A battery of physical tests identifies the most promising young archers

The summary today

The available studies say something narrower than what is heard in clubs. A series of Malaysian studies from the same research group shows that a few physical measures split young archers into high and low potential with accuracies between 82.5 and 97.5%, but the groups are built from the same measures and the same score that then recognise them: the accuracy is largely circular, and the models go astray mainly on the weaker archers (in the best model based on body measurements alone the specificity is 60%). The only test on archers kept out of the training data involves ten youngsters, and it misclassifies one. The other batteries add no evidence: the one used in three Indonesian clubs on 40 children places nobody below the middle band, so it does not discriminate downwards, and those built with more care, with the judgement of seven experts or a test repeated a day later, show that the instrument is consistent or repeatable, not that it picks the right youngsters. No study has followed the tested youngsters over time to see who reached the top: there is evidence in favour, but it concerns current level and rests on a single sample of fifty archers, analysed several times over.

Also phrased as: "tests to select young archers"; "talent identification in archery"; "talent test battery"; "selecting youngsters for national centres"; "how to spot a promising young archer"

Evidence

• 2015
  Context Archers Published later

  160 primary school pupils in Manokwari enrolled in a youth programme, a battery of physical tests plus a shooting test: boys come out very good in 22.5% and good in 22.5% of cases, girls in 15 and 30%. The categories are built on the distribution of the sample (composite T scores), so the share of good children is a product of the method and does not say how many are promising; regional journal.

  IDENTIFIKASI BAKAT OLAHRAGA PANAHAN PADA SISWA SEKOLAH DASAR DI KABUPATEN MANOKWARI · Nikanor Asaribab, Siswantoyo

• 2016
  Context Archers Published later

  17 young archers from the Terengganu programme, aged 13 to 19: a principal component analysis of age, body mass index, flexibility, push-ups, one-minute sit-ups and estimated VO2max finds two components (KMO 0.70, cumulative variance 64.36%), the second made up of push-ups (0.76) and flexibility (0.75). It shortens the battery but does not compare it with the score; the text also gives a second split of the variance, inconsistent with the table, and the venue has doubtful editorial standards.

  The Application of Principal Components Analysis to Recognize Essential Physical Fitness Components among Youth Development Archers of Terengganu, Malaysia Questionable venue · Rabiu Muazu Musa, Mohamad Razali Abdullah, Ahmad Bisyri Husin Musawi Maliki, Norlaila Azura Kosni, Mainul Haque

• 2018
  For Archers Published later

  50 young Malaysian archers aged 17.0 ± 0.6 from several programmes: a hierarchical cluster analysis of six physical measures (grip strength, jumps, static balance, arm and core strength) plus a six-arrow score splits the group into high and low potential, and an SVM classifier trained on the same measures recognises the group with 97.5% accuracy for the linear, quadratic, cubic and medium RBF kernels, while for the other two models tried no value is reported. The groups come from the same variables that then recognise them, so the accuracy is circular; only the abstract is in the library.

  The identification of high potential archers based on fitness and motor ability variables: A Support Vector Machine approach · Zahari Taha, Rabiu Muazu Musa, Anwar P.P. Abdul Majeed, Muhammad Muaz Alim, Mohamad Razali Abdullah

• 2018
  For Archers Published later

  A book chapter from the same Malaysian group: from body measurements alone a linear SVM, the best of six models, recognises the two performance groups with 88% accuracy, 100% sensitivity and 60% specificity (5-fold cross-validation), so it always recognises the stronger archers and often goes astray on the weaker ones; the worst, fine RBF, classifies every low-performance case as high performance (specificity 0%). The groups come from clustering and the sample is not described.

  Anthropometry Correlation Towards Archery Performance · Rabiu Muazu Musa, Zahari Taha, Anwar P. P. Abdul Majeed, Mohamad Razali Abdullah

• 2018
  Context Archers Published later

  40 children aged 7 to 12 from three clubs in the Sleman district, a battery of body measurements (height, weight, arm span), four physical tests and a shooting test of 6 ends at 15 m: 5 very gifted (12.5%), 24 gifted (60%), 11 fairly gifted (27.5%) and none in the lower bands. A selection that places no child at the bottom does not separate the promising from the rest; the battery is not validated, there are no raw data, regional journal.

  IDENTIFIKASI BAKAT ISTIMEWA PANAHAN DI KABUPATEN SLEMAN · Yudik Prasetyo, Ahmad Nasrulloh, Komarudin

• 2018
  Context Archers Published later

  50 young Malaysian archers (37 male and 13 female), score over six arrows at 50 m: k-means on the score and resting parameters (heart rate, blood pressure, respiratory rate, calorie intake) forms two groups, which a linear SVM recognises with 94% accuracy (MCC 0.869; quadratic 88%, cubic 84%). These are resting measures rather than physical tests, and the groups are built from the same variables that recognise them.

  The employment of Support Vector Machine to classify high and low performance archers based on bio-physiological variables · Zahari Taha, Rabiu Muazu Musa, Anwar P P Abdul Majeed, Mohamad Razali Abdullah, Muhammad Amirul Abdullah, Mohd Hasnun Arif Hassan, Zubair Khalil

• 2018
  For Archers Published later

  Very probably the same 50 young Malaysian archers (37 boys and 13 girls, aged 13 to 20) as the paper with the SVM classifiers: six physical tests (grip strength, vertical jump, standing long jump, static balance, arm and core strength) and one end of six arrows at 50 m. A k-means analysis of tests and score forms the high- and low-potential groups, and a neural network with ten neurons recognises them with 94% accuracy (sensitivity 96.67%, specificity 90%). The groups come from the same measures, no group is held back for testing and nobody was followed over time: it is not an independent confirmation.

  Talent Identification of Potential Archers Through Fitness and Motor Ability Performance Variables by Means of Artificial Neural Network · Zahari Taha, Rabiu Muazu Musa, Anwar P. P. Abdul Majeed, Mohamad Razali Abdullah, Mohd Hasnun Arif Hassan

• 2018
  Context Archers Published later

  Chapter of a book by the same Malaysian research group: five physical tests (grip, arm, core, static balance, vertical jump) combined with a questionnaire on coping with stress (ACSI) and the shooting score form two groups by k-means, which a cosine-distance k-NN recognises with 82% accuracy (sensitivity 90.32%, specificity 68.42%). The figure of 50 archers is inferred from the table, and the tests almost entirely match those in the paper on the 50 young Malaysians. Context rather than support: with the questionnaire included, the accuracy cannot be credited to the physical tests alone.

  Psycho-Fitness Parameters in the Identification of High-Potential Archers · Rabiu Muazu Musa, Zahari Taha, Anwar P. P. Abdul Majeed, Mohamad Razali Abdullah

• 2018
  Context Archers Published later

  50 young Malaysian archers (37 boys and 13 girls, aged 13 to 20), one end of six arrows at 50 m with sensors: body sway, bow movement, activity of one finger flexor and one extensor, static balance. A k-means analysis of these measures and the score forms two groups, which three SVMs recognise in 70% (the finest model) and 64% of cases. These are biomechanical measures of the shot rather than physical tests; in the charts the only clear difference between the groups is the score, and the group sizes are not given, so 64% may not beat a model that always gives the same answer.

  The Application of Support Vector Machine in Classifying Potential Archers Using Bio-mechanical Indicators · Zahari Taha, Rabiu Muazu Musa, Anwar P. P. Abdul Majeed, Mohamad Razali Abdullah, Muhammad Amirul Abdullah, Mohd Hasnun Arif Hassan

• 2018
  Context Archers Published later

  50 young Malaysian archers (37 boys and 13 girls, aged 13 to 20): a cluster analysis of score and resting or self-reported measures (heart rate, blood pressure, breathing rate, calorie intake) forms three potential groups, and a k-NN recognises them in 84% of cases with one neighbour, 78% with ten and 56% with a hundred, when it puts everyone in the largest group. As in the companion paper with the SVM, the variables are not physical tests and the groups come from the same measures that recognise them.

  Classification of High Performance Archers by Means of Bio-physiological Performance Variables via k-Nearest Neighbour Classification Model · Zahari Taha, Rabiu Muazu Musa, Anwar P. P. Abdul Majeed, Mohamad Razali Abdullah, Ahmad Fakhri Ab. Nasir, Mohd Hasnun Arif Hassan

• 2019
  For Archers Published later

  50 young Malaysian archers aged 17.0 ± 0.56: a regression on six physical measures finds only core strength and the vertical jump linked to a one-end score, a cluster analysis of these measures and the score forms high and low potential groups, and a weighted k-NN recognises them with 82.5 ± 4.75% accuracy (logistic regression 80%). Tried on 10 archers not used in training, it misclassifies one of the 3 low-potential archers and none of the 7 high-potential ones. The definition of the groups remains circular; age, tests, make-up (37 male and 13 female) and shooting test (six arrows at 50 m after four practice shots) match those of the group's 2018 papers, so it is in all likelihood the same sample.

  A machine learning approach of predicting high potential archers by means of physical fitness indicators · Rabiu Muazu Musa, Anwar P. P. Abdul Majeed, Zahari Taha, Siow Wee Chang, Ahmad Fakhri Ab. Nasir, Mohamad Razali Abdullah

• 2022
  Context Archers Published later

  A battery of nine tests (height, weight and arm span, kinaesthetic perception, sit-and-reach, balance, wall sit, side lying, grip dynamometer, VO2max by shuttle test, a shooting test) rated by 7 Indonesian experts: Aiken's V from 0.809 for the testing procedures to 0.952 for the relevance of the items, above the threshold of 0.76 to 0.86. Tried on 61 archers from three clubs, the internal consistency of the 11 items is moderate (Cronbach's alpha 0.635). It validates the instrument, not whether it picks out the promising; conference proceedings.

  Validity and Reliability of Talent Identification Instrument in Archery · Wisnu Nugroho, Endang Rini Sukamti, Ratna Budiarti, Sb. Pranatahadi

• 2022
  Context Archers Published later

  31 primary school children from Yogyakarta and Central Java enrolled in a club, a battery of ten tests including a shooting test of 6 ends at 15 m: 12 very gifted, 5 gifted, 8 fairly gifted, 6 not gifted and none not gifted at all. Frequencies only, with no check on what those children go on to become; regional journal.

  Analysis of Anthropometry, Physical Conditions, and Archering Skills as the Basic for Identification of Talent in the Sport of Arrow · Yudik Prasetyo, Okky Indera Pamungkas, Heru Prasetyo, Susan Susanto

• 2025
  Context Archers Published later

  14 young recurve archers at an Indonesian national youth centre, a battery of seven specific tests (qualification and elimination scores, one-minute rapid shooting at 5 m, maximal draws in one minute, holding at the clicker, a level test, shooting to exhaustion at 5 m): repeated a day apart it gives r = 0.893 (p < 0.01). It says the instrument measures repeatably, not that it predicts who will go far; regional journal.

  Development of an Archery Technique Test Instrument: Objective Evaluation to Improve Athlete Performance · Rofa Ichsandi, Riko Adi Baskoro, Agung Hadi Saputro, Muhammad Nur Aziz Andriawan

Status history

1. 2015 Not studied The first battery applied to children tests nothing: of 160 primary school pupils in Manokwari, given physical tests and a shooting test, about 45% come out good or very good, but the categories are built on the distribution of the sample itself, so that percentage is a product of the method rather than a measure of how many are promising. There is no evidence on the question yet.
2. 2016 Not studied In 17 young archers from the Terengganu programme a principal component analysis reduces six physical measures to two clusters, one made up of push-ups and flexibility. It is a first step towards choosing which tests to keep, but no measure is compared with the score and the venue has doubtful editorial standards: the claim stays not studied.
3. 2018 Reported The first evidence in favour arrives, and its limit with it. Two papers from the same Malaysian research group split young archers into high and low potential starting from physical tests (97.5% accuracy) and from body measurements alone (88%), but the groups are built on the same variables that then recognise them, and the best anthropometric model goes astray four times in ten on the weaker archers. A third paper does the same with resting heart rate and blood pressure. The same year the group publishes four more papers that very probably reanalyse the same fifty archers: a neural network on six physical tests plus the score recovers the two groups in 94% of cases, a classifier that adds a stress-coping questionnaire to the tests reaches 82%, and two models built on resting measures or biomechanical sensors reach 84 and 70% in their best version. More analyses of the same data do not add up to more evidence. Also in 2018, a battery used on 40 children from Indonesian clubs places nobody below the middle band, so it does not discriminate downwards; but it is a different battery, unvalidated and with no outcome to predict, and it says nothing about the Malaysian one. The evidence in favour is limited and circular: reported.
4. 2019 Reported A new algorithm, on fifty young Malaysian archers with the same mean age, the same make-up (37 boys and 13 girls) and the same tests as in 2018, recognises the two groups with 82.5% accuracy: in all likelihood this is a fresh analysis of the same data rather than a replication. It does make the first attempt to escape the circularity: tried on ten archers kept out of the training data, it misclassifies one of the three low-potential ones. The groups are still defined by the same measures and the score, and nobody has followed the archers over time: the status stays reported.
5. 2022 Reported The battery is built with more method: seven Indonesian experts rate nine tests, from body measurements to a shooting test, with an Aiken's V between 0.809 and 0.952, highest for the relevance of the content and lowest for the practicality of the procedure. Tried on 61 archers from three clubs, however, the internal consistency of the eleven items is only moderate (alpha 0.635), and another battery on 31 children returns only how many are very gifted, gifted or not gifted. A shared instrument now exists and only partly holds up, and nobody has yet followed the selected youngsters over time: reported.
6. 2025 Reported A battery of seven recurve-specific tests, tried on 14 athletes at an Indonesian national youth centre, gives stable results when repeated a day later (r = 0.893). It is an internal validation, not a prediction: it says the instrument measures repeatably, not that whoever does well today will go far, and the status stays reported.

Related

Related articles

• Talent identification: tests and assessment batteriesWhat talent tests in archery actually measure: coaches' ratings, physical tests, attention, years of practice and predictive models, and where their limits lie.

Research lines

• 48 studiesWhat predicts the score, and how talent selection tried to use it
• 63 studiesThe archer's physical profile: strength, endurance and size

Other claims in this domain

• ReportedA few weeks of archery improves children's balance1 study · 32 archers measuredStatus since 2009
• ReportedA mobility and stretching programme improves an archer's flexibility1 study · 30 archers measuredStatus since 2025
• ReportedA spell of reduced training makes the score drop1 study · 9 archers measuredStatus since 2021
• ReportedA strength programme away from the range increases an archer's measured strength15 studies · 316 archers measuredStatus since 2018
• ReportedA yoga programme raises an archer's score4 studies · 250 archers measuredStatus since 2021

See all claims on physical training