5
12.0±5.8* 15.0±6.6 13.5±6.4
*Significant differences between groups within study design