Professional Documents
Culture Documents
Abstract In this paper we study the golf putt learning procedure for unskilled
players. Segmented-trend estimates of the learning process are obtained via Fourier
truncated series. The geometric error distribution is modeled and statistical tests are
performed for the localization of the putting strokes. The experimental results provide
a sound basis for ameliorating the teaching of this sport task.
1 Introduction
The process of motor learning is, fundamentally, inferred by the level of motor perfor-
mance when a motor action or task is carried out. In the most of the cases, it is done by
analyzing the performance or the learning curve. Then, taking into account its specici-
ties, we characterize the evolutive process of acquisition of capacities and the learning
method.
Key words and phrases. motor variability, approximation function, Fourier series, error ellipse, linear
regression, statistical tests, learning curves, motor control.
strokes done by 5 unskilled golf players, in order to evaluate the effect of 1050 putting
strokes (for each one) in the learning of this golf movement.
Using the extensive series of data for each novice player, we suggest and make use of,
as an alternative to the method presented in [1], a novel method for the estimation of the
learning curve, which calls truncated Fourier series in order to get the turning points of
the learning procedure.
Moreover, we explore the geometric error positions in order to evaluate the learning
performance, namely obtaining the best least squares error ellipses and we study the dis-
tribution among quadrants by testing, statistically, its equiprobability.
The observed variability of the motor performance leads us to establish new approaches
and methodologies in order to obtain the full information, at it is possible, provided by
the experiments, pretending to be close to the real data given by the extrinsic measures,
such as the radial errors, in particular when we deal with novice players.
This article is comprised by ve sections. In the rst one, we describe the main novice
player of this work. The second one is devoted to the explanation of the held experimen-
tal procedure. In the third one, we introduce a new method for approximating by specic
functions the raw data of radial errors. The geometric error distribution and some statisti-
cal tests are described and its results are presented in section number four. Finally, in the
last section, we establish some conclusions and pave the way to future work in this area.
Figure 1: Novice players, hole positions and the dened coordinate system.
Figure 2: Error points and linear tendency of the novice player one.
So, in order to obtain a function describing the error behaviour of the novice player, we
suggest the method presented below, based on an approximation function. This function
has several purposes: to determine the change points of the learning trends, to conclude
about the learning evolution and, even, to establish corrective measures of the learning
task.
Therefore, for reasons of numerical stability and capacity of prevision, we suggest an
alternative method to the one presented in [1].
We propose for segmented-trend estimate of the learning trend the following method,
with three steps:
n
X j2 j2
qn (t) = a0 + aj cos t + bj sin t (1)
j=1
T T
t0 t1 t2 t3 t4 t5 t6 t7 t8 t9
0 55 217 327 421 563 716 807 934 1050
Finally, after performing the last step of our process, we obtain an estimate of the lear-
ning trend as shown in gure 3.
220
200
180
160
140
Radial Error (cm)
120
100
80
60
40
20
0
0 100 200 300 400 500 600 700 800 900 1000
Attempt
and has major changes along the learning task. We observe, also, a signicative improve-
ment in what concerns to performance in the rst 327 putting strokes. Moreover, the best
series of results occurs in the penultimate series of essays. We see, in the constructed seg-
mented line, three platforms of learning and performance, happening two of them before
the putting stroke number 422.
We think that the rest stops between all the six series of 1075 putting strokes have
different effects in the performance of the novice player. As an example, in the last series
we observe a degradation of the performance in opposition to what occurs in the fth
series. The reasons for this behaviour of observed performance deterioration, could be
effects of boredom and, eventually, accumulation of fatigue.
Typically, in the case of golf, a novice player produces a error scatter roughly seen as
an ellipsoidal gure. Moreover, the ellipse shape has a major axis approximately aligned
with the putting stroke direction. Furthermore, the hole point is not near the ellipses
center but close to one of its focus. This behavior is expected as a consequence of the
specicity of this sport task.
Taking that into account, for each novice player we obtained a scatter diagram and
constructed the best least square t error ellipse. With that purpose, we applied a least
square method in order to nd the best ellipse parameters in the least square sense. The
axis lengths and an angle dening the ellipses orientation. These results could provide
useful information prior to study, planning and correcting the learning task, namely to
adjust the direction and the strength of the putting stroke. In order to illustrate the latter,
in gure 5 we show the error ellipse for the novice player one, together with the scatter
diagram.
Figure 5: Scatter diagram and best least square t ellipse for novice player one.
With the purpose of comparing the novice players error behaviour, we can calculate and
plot all the error ellipses in the same diagram. We can not forget that the available data
for each novice player is not obtained by the same experimental set (different positions
and directions of putting stroke). However, for novice players using the same initial
conditions, we think that this is a valid and measurable comparison (see the illustration in
gure 6).
As stated before, the results provided by the ve novice players show a real error ten-
dency with a shape of an ellipse. Essentially, the novice players give privilege to the error
in length in excess, meaning with the above sentence that they putting stroke the ball more
often with excessive strength than with reduced vigor, because, in the latter case, the ball
stops before the hole. This kind of strategy is recommended as adequate in learning and
practice of the putting stroke task. As expected, the error in direction is much smaller
than in the distance.
Bearing in mind that this kind of experimental work intends to pave the way to the
improvement of the learning process, we also made a brief statistical study of the error
positions distribution. First, we were interested in get information about the distribution
between the four quadrants of the dened coordinate system. The obtained histogram is
shown in gure 7.
With that purpose, we made a Chi-square statistical test in order to check if the quadrant
distribution is uniform (same expected probability to fall in each quadrant). The SPSS
Figure 7: Histogram of the error distribution for the quadrants of the novice player one.
Figure 9: SPSS Chi-square tests output for equiprobability between the rst and the se-
cond quadrants and between the third and the fourth quadrants.
equiprobability between right and left side of the hole and strength of the putting stroke
to put the ball before or beyond the hole. As it can be seen in gure 10, the left and right
directions are equiprobable but the positions beyond the hole are much more probable
three and four but the fth one results were not compatible with the previous ones due to
the orientation of the putting stroke direction.
Figure 10: SPSS Chi-square tests output for equiprobability for direction and strength of
the putting strokes.
The obtained statistical results lead us to the following (learning) conclusions. Never-
theless the dispersion of the putting strokes is huge, in fact there is no tendency to the
left or to the right side of the hole (they are equally distributed). Moreover, there is no
systematic direction error but only a very high dispersion of the putting strokes. In what
concerns to the strength of the putting stroke, it is predominantly high, leading to get
positions beyond the hole.
points.
It is also important to study the correlation between other variables, such as the type of
putter (and its direction) and the error positions.
Another approach to make the study more real it is to take into account the shape of the
surface of the green, specially the slopes of the green.
For future research, we think that is necessary to consider learning effects in tests of
retention and transfer, as well as the analysis of traditional performance measures such as
the variable, constant and absolute errors.
References
[1] A. Diniz, J. Barreiros, R. Benda and N. Crato, Modelling learning curves: A segmented-trend
method, Proceedings of the International Symposium on Measurement, Analysis and Modeling:
319-324 (2007).
[2] E. Maor, Trigonometric Delights, Princeton University Press, Princeton, USA, 2002.
[3] J. Verbesselt, R. Hyndman, G. Newnham and D. Culvenor Detecting trend and seasonal changes in
satellite image time series, Remote Sensing of Environment 114 (1): 106-115 (2010).