An analysis of training loads in elite under 18 Australian Rule
Transcript Of An analysis of training loads in elite under 18 Australian Rule
Analysis of Training Loads in Elite Under 18 Australian Rule Football Players
This is the Accepted version of the following publication
Cust, Emily, Elsworthy, Nathan and Robertson, Samuel (2018) Analysis of Training Loads in Elite Under 18 Australian Rule Football Players. Journal of Strength and Conditioning Research, 32 (9). 2521 - 2528. ISSN 1064-8011
The publisher’s official version can be found at https://dx.doi.org/10.1519/JSC.0000000000002392
Note that access to this version may require subscription.
Downloaded from VU Research Repository https://vuir.vu.edu.au/37102/
Journal of Strength and Conditioning Research Publish Ahead of Print DOI: 10.1519/JSC.0000000000002392
An analysis of training loads in elite under 18 Australian Rule football players Emily E. Cust1,2, Nathan Elsworthy1, & Sam Robertson1, 2
1Institute of Sport, Exercise and Active Living (ISEAL),
Victoria University,
Ballarat Road, Melbourne, Victoria, Australia, 8001
2Western Bulldogs Football Club, Whitten Oval,
TED
Barkly Street,
West Footscray,
Melbourne,
P Victoria,
Australia, 3210
E Running title: U18 Australian football training loads
Submission type: Original Investigation Key words: team sports; session RPE; talent identification; internal
C Address for correspondence:
Emily Cust Victoria University,
C Footscray Park.
Ballarat Road, Melbourne,
A Victoria,
Australia 8001
loads;
junior
athlete
Email: [email protected]
Telephone: +61 (0)448810237
Word count: 3,598 Abstract word count: 249 Number of tables: 2 Number of figures: 2
Copyright ª 2017 National Strength and Conditioning Association
U18 Australian football training loads 1 1 Abstract 2 Differences in training loads (TL) between under 18 (U18) Australian Rules football (AF) 3 State Academy selected and non-selected players were investigated. Players were categorised 4 relating to their highest representative level; State Academy selected (n = 9) and TAC Cup 5 level players (n = 38). Data were obtained from an online training-monitoring tool 6 implemented to collect player training and match information across a 20 - week period 7 during the regular season. Parameters modelled included AF skills, strength, and other sport 8 training sessions. Descriptive statistics (mean ± SD) and between-group comparisons
D 9 (Cohen’s d) were computed. A J48 decision tree modelled which TL variables could predict E 10 selection level. Pooled data showed 60% of weekly training duration consisted of AF training
11 sessions. Similar AF TL were reported between State Academy and TAC Cup players (1578
T 12 ± 1264 arbitrary units (AU) v 1368 ± 872 AU; d = .05). While higher TL were reported for
13 State selected players comparative to TAC Cup in total training (d = .20), core stability (d =
P 14 .36), flexibility (d = .44), on-feet conditioning (d = .26), and off-feet conditioning (d = .26). E 15 Decision tree analysis showed core stability duration and flexibility TL the most influential
16 parameters in classifying group selection (97.7% accuracy TAC Cup level; 35.8% accuracy
C 17 State Academy level). Insights of U18 AF players’ weekly training structures, loads, and
18 characteristics of higher achieving players are provided. This study supports the application
C 19 of training diaries and session rating of perceived exertion (sRPE) for TL monitoring in
20 junior athletes.
A 21
22 Key Words: team sports; session RPE; talent identification; internal loads; junior athletes 23
Copyright ª 2017 National Strength and Conditioning Association 1
U18 Australian football training loads 2
24 INTRODUCTION
25 The Australian Football League (AFL) has established a talent development pathway for
26 junior players aimed at identifying, fostering, and progressing players towards an elite
27 Australian Rules football (AF) career. Levels including State Academies and National
28 Championships for age groups ranging from Under 14 to Under 18 years (U14 - U18 years),
29 are implemented nationwide and run along-side each State’s participation pathways. In key
30 relevance to this study, the Transport Accident Commission (TAC) Cup is a Victoria state-
31 wide U18 representative competition for players to compete in high quality football and
D 32 developmental opportunities. The competition acts as one of the primary recruitment grounds
E 33 for selection into the Victorian State Metropolitan or Country teams, National Academy, and
34 scouting process for AFL clubs and semi-professional State league clubs.
T 35
Talent development and training practices for junior elite AF players are evolving to
36 incorporate a more scientific and measured approach as seen in the senior elite competitions.
P 37 The increased use of global positioning system (GPS) technology, individual athlete load E 38 monitoring (25), and online athlete self-reporting applications reflects a greater focus on grass
39 root development of AF players. An increased understanding of physical demands on players
C 40 from previous studies looking into junior elite AF match profiles (2, 21, 22) and athlete loads (12,
41 13) has also allowed for ongoing refinement of coaching practices and athlete management.
C 42 For example, match physical and technical differences between elite U16 and U18 AF
43 players have been reported (24), including contested marks, clearances, total marks, and
A 44 relative distance (m.min-1). Greater statistical information of junior players could contribute
45 to improving progression and retention of talented players into the senior elite leagues. Apart
46 from the use of this data for match play performance enhancement, coaches could further
47 adapt training to suit age level, developmental stage, and playing position. Again, ensuring
48 appropriate loads are administered and effectively monitored.
Copyright ª 2017 National Strength and Conditioning Association 2
U18 Australian football training loads 3
49
Talented players may be exposed to higher training load (TL) in order to complete the
50 required tasks for selection at various levels of sport talent pathways (10). For example, U18
51 TAC Cup players may be involved in local club and school football competitions, or other
52 sports (e.g., basketball), whilst potentially being selected in State and National Academies.
53 The impacts of these additional training loads specifically on U18 AF player development is
54 not yet fully known. By using self-reported training measures, this study will examine the
55 training characteristics of U18 TAC Cup players throughout the 2016 playing season. 56 Previous studies have reported on the physical and match demands of TAC Cup players (12,
D 57 13). But it is not yet known the breakdown of total TL including extra training activities such
E 58 as participating in other organised sports simultaneously. Previous research on junior rugby
59 union players concluded that commitment to several levels of rugby teams, training and
T 60 matches, combined with outside sports participation created numerous high-load and impact
61 sessions throughout a week (10).
P 62
A previous systematic review (7) of the major football codes (American, AF, Gaelic,
E 63 rugby codes and soccer) examining the relationship between workloads, performance, injury,
64 and illness in adolescent male players acknowledged the need for further research in the area.
C 65 Particularly, training does-response relationships and effects of additional training. Results
66 indicated significant positive relationships between physical stress and traumatic injury,
C 67 furthermore that training duration was significantly associated with illness (7). Consistent
68 study results from multiple youth sports indicate a linear relationship between hours
A 69 participated and injury risk; greater than 16 hours per weeks specifically (4). Yet there are
70 changing views with evidence to suggest that appropriately prescribed and monitored high 71 TL will develop physical qualities in athletes that provide a protective effect against injury (8).
72
Copyright ª 2017 National Strength and Conditioning Association 3
U18 Australian football training loads 4
73
The aim of this study was to determine whether differences in TL existed between the
74 selection level of U18 AF players during the regular playing season. Furthermore, to
75 determine which combination of training type parameters would classify a player’s training
76 week and level as either a TAC Cup player or higher selected State-team player. It was
77 hypothesized that higher selected State Academy players would record greater AF specific
78 training and associated developmental training such as strength sessions. This would be
79 accompanied by lower other outside sport involvement comparative to TAC Cup level
80 players.
D 81 E 82 METHODS
83 Subjects
T 84 A sample of 47 players registered with two TAC Cup clubs was available for participation in
85 the study (n = 17 club 1; n = 30 club 2). Participants were categorised into two groups based
P 86 on their highest representative level as supplied by the TAC Cup clubs; State Academy E 87 selected (n = 9; male, age: 16.9 ± 0.3 years) or TAC Cup level (n = 38; male, age: 16.8 ± .8
88 years) therefore not selected in the higher State Academy level. The players trained and
C 89 competed in matches for their TAC Cup club, school team, local team, or State squad based
90 on coaches’ selection, prior commitment requirements, and player availability during the data
C 91 collection period. Training sessions for both TAC Cup clubs were held on Monday, Tuesday
92 and Thursday evenings. The study and its methods were approved by the relevant Human
A 93 Research Ethics Committee. Parental or guardian signed consent was obtained for all players
94 under 18 years of age.
95
96
97
Copyright ª 2017 National Strength and Conditioning Association 4
U18 Australian football training loads 5
98 Experimental Approach to the Problem 99 Data were collected over a 20-week period during the regular playing season of the 2016 100 TAC Cup competition from rounds one to 16 inclusive (including four bye rounds). 101 Participants were provided with access to an online training monitoring tool (Smartabase: 102 Version 4.835, Fusion Sport, Queensland, Australia) for the purpose of self-reporting daily 103 training activity. Prior to the season, players were educated on how to correctly fill out the 104 diaries, including categorising training types and recording RPE scores. Players were 105 instructed to enter individual data each day related to all training undertaken throughout the
D 106 2016 TAC Cup competition (March to August) in the set questionnaire. The completion of E 107 the diaries was self-directed from a player’s perspective which may have created possibility
108 for players to misclassify certain sessions based on their own subjective interpretation of the
T 109 education mentioned above. The training load parameters included for modelling were: AF
110 training – scheduled sessions with their AF team; other sport training – any training or
P 111 competition undertaken with another sport outside AF; core stability – specific core work E 112 conducted in an athlete’s own time from a recommended program provided by the club’s
113 strength coach; strength training – dedicated strength sessions either with their AF club or on
C 114 own; flexibility – dedicated flexibility sessions conducted on own from a recommended
115 program provided by the club’s strength coach; on-feet conditioning – all dynamic
C 116 conditioning (e.g. run intervals, plyometrics); off-feet conditioning – all static or passive
117 conditioning work (e.g. stretching); total training – sum of all training conducted from each
A 118 training type.
119 Procedures 120 Internal TL was calculated through the session rating of perceived exertion (sRPE) method 121 by multiplying the total training duration (min) by the sRPE rating from the CR10 scale (AU) 122 (5). All raw data exported from the Smartabase software was imported into a custom designed
Copyright ª 2017 National Strength and Conditioning Association 5
U18 Australian football training loads 6
123 Microsoft Excel™ spreadsheet (Microsoft Corporation, Redmond, USA), and pre-processed 124 (17). Any identified abnormalities such as incorrectly entered time format data (reporting in
125 hours instead of minutes), or inconsistencies in recording a zero or leaving blank in entries
126 were rectified. Players were coded with an assigned identification number to de-identify the
127 data; and then level coded based on highest squad selection, State Academy (1) or TAC Cup
128 level (2). Cleaned data were organized to show all measures across a single row for each 129 player on each day of data entry provided, and weekly averages calculated. This resulted in 130 726 individual weekly load profiles for analysis.
D 131
E 132 Statistical Analysis
133 Descriptive data are presented as mean ± standard deviation (SD). The effect size (ES) for
T 134 each measure for between group distances was calculated using Cohen’s d statistic on a
135 customised Microsoft Excel™ spreadsheet, indicating a small or trivial (d = 0 - .2), moderate
P 136 (d = .2 - .5), large (d = .5 - .8), and very large (d > .8) effect (3). The confidence interval (CI)
E 137 was expressed as 90% representing the uncertainty in each effect and as probability that the
138 true effect was considerably positive or negative (14).
C 139
In addition to quantifying the differences between the two groups, a supervised
140 learning model was developed to provide a classification prediction for State Academy
C 141 selected and non-selected participants based on TL parameters. Given the uneven group
142 numbers, multiple blank events for some categories as well as ‘zeroes’ recorded in some
A 143 weeks, a number of data transformation techniques were attempted in order to normalise the
144 data. All of these were unsuccessful however, meaning that a non-parametric, machine
145 learning approach was implemented. Specifically, using the ‘RWeka’ package in R (R 146 Computing Environment) (15, 23). A J48 decision tree modelled each of the weekly load
147 profiles included in the dataset to classify player selection level in relation to TL measures.
Copyright ª 2017 National Strength and Conditioning Association 6
U18 Australian football training loads 7
148 All eight load parameters were included in the model, whilst a confidence value of 0.25 was 149 set and a minimum support of 10 instances required in order for a node to split. Model 150 performance was reported as classification accuracy of both groups and compared to the null 151 model.
152
153 RESULTS 154 The breakdown of weekly training duration types indicated that the majority of training for 155 this cohort was AF based sessions followed by strength training (Table 1); which is also
D 156 reflected in weekly sRPE TL (Table 1).
157
E 158
**Table 1 near here**
T 159
**Table 2 near here**
160
P 161
State Academy selected players in comparison to TAC Cup players had higher
E 162 weekly training durations in core stability (ES = 0.40; CI = -0.16 to -0.64), strength (0.23;
163 001 to -0.47), flexibility (0.37; -0.13 to -0.61), on-feet conditioning (0.28; -0.04 to -0.52), and
C 164 off-feet conditioning (0.26; -0.02 to -0.50) (Table 2). State Academy selected players also
165 showed higher weekly training loads in total training (ES = 0.20; CI = 0.04 to -0.44), core
C 166 stability (0.36; -0.12 to -0.60), flexibility (0.44; -0.20 to -0.68), on-feet conditioning (0.26; -
167 0.02 to -0.50), and off-feet conditioning (0.26; -0.02 to -0.50) (Table 2). In breaking down
A 168 training sRPE loads for each training type across four-week blocks between the two groups,
169 marked TL differences showed TAC Cup level players has larger loads in weeks 13, 14 and
170 15 compared to State selected players (Figures 1a and 1b). Other sports reported in the
171 training diaries included volleyball, rowing, swimming, soccer, hockey, tennis, athletics,
172 basketball, bike riding, own gym sessions, and netball.
Copyright ª 2017 National Strength and Conditioning Association 7
U18 Australian football training loads 8
173
**Figures 1a, 1b, and 1c near here**
174
175
Decision tree evaluation analysed a total of 567 training weeks (78.1% of total
176 sample) including TAC Cup level players, and 159 weeks were reported including State
177 Academy selected participants. Results indicate that core stability duration and flexibility TL
178 are the most important interaction in parameters to classifying the two groups (Figure 2). This 179 is shown by the tree terminating down the right side at nodes 1 and 2 after just one branch 180 from the root node, weekly core stability duration greater than 33 minutes to weekly
D 181 flexibility TL. On the left side of the figure, the interaction between higher weekly off-feet E 182 conditioning durations and weekly AF TL is also suggested as a strong predictor of player
183 selection level, classifying TAC Cup level 23 out of the 31 weeks (node 4) and State
T 184 Academy 10 out of the 12 weeks (node 5). The asymmetry in the decision tree output
185 indicates that TAC Cup level and State Academy training behaviour have different nuances.
P 186 There are greater interactions in parameters to classify TAC Cup level players based on their E 187 training characteristics (nodes 2 – 4, 6, 7, 9) than State level players (nodes 1, 5, 8, 10).
188 Model performance was reported as 83.3%, which constituted only a moderate improvement
C 189 on the 78.1% null model. Of this, the model displayed an accuracy of 97.7% in classifying
190 TAC Cup level players (554 of 557 weeks) and 35.8% accuracy in classifying State Academy
C 191 players (51 of 157 weeks).
192
A 193
**Figure 2 near here**
194
195
196
197
Copyright ª 2017 National Strength and Conditioning Association 8
This is the Accepted version of the following publication
Cust, Emily, Elsworthy, Nathan and Robertson, Samuel (2018) Analysis of Training Loads in Elite Under 18 Australian Rule Football Players. Journal of Strength and Conditioning Research, 32 (9). 2521 - 2528. ISSN 1064-8011
The publisher’s official version can be found at https://dx.doi.org/10.1519/JSC.0000000000002392
Note that access to this version may require subscription.
Downloaded from VU Research Repository https://vuir.vu.edu.au/37102/
Journal of Strength and Conditioning Research Publish Ahead of Print DOI: 10.1519/JSC.0000000000002392
An analysis of training loads in elite under 18 Australian Rule football players Emily E. Cust1,2, Nathan Elsworthy1, & Sam Robertson1, 2
1Institute of Sport, Exercise and Active Living (ISEAL),
Victoria University,
Ballarat Road, Melbourne, Victoria, Australia, 8001
2Western Bulldogs Football Club, Whitten Oval,
TED
Barkly Street,
West Footscray,
Melbourne,
P Victoria,
Australia, 3210
E Running title: U18 Australian football training loads
Submission type: Original Investigation Key words: team sports; session RPE; talent identification; internal
C Address for correspondence:
Emily Cust Victoria University,
C Footscray Park.
Ballarat Road, Melbourne,
A Victoria,
Australia 8001
loads;
junior
athlete
Email: [email protected]
Telephone: +61 (0)448810237
Word count: 3,598 Abstract word count: 249 Number of tables: 2 Number of figures: 2
Copyright ª 2017 National Strength and Conditioning Association
U18 Australian football training loads 1 1 Abstract 2 Differences in training loads (TL) between under 18 (U18) Australian Rules football (AF) 3 State Academy selected and non-selected players were investigated. Players were categorised 4 relating to their highest representative level; State Academy selected (n = 9) and TAC Cup 5 level players (n = 38). Data were obtained from an online training-monitoring tool 6 implemented to collect player training and match information across a 20 - week period 7 during the regular season. Parameters modelled included AF skills, strength, and other sport 8 training sessions. Descriptive statistics (mean ± SD) and between-group comparisons
D 9 (Cohen’s d) were computed. A J48 decision tree modelled which TL variables could predict E 10 selection level. Pooled data showed 60% of weekly training duration consisted of AF training
11 sessions. Similar AF TL were reported between State Academy and TAC Cup players (1578
T 12 ± 1264 arbitrary units (AU) v 1368 ± 872 AU; d = .05). While higher TL were reported for
13 State selected players comparative to TAC Cup in total training (d = .20), core stability (d =
P 14 .36), flexibility (d = .44), on-feet conditioning (d = .26), and off-feet conditioning (d = .26). E 15 Decision tree analysis showed core stability duration and flexibility TL the most influential
16 parameters in classifying group selection (97.7% accuracy TAC Cup level; 35.8% accuracy
C 17 State Academy level). Insights of U18 AF players’ weekly training structures, loads, and
18 characteristics of higher achieving players are provided. This study supports the application
C 19 of training diaries and session rating of perceived exertion (sRPE) for TL monitoring in
20 junior athletes.
A 21
22 Key Words: team sports; session RPE; talent identification; internal loads; junior athletes 23
Copyright ª 2017 National Strength and Conditioning Association 1
U18 Australian football training loads 2
24 INTRODUCTION
25 The Australian Football League (AFL) has established a talent development pathway for
26 junior players aimed at identifying, fostering, and progressing players towards an elite
27 Australian Rules football (AF) career. Levels including State Academies and National
28 Championships for age groups ranging from Under 14 to Under 18 years (U14 - U18 years),
29 are implemented nationwide and run along-side each State’s participation pathways. In key
30 relevance to this study, the Transport Accident Commission (TAC) Cup is a Victoria state-
31 wide U18 representative competition for players to compete in high quality football and
D 32 developmental opportunities. The competition acts as one of the primary recruitment grounds
E 33 for selection into the Victorian State Metropolitan or Country teams, National Academy, and
34 scouting process for AFL clubs and semi-professional State league clubs.
T 35
Talent development and training practices for junior elite AF players are evolving to
36 incorporate a more scientific and measured approach as seen in the senior elite competitions.
P 37 The increased use of global positioning system (GPS) technology, individual athlete load E 38 monitoring (25), and online athlete self-reporting applications reflects a greater focus on grass
39 root development of AF players. An increased understanding of physical demands on players
C 40 from previous studies looking into junior elite AF match profiles (2, 21, 22) and athlete loads (12,
41 13) has also allowed for ongoing refinement of coaching practices and athlete management.
C 42 For example, match physical and technical differences between elite U16 and U18 AF
43 players have been reported (24), including contested marks, clearances, total marks, and
A 44 relative distance (m.min-1). Greater statistical information of junior players could contribute
45 to improving progression and retention of talented players into the senior elite leagues. Apart
46 from the use of this data for match play performance enhancement, coaches could further
47 adapt training to suit age level, developmental stage, and playing position. Again, ensuring
48 appropriate loads are administered and effectively monitored.
Copyright ª 2017 National Strength and Conditioning Association 2
U18 Australian football training loads 3
49
Talented players may be exposed to higher training load (TL) in order to complete the
50 required tasks for selection at various levels of sport talent pathways (10). For example, U18
51 TAC Cup players may be involved in local club and school football competitions, or other
52 sports (e.g., basketball), whilst potentially being selected in State and National Academies.
53 The impacts of these additional training loads specifically on U18 AF player development is
54 not yet fully known. By using self-reported training measures, this study will examine the
55 training characteristics of U18 TAC Cup players throughout the 2016 playing season. 56 Previous studies have reported on the physical and match demands of TAC Cup players (12,
D 57 13). But it is not yet known the breakdown of total TL including extra training activities such
E 58 as participating in other organised sports simultaneously. Previous research on junior rugby
59 union players concluded that commitment to several levels of rugby teams, training and
T 60 matches, combined with outside sports participation created numerous high-load and impact
61 sessions throughout a week (10).
P 62
A previous systematic review (7) of the major football codes (American, AF, Gaelic,
E 63 rugby codes and soccer) examining the relationship between workloads, performance, injury,
64 and illness in adolescent male players acknowledged the need for further research in the area.
C 65 Particularly, training does-response relationships and effects of additional training. Results
66 indicated significant positive relationships between physical stress and traumatic injury,
C 67 furthermore that training duration was significantly associated with illness (7). Consistent
68 study results from multiple youth sports indicate a linear relationship between hours
A 69 participated and injury risk; greater than 16 hours per weeks specifically (4). Yet there are
70 changing views with evidence to suggest that appropriately prescribed and monitored high 71 TL will develop physical qualities in athletes that provide a protective effect against injury (8).
72
Copyright ª 2017 National Strength and Conditioning Association 3
U18 Australian football training loads 4
73
The aim of this study was to determine whether differences in TL existed between the
74 selection level of U18 AF players during the regular playing season. Furthermore, to
75 determine which combination of training type parameters would classify a player’s training
76 week and level as either a TAC Cup player or higher selected State-team player. It was
77 hypothesized that higher selected State Academy players would record greater AF specific
78 training and associated developmental training such as strength sessions. This would be
79 accompanied by lower other outside sport involvement comparative to TAC Cup level
80 players.
D 81 E 82 METHODS
83 Subjects
T 84 A sample of 47 players registered with two TAC Cup clubs was available for participation in
85 the study (n = 17 club 1; n = 30 club 2). Participants were categorised into two groups based
P 86 on their highest representative level as supplied by the TAC Cup clubs; State Academy E 87 selected (n = 9; male, age: 16.9 ± 0.3 years) or TAC Cup level (n = 38; male, age: 16.8 ± .8
88 years) therefore not selected in the higher State Academy level. The players trained and
C 89 competed in matches for their TAC Cup club, school team, local team, or State squad based
90 on coaches’ selection, prior commitment requirements, and player availability during the data
C 91 collection period. Training sessions for both TAC Cup clubs were held on Monday, Tuesday
92 and Thursday evenings. The study and its methods were approved by the relevant Human
A 93 Research Ethics Committee. Parental or guardian signed consent was obtained for all players
94 under 18 years of age.
95
96
97
Copyright ª 2017 National Strength and Conditioning Association 4
U18 Australian football training loads 5
98 Experimental Approach to the Problem 99 Data were collected over a 20-week period during the regular playing season of the 2016 100 TAC Cup competition from rounds one to 16 inclusive (including four bye rounds). 101 Participants were provided with access to an online training monitoring tool (Smartabase: 102 Version 4.835, Fusion Sport, Queensland, Australia) for the purpose of self-reporting daily 103 training activity. Prior to the season, players were educated on how to correctly fill out the 104 diaries, including categorising training types and recording RPE scores. Players were 105 instructed to enter individual data each day related to all training undertaken throughout the
D 106 2016 TAC Cup competition (March to August) in the set questionnaire. The completion of E 107 the diaries was self-directed from a player’s perspective which may have created possibility
108 for players to misclassify certain sessions based on their own subjective interpretation of the
T 109 education mentioned above. The training load parameters included for modelling were: AF
110 training – scheduled sessions with their AF team; other sport training – any training or
P 111 competition undertaken with another sport outside AF; core stability – specific core work E 112 conducted in an athlete’s own time from a recommended program provided by the club’s
113 strength coach; strength training – dedicated strength sessions either with their AF club or on
C 114 own; flexibility – dedicated flexibility sessions conducted on own from a recommended
115 program provided by the club’s strength coach; on-feet conditioning – all dynamic
C 116 conditioning (e.g. run intervals, plyometrics); off-feet conditioning – all static or passive
117 conditioning work (e.g. stretching); total training – sum of all training conducted from each
A 118 training type.
119 Procedures 120 Internal TL was calculated through the session rating of perceived exertion (sRPE) method 121 by multiplying the total training duration (min) by the sRPE rating from the CR10 scale (AU) 122 (5). All raw data exported from the Smartabase software was imported into a custom designed
Copyright ª 2017 National Strength and Conditioning Association 5
U18 Australian football training loads 6
123 Microsoft Excel™ spreadsheet (Microsoft Corporation, Redmond, USA), and pre-processed 124 (17). Any identified abnormalities such as incorrectly entered time format data (reporting in
125 hours instead of minutes), or inconsistencies in recording a zero or leaving blank in entries
126 were rectified. Players were coded with an assigned identification number to de-identify the
127 data; and then level coded based on highest squad selection, State Academy (1) or TAC Cup
128 level (2). Cleaned data were organized to show all measures across a single row for each 129 player on each day of data entry provided, and weekly averages calculated. This resulted in 130 726 individual weekly load profiles for analysis.
D 131
E 132 Statistical Analysis
133 Descriptive data are presented as mean ± standard deviation (SD). The effect size (ES) for
T 134 each measure for between group distances was calculated using Cohen’s d statistic on a
135 customised Microsoft Excel™ spreadsheet, indicating a small or trivial (d = 0 - .2), moderate
P 136 (d = .2 - .5), large (d = .5 - .8), and very large (d > .8) effect (3). The confidence interval (CI)
E 137 was expressed as 90% representing the uncertainty in each effect and as probability that the
138 true effect was considerably positive or negative (14).
C 139
In addition to quantifying the differences between the two groups, a supervised
140 learning model was developed to provide a classification prediction for State Academy
C 141 selected and non-selected participants based on TL parameters. Given the uneven group
142 numbers, multiple blank events for some categories as well as ‘zeroes’ recorded in some
A 143 weeks, a number of data transformation techniques were attempted in order to normalise the
144 data. All of these were unsuccessful however, meaning that a non-parametric, machine
145 learning approach was implemented. Specifically, using the ‘RWeka’ package in R (R 146 Computing Environment) (15, 23). A J48 decision tree modelled each of the weekly load
147 profiles included in the dataset to classify player selection level in relation to TL measures.
Copyright ª 2017 National Strength and Conditioning Association 6
U18 Australian football training loads 7
148 All eight load parameters were included in the model, whilst a confidence value of 0.25 was 149 set and a minimum support of 10 instances required in order for a node to split. Model 150 performance was reported as classification accuracy of both groups and compared to the null 151 model.
152
153 RESULTS 154 The breakdown of weekly training duration types indicated that the majority of training for 155 this cohort was AF based sessions followed by strength training (Table 1); which is also
D 156 reflected in weekly sRPE TL (Table 1).
157
E 158
**Table 1 near here**
T 159
**Table 2 near here**
160
P 161
State Academy selected players in comparison to TAC Cup players had higher
E 162 weekly training durations in core stability (ES = 0.40; CI = -0.16 to -0.64), strength (0.23;
163 001 to -0.47), flexibility (0.37; -0.13 to -0.61), on-feet conditioning (0.28; -0.04 to -0.52), and
C 164 off-feet conditioning (0.26; -0.02 to -0.50) (Table 2). State Academy selected players also
165 showed higher weekly training loads in total training (ES = 0.20; CI = 0.04 to -0.44), core
C 166 stability (0.36; -0.12 to -0.60), flexibility (0.44; -0.20 to -0.68), on-feet conditioning (0.26; -
167 0.02 to -0.50), and off-feet conditioning (0.26; -0.02 to -0.50) (Table 2). In breaking down
A 168 training sRPE loads for each training type across four-week blocks between the two groups,
169 marked TL differences showed TAC Cup level players has larger loads in weeks 13, 14 and
170 15 compared to State selected players (Figures 1a and 1b). Other sports reported in the
171 training diaries included volleyball, rowing, swimming, soccer, hockey, tennis, athletics,
172 basketball, bike riding, own gym sessions, and netball.
Copyright ª 2017 National Strength and Conditioning Association 7
U18 Australian football training loads 8
173
**Figures 1a, 1b, and 1c near here**
174
175
Decision tree evaluation analysed a total of 567 training weeks (78.1% of total
176 sample) including TAC Cup level players, and 159 weeks were reported including State
177 Academy selected participants. Results indicate that core stability duration and flexibility TL
178 are the most important interaction in parameters to classifying the two groups (Figure 2). This 179 is shown by the tree terminating down the right side at nodes 1 and 2 after just one branch 180 from the root node, weekly core stability duration greater than 33 minutes to weekly
D 181 flexibility TL. On the left side of the figure, the interaction between higher weekly off-feet E 182 conditioning durations and weekly AF TL is also suggested as a strong predictor of player
183 selection level, classifying TAC Cup level 23 out of the 31 weeks (node 4) and State
T 184 Academy 10 out of the 12 weeks (node 5). The asymmetry in the decision tree output
185 indicates that TAC Cup level and State Academy training behaviour have different nuances.
P 186 There are greater interactions in parameters to classify TAC Cup level players based on their E 187 training characteristics (nodes 2 – 4, 6, 7, 9) than State level players (nodes 1, 5, 8, 10).
188 Model performance was reported as 83.3%, which constituted only a moderate improvement
C 189 on the 78.1% null model. Of this, the model displayed an accuracy of 97.7% in classifying
190 TAC Cup level players (554 of 557 weeks) and 35.8% accuracy in classifying State Academy
C 191 players (51 of 157 weeks).
192
A 193
**Figure 2 near here**
194
195
196
197
Copyright ª 2017 National Strength and Conditioning Association 8