Pilates and balance in young archers: Park 2016 study

Twenty high-school archers from Daejeon, registered with the Korean federation and with at least five years of practice, split into ten who added twelve weeks of mat Pilates to their normal training and ten who carried on with the normal training alone. Three sixty-minute sessions a week, warm-up, work and cool-down, intensity declared between 8 and 13 on the Borg scale, that is from very light to somewhat hard. Two things were measured at the end, static and dynamic balance, on the same board.

Static balance did not move. In the Pilates group the weight distribution went from 51.10 to 50.80 per cent forwards and from 51.10 to 50.20 to the left, neither significant, and the control group did not move either. The only significant difference is between the two groups at the end, on left and right: the Pilates group reaches 50.20 and 49.80, the control group stays at 51.70 and 48.30. One and a half percentage points of weight between one foot and the other.

Dynamic balance moved, and by a good deal. The test measures how far and how fast an archer takes the centre of pressure towards an indicator lighting up in eight directions. In the Pilates group the overall score went from 38.6 to 47.5, forwards from 40.7 to 46.9, right from 38.5 to 50.4, down-right from 31.9 to 42.2. In the control group the overall score stayed where it was, 39.2 against 38.3, with five directions slightly down and four slightly up, and this is the most useful figure in the whole article, because it says that repeating the test three months later does not improve it by itself. In the final comparison between groups, Pilates sits higher forwards, up-right, down-right and overall.

Nobody shot an arrow inside the study. No score, no measurement of the shooting posture, no bow sway: the article opens by saying it wants data useful to competitive performance and to the stability of the shot, and closes with a score on a board. Body composition too, which the abstract declares measured before and after, appears only as a baseline snapshot: after twelve weeks not one value of it is published.

The study ran from December 2015 to March 2016, that is outside the competitive season. Twenty high-school archers from Daejeon, registered as athletes with the Korean federation, with more than five years of career: mean age seventeen years and three months, career 6.6 ± 1.4 years. Ten in the Pilates group, ten in the control group. How they were split is written nowhere: we read that twenty volunteers had understood the purpose of the study and intended to take part, and there is not a line on randomisation, matching or any allocation criterion. The participants' sex is not stated anywhere in the article either.

The programme is twelve weeks, three times a week, sixty minutes a session, ten of warm-up, forty of work and ten of cool-down, on top of normal school training rather than instead of it. Nineteen mat Pilates exercises, from breathing to swimming to the shoulder bridge, tabulated with sets and repetitions that grow every four weeks and with intensity held between 8 and 10 on the Borg scale in the opening and closing phases and between 11 and 13 in the main work. The control group does only its usual training. The comparison, then, is not between Pilates and another activity: it is between three extra hours a week and nothing extra.

Static balance is measured as weight distribution on the board, with the archer standing, looking at a screen and pressing towards a central indicator. Four values are reported as percentages, forwards, backwards, left and right. Dynamic balance is the limits-of-stability test on the same machine: the indicator lights up in random order in the eight directions, movement time and hold are recorded, and out come eight directional scores plus an overall one. Body composition is taken with a bioimpedance device, with an empty bladder and after five minutes of rest.

The statistics are non-parametric. The authors write that they checked "the degree of equal variance" with the Kolmogorov-Smirnov test and found that no data were normally distributed: those are two different things in the same sentence, since that test looks at the shape of the distribution and not at homogeneity of variance. Within groups they use the Wilcoxon signed-rank test, between groups the Mann-Whitney test, with the threshold at 0.05. The between-group comparisons are made on final values only, the baseline differences never go through a test, and across the thirty-nine comparisons no correction is applied.

The most serious problem is in the column of Z values in the dynamic balance table. With ten pairs of measurements, the Wilcoxon signed-rank test has a rank sum running from 0 to 55, with a mean of 27.5 and a standard deviation of 9.81: the most extreme Z that test can produce on ten pairs is 2.80, and pushing the tie correction to the limiting case where all differences have the same magnitude does not get past 3.16. The table reports, for the Pilates group, 3.637 forwards, 4.052 down, 5.655 overall and 5.859 down-right. Four numbers that test cannot give on that sample.

There is a second signal in the same place, and it helps to understand. The asterisks do not follow the Z values. With Z equal to 3.637 the probability under the normal distribution is 0.00028, that is three asterisks, and the article puts two; with 2.644 it is 0.0082, two asterisks, and it puts one; with 4.052 it is 0.00005 and it puts two; with 2.081 it is 0.037, one asterisk, and it puts none. In the control group the down-right value, 2.819, would give 0.0048 and stays without asterisks, while the text describes that group as showing no change although it goes from 26.20 to 30.90, almost a fifth more. Across eighteen within-group comparisons, five carry asterisks that do not match their own Z. The two columns do not come from the same calculation, and the article's conclusions rest on the asterisks.

Static balance, moreover, is not sway: it is weight distribution, and its four values are two. In all eight cells forwards plus backwards makes exactly 100.00 and left plus right makes exactly 100.00, with identical standard deviations inside each pair and mirrored Z values of reversed sign, 1.152 and -1.152, 1.868 and -1.868, 2.923 and -2.923. The result announced in the conclusions, improved static balance on the left and right parts, is therefore one result counted twice. And within the group nothing changed, Z equal to 1.868 and no asterisk: the significance comes from the final comparison between groups, while at the start the Pilates group stood at 51.10 and the control group at 51.60 with no test checking it.

The two groups did not start out alike, and their own table says so. The Pilates group is 175.6 ± 2.1 centimetres tall against 170.7 ± 4.9 and weighs 79.7 ± 5.2 kilograms against 73.2 ± 7.0: redoing the comparison, height gives p equal to 0.012 and weight 0.032. With twenty youngsters split into two groups with no statement of how, five centimetres and six kilograms of difference call for an explanation, and the article does not even carry the test that detects them. On down-right dynamic balance too, the direction with the strongest result, the Pilates group started from 31.9 against 26.2.

One calculation does not close. The Pilates group's body mass index is given as 24.20, but 79.67 kilograms over 1.756 metres makes 25.84; in the control group 73.23 over 1.707 makes 25.13 against the published 25.04. The mean of a ratio is not the ratio of the means, granted, but with standard deviations this small that correction is worth less than two tenths of a point, not one and a half.

On the design the underlying limit remains. The Pilates group did thirty-six extra hours of supervised activity and the control group nothing extra, so any other added activity might produce the same shift on a board: the study does not show that Pilates is needed, or that the core is, it shows that training more beats training as before. Nobody was blinded, neither the youngsters nor those measuring, and the dynamic test depends on volition, because taking the centre of pressure further and faster towards an indicator is a performance one can decide to do better.

Finally the citation apparatus, which can be checked against the article's own reference list in ten minutes. That archers mostly have an unbalanced posture is attributed to a study on pelvis morphology and the Cobb angle in adolescent idiopathic scoliosis; that Pilates improves physical stability, to an article on the design of an instrumented bowling ball; that stabilising exercise improves an archer's shooting performance, to a paper on the morphological characteristics of male bowling players; that respiratory stability weighs more than the mental factor, to a randomised trial of sling exercise for whiplash-associated disorders.

For deciding how to train it offers little. It says that twenty Korean youngsters who add three hours of mat Pilates a week take the centre of pressure further and faster on a board, and it says nothing about what happens on the target, because nobody shot inside the study. Between shifting weight and putting the arrow in the gold lies the whole chain this work does not measure.

What can be taken away, carefully, is three things.

The first is not the Pilates effect, it is the behaviour of the control group: ten archers repeating the same test after twelve weeks with nothing extra do not improve it, five directions slightly down and four slightly up, overall unchanged. Anyone trying a balance programme therefore has a reference point for understanding that an improvement on the board is not simply practice at the board. It is the reason a control group, even of ten people, is worth more than a sample twice the size without one.

The second is that static balance does not move, and that measured this way there may have been nothing to move: these archers already stood at 51 against 49 between one foot and the other before starting, practically even, although the article opens by saying that archers mostly have an unbalanced posture. A nearly symmetrical weight distribution is what one finds in registered youngsters with six years of practice, and watching it as a marker of progress does not promise much.

The third is the dose, the most concrete number in the article: sixty minutes, three times a week, twelve weeks, at light to moderate intensity, added rather than substituted. Thirty-six hours of mat work for a gain nobody has linked to a score.

For a coach the consequence is simple. If Pilates goes into the programme as general base work, no problem: it is extra supervised activity, and at that age it does good anyway. If it goes in as an intervention on accuracy, this study is not the argument to bring. For that, the two things missing here are needed, the score among the outcomes and a comparison group doing another activity for the same length of time.