Authors: Pierre-Luc Yao1, Vincent Huard Pelletier1, and Jean Lemoyne1

1 Department of Human Kinetics, Université du Québec à Trois-Rivières, Trois-Rivières, QC, Canada

Corresponding Author:
Pierre-Luc Yao, PhD
3351 Boulevard des Forges
Trois-Rivières, QC, Canada, G8Z 4M3
[email protected]
819 376-5011, ext. 3793

Pierre-Luc Yao, PhD a lecturer and internship coordinator in the Department of Human Kinetics at Université du Québec à Trois-Rivières in Trois-Rivières, Québec. His research interests include psychometrics, sport retirement impacts and athlete development.

Vincent Huard Pelletier, MSc, PhD(c) is currently a doctoral student at Université du Québec à Trois-Rivières. Vincent research interest include athlete development, physical activity behavior amongst athletes.

Jean Lemoyne is professor of physical education in the Department of Human Kinetics at the Université du Québec à Trois-Rivières. His research interests are practice of sport amongst teens and young adults, performance evaluation in sports, advanced statistics in sports.

Testing the predictive validity of combine tests among junior elite football players: an 8-yr follow-up

ABSTRACT

Purpose: The objective of this study was to assess the relationship and contribution of physical performance test results on the final selection of an elite under-18 football selection camp. Methods: Data were drawn from 2 876 players divided into seven position groups (DB, DL, OL, LB, QB, RB, and WR) collected over an 8-year span. Players’ evaluations included performance tests (10-yd dash, 20-yd dash, 40-yd dash, 20-yd pro agility shuttle, 3-cone drill, broad jump, vertical jump, power max test) and anthropometric measures (height and weight). Student t tests were calculated for selected and non-selected groups for all positions. Results: Mean comparisons showed that for most measures, selected players obtained significantly better results than non-selected players. Linear regression models were generated for all groups, and every position was found to have its own unique prediction model. The best models were those of the DL (R2 = 0.222), OL (R2 = 0.207) and LB (R2 = 0.204), and the overall explained variance for each model was considered low (R2 = 0.173). Weight, height and 40-yd dash were the most predominant factors in all models. Conclusion: Individually, selection camp results effectively discriminate between selected and non-selected players; together, however, they explain only a limited part of the final selection for each position. Applications in sport: These results suggest that the predictive capacity of the football combine could be improved in terms of the selection of elite football players.

Keywords: football, selection, physical testing, regression 

INTRODUCTION

Football requires multiple attributes conducive to success based on either individual or team performance.  Football teams competing at a high level must look for individuals to help them achieve victory. Very little is known to date about predictors of talent in elite sport contexts (10). A talent identification process, in fact, is used to find the best possible or most promising athletes to meet their team’s needs. This choice is often based on pre-determined criteria to objectively and subjectively distinguish athletes who possess abilities conducive to the success of a sport organization.

In most cases, physical tests measure criteria based on functional skills and fitness. Coaches or recruiters use the results to better appreciate football playing ability (FPA) and make decisions regarding team composition. The relationship between football playing ability and physical testing has garnered a good deal of attention over the past few decades (8). In North America, the “combine” testing approach is a well-established process (9) used for the selection of college, amateur and professional players. Most studies on this subject have been conducted in the United States at both the professional (NFL) and college (NCAA) levels (5, 9, 16, 21, 22), with a few studies at the high school level (8, 9) as well. The usefulness of the combine has already been questioned regarding its lack of relationship to future performance in the NFL (11). It may be no indication of talent on the football field (19).

To our knowledge, no comprehensive study has yet been done on football selection in the province of Québec (Canada). Over 14 000 student-athletes practice this sport in Québec (20). Interestingly, teams from Québec perform best at the varsity sport level, winning 7 of the 10 last Vanier Cups (U Sport championships), the Canadian equivalent of the BCS championships in the United States (24). Moreover, Québec’s under-18 provincial teams (Team Québec) have won six gold medals and finished in the top three at the Football Canada Cup every year since 2008 (6). Considering that most of these teams follow a process to recruit the best players possible, one may reasonably ask what factors influence the selection process. Team Québec holds recurrent selection camps, the largest in the province, in which some 350 players perform multiple physical tests. To date, the relationship between the results of selection camp rosters and physical testing has not been studied in a Québec context; doing so could shed light on how certain test results assist in the selection of players for an elite football team. The combine does in fact have certain advantages: it can assess the skill level required to be drafted at the professional level for each position, give coaches opportunities to gather information of possible interest about players, and provide the minimal physical requirement for playing at the next level (16).

This study aims to verify the relationship between the performance results of athletes at selection camps and the final selection of players for an elite under-18 football team. More specifically, its objectives are twofold: 1) to validate the discriminant and predictive capacity of the combine testing done at Team Québec’s selection camps, and 2) to analyze and evaluate the contribution of each test to predicting the outcome of selection across different football positions. The purpose here is to inform athletes and coaches about the criteria and thresholds most likely to help them identify their province’s best football prospects.

METHODS

Experimental approach

The present study design proposes a secondary data analysis of the data collected from seasons 2011 to 2018 at Football Québec. The data were aggregated for the 8-year study period to obtain a large and representative sample. The combine testing results, and anthropometric measures (height and weight) were identified as independent variables used to predict the outcome of selection status (selected or not) on Team Québec’s roster. The selection status (dependent) was obtained from the 2011 to 2018 Team Québec rosters, and each player in the sample was assigned to either the selected (1) or the unselected (0) group. The first objective was to see if the combine’s tests could determine affiliation to each group. The second was to estimate a predictive model (multiple regression) for each position and find the best possible combinations of physical-technical attributes for predicting selection.

Participants

The present sample includes 2 876 players (100% boys: 16.65 ± 0.53 years old) who were tested between seasons 2011 and 2018 at Team Québec’s selection camps. The sample was sub-categorized into seven groups according to player position: defensive back (DBs, n = 580), defensive line (DL, n = 415), linebacker (LB, n = 381), offensive line (OL, n = 383), quarterback (QB, n = 218), running back (RB, n = 316) and wide receiver (WR, n = 583). All players played in high school (99%) or at the pre-university level (1%) (known as college in Québec). The average selection rate for all positions was 11.68% ± 2.2%. The Institutional Ethics Committee of the Université du Québec à Trois-Rivières approved this study (Certificate number: CDERS-17-11-06.11).

Procedures

High school players from different competition levels are invited to perform at Team Québec’s selection camp each year. The performance includes game situation drills and standardized physical tests. A qualified physical trainer and football coaches oversee testing protocols at each Team Québec tryout. All coaches involved with evaluations are from higher football categories (pre-university, junior and university).

Variables and instruments: Football Quebec’s combine testing protocol

As for most elite teams, the combine testing approach of Football Québec’s under-18 team consists of a protocol conducted each year to ensure selection of the best players possible. To observe stakeholders’ current protocols, we used data from the results of testing protocols from seasons 2011 to 2018. In the case of tests administered twice, the best result was considered.

40-yard dash

The 40-yard dash (40Y) is a test of acceleration and maximal speed on a 36.58 m distance used in different kinds of performance-oriented events (14).  Before performing the test, the athlete must have time to warm up and run at submaximal speed for several minutes. Once ready, he positions himself on the starting line with a 3-point stance and waits for the auditory signal to sprint as fast as possible in a straight line directly towards the end line (2). In this specific protocol, a split time is also recorded after 10 yards (9.15 m) to assess the athlete’s speed and acceleration. Antecedent research establishes the reliability of the 40Y (15).

Vertical Jump

The vertical jump (VJ) is a common measure of lower body muscular power. In this study, vertical jumps were performed using the Vertec device (JUMPUSA, Sunnyvale, CA, USA). The vertical jump is an accessible and highly popular field test used to assess jump height (3). To perform the test, the athlete quickly flexes his knees and hips, swings his arms backward and jumps explosively. To register a valid score, he must hit the highest possible horizontal vane of the Vertec device with one or both arms. Vertical jump height is calculated by subtracting the player’s standing reach height from the height of the highest vane touched. The best of three trials is recorded to the nearest half inch (2). Young, Macdonald, Heggen and Fitzpatrick (25) report an intraclass correlation coefficient of 0.94 and a coefficient of variation of 3.8% when measuring vertical jump height using a device like the Vertec.

Standing broad jump

The standing broad jump (BJ), like the vertical jump, is a test of lower body power. However, the two jumps have distinct biomechanical and functional differences, which justifies the choice to test them both (18). To perform the test, the athlete produces a countermovement and jumps forward as far as possible. He must land on his feet for the score to be registered. If he fails to do so, the trial must be repeated. The best of three trials is recorded to the nearest 1cm (2). Antecedent research shows acceptable validity for the standing broad jump test (4).

Pro Agility Shuttle

The Pro Agility Shuttle (20S) is a well-known performance test that has long been used on the field as an indicator of acceleration, deceleration, anaerobic power and change of direction (16). To perform the test, the athlete sprints 5 yards in one direction, touches the line with one hand, turns back and then sprints 10 yards in the opposite direction before touching the line. He then sprints back to the start/finish line and the time recording stops. The athlete must perform two trials, one with movement initiation from the left and one from the right, and the best attempt is used for analysis (2).

Three-Cone Drill

The Three-Cone Drill (3C) is a hand-timed field test that measures power, agility and change of direction (16) and which has been validated in the past (12, 23). Prior to the execution of the test three cones are placed in an ‘L’ shape, 5 yards from each other. To perform the test, the athlete positions himself behind the starting line (at cone #1) in a 3-point stance. He then sprints directly in front of him to cone #2 and touches it before returning to cone #1. Without stopping, the athlete sprints back to cone #2, corners it and then sprints to cone #3, which is 5 yards lateral to cone #2. Still sprinting, the athlete returns to cone #2, corners it and returns to cone #1 as fast as possible to complete the test (16).

MAXX Test

The MAXX test is not as common as the previous tests and, to our knowledge, has not often been used in another known Quebec football organization. The test is performed on a HI Trainer, a non-motorized treadmill device that aims to assess a player’s power input and acceleration (17). To perform the test, the athlete leans against the chest support pad and sprints as fast as possible. His max power output and the time required to achieve it are recorded by the HI Trainer and serve as two distinct results in this study. In the present sample, HI Trainer was administered only to DL, LB and OL.

Statistical analyses

Each sample’s data were verified for presence outliers and winsorization was applied as needed.  The asymmetrical distributions of all variables were adjusted for normality using a log transformation. Data analyses involved descriptive statistics and group comparisons (e.g., student’s t tests) to attest the discriminant capacity of each component of the testing protocol (see Table 1). For the study’s first objective, regression equations and total explained variance (R2) were generated using the backward stepwise multiple regression procedure to find the best predictive model for each position. The dependent variable was the outcome of the selection project, which was binary (selected or not). For the second objective, Pearson r and regression models were estimated for all seven positions to provide an accurate prediction for each type of player. All statistical analyses were obtained using SPSS version 26 (IBM Statistics), and statistical significance for regression coefficients and t tests was set at p ≤ 0.05.

RESULTS

Objective 1: Group affiliation

Student t test results were used for each position to determine if selected players had better test results than those not selected. Selected players were significantly heavier and taller except for QBs (HT: t214 = -1.454, p = 0.147) and RBs (HT: t311 = -1.709, p = 0.088). Selected players had significantly better times at the 40-yard dash and the Pro agility shuttle except for the OLs (40Y: t370 = 0.844, p = 0.399; 20S: t378 = 1.617, p = 0.107). Max power output was significantly better for the selected DLs (t406 = -3,697, p < 0.01), LBs (t358 = -3.106, p < 0.01) and OLs (t369 = -6.457, p < 0.01). On the other hand, no significant difference was found for time to max power for these three positions.

Objective 2: Tests’ contribution to selection

Regression equations were obtained to find the best combinations of physical attributes to explain selection in Team Québec (Table 1). Weight (WT) was a predictor included in the seven models, as was the 40-yard dash with exception of the OL. In contrast, broad jump (BJ) and max power output (PW) were only included in the OL and the LB respectively, and the speed to power output (SP) was not part of any regression model.

Table 1: Linear regression model, standardized coefficients, and total explained variance by position

  WT (lb) HT (in) 10Y 40Y 20S VJ BJ 3C SP PW R2
DB 0.191** 0.098 0.116 -0.196* – 0.165** – – – – 0,154
DL 0.286** 0.270** – -0.218* -0.181* – – 0.127 – – 0,222
LB 0.245** – – -0.162* -0.116 – 0.191** – – – 0,204
OL 0.149* 0.255** – – – – – -0.104 – 0.188** 0,207
QB 0.266** – – 0.137 – 0.396** – – – – 0,148
RB 0.231** – – -0.397** -0.117 -0.179* – – – – 0,137
WR 0.119 0.107 0.170* -0.276** -0.194** – – – – – 0,140

Note. DB = Defensive back, DL = Defensive line, LB = Linebacker, OL = Offensive line, QB = Quarterback, RB = Running back, WR = Wide receiver, WT = Weight, HT Height, 10Y = 10-yd dash, 40Y = 40-yd dash, 20S = 20-yd shuttle, VJ = Vertical jump, BJ = Broad jump, 3C = 3 three-cone drill, SP = Maxx speed output, PW = Maxx power output. Significant regression coefficient: * p < 0.05, ** p < 0.01.

The average amount of explained variance for all models was R2 = 0.173 (17%), ranging between R2 = 0.140 and R2 = 0.222: DB (R2 = 0.154), DL (R2 = 0.222), LB (R2 = 0.204), OL (R2 = 0.207), QB (R2 = 0.148), RB (R2 = 0.137), WR (R2 = 0.140). These results are illustrated and summarized in Figure 1.

Figure 1:

Figure 1

Note. Position-specific predictors for team selection for Team Québec’s football (2011-2018) Each position represented using either a square (defense) or circle (offense), is associated with their predicting variable included their specific selection prediction model. All predictors are significant at p < 0.05

DISCUSSION

The main objective of this research was to assess the predictive capacity of the physical tests used in Équipe Québec football selection camps to predict players’ selection to the team.  The overall sample was divided into position groups, which formed the basis of the analyses performed. T-tests provided a first indication of the discriminant capacity of the physical tests. Most of the comparisons (44 out of 62; 71%) were significant and favoured the selected group suggesting that testing protocols are somewhat relevant for talent selection. This differs from Ghigiarelli (8) when comparing recruited high school football players with those highly recruited. His study shows that the highly recruited players had better results than their recruited counterparts on 16 out of 50 tests and measures. This difference in results may be due to the classification process and recruiting criteria. Notably, all selected players regardless of position were heavier than those in the non-selected group, and most were taller except for QB and RB.  This result was anticipated for positions such as OL and DL since both these groups are expected to be morphologically massive and sturdy and are required to engage other heavy opponents at every play (7). Selected players were heavier and taller for most positions, which accords with Ghigiarelli (8), who suggests there is a relationship between larger physical size and selection potential among football players. Bigger players may have reached or will potentially reach a physical maturity benchmark comparable to the pre-university or university level and may therefore find favour in the eyes of higher category coaches (8). The selected players were faster at the 40-yard-dash and the 20-yard shuttle test except for the OL and had a better acceleration on 10 yards except for the OL and QB. This was expected because speed, acceleration, and agility what an athlete needs to play football from the very start of his development (7, 16, 22). These results serve as a reference point for future selection camps as regards the basic level of performance expected for each position to be selected to the team (7). However, they suggest only that most tests taken individually may help determine which player has the best physical abilities for eligibility.

This study’s second objective was to validate the discriminant and predictive capacity of the combined physical tests used at the Team Québec selection camps. The regression analysis was computed to determine the best combination of tests to explain selection at each position. The most successful prediction models were those of linemen (DL and OL). To be selected on the final roster, DLs were expected to be tall and heavy, but to be fast on a straight line (40Y) and be able to change directions quickly and effectively (20S,3C). These elements can be linked to the role of defensive lineman, which is to physically oppose offensive linemen and chase quarterbacks and/or ball carriers in the offensive backfield. On the opposite end, the OL model included weight, height and power output. In the present case, it seems that selected offensive linemen were expected to be big and capable of generating power to move their defensive opponents (7).

Regarding skill positions, DBs and WRs had remarkably similar regression models. Both were likely to be big players able to demonstrate speed and acceleration, but the receiver’s ability to change direction was more important. More specifically for defensive backs, the explosive power of the lower limbs was a more important factor for selection. A receiver’s main role is to be a target for the quarterback and catch the ball. This implies the ability to run different routes with speed and the agility to get open. DBs, on the other hand, often must compensate for taller receivers by being able to challenge a catch to the highest point possible. One might assume that the coaches’ criteria for selecting one position would be to counter the other one given they are competing. The same can be said for RB and LB. Both positions had essentially the same models, which included criteria such as weight, speed, and agility. The main difference was that lower limb power was expressed by the broad jump for LB and the vertical jump for RB.

Notwithstanding its contribution to scientific knowledge in the field of talent selection, this research has certain limitations. As McGee and Burkett (16) indicate, selection camp physical performance measures physical capabilities only, and many other factors are important to a football player. Since the explained variance for all models was modest, oscillating between 17% and 22%, We can assume that selection is based on additional factors that are not considered in Football Québec’s selection process. The physical results obtained at the selection camp reflect only a small portion of the players’ skills (8). Combine tests are conducted with no physical contact and with minimal decision-making opportunities, the two major components of a real football game (11). Considering additional factors would certainly help improve the protocol’s predictive validity. Other psychosocial attributes such as determination, toughness, self-confidence, coping resources and teamwork would provide a better estimation of a player’s potential (16).  Andrew, Potgieter and Grobbelaar (1), for example, found that top rank players in rugby were better at coping with adversity and had more confidence and achievement motivation than lower rank players. It is reasonable to believe such factors also play a role in a related sport like football. The authors conclude that psychological skills are clearly related to team selection, and more emphasis should consequently be placed on developing these qualities (1). Finally, previous on-field performance has been shown to be a good predictor of success at the professional level (13) and could be another element in helping coaches select their players.

Although the physical trainer and coaches were qualified by Football Québec to oversee testing, the authors cannot comment on the rigor of the data mining or provide test-retest reliability correlations measures for the current data. This study was observational, and the researchers did not examine past protocols. Future research in a football selection setting should include procedures of primary data collection, which would combine the physical testing expertise of physical trainers, the sports knowledge of coaches and an experimental design by researchers. Such protocols would allow for better control over potential experimental bias. The data obtained would be more valid and reliable and provide an improved picture of on-field performance. Upcoming studies on football selection should account for specific on-field abilities per position, mental skills, and coaches’ perceptions.

CONCLUSION

This article reinforces the importance of physical testing in the selection of football athletes. Indeed, players selected in Team Quebec have accumulated better test results than those not selected. Moreover, one of the strengths of this study is to establish the predictive power of the physical test battery on the selection of players according to their position. Certain tests such as the 40-yard dash or the Pro Agility Shuttle test proved to be excellent predictors of selection for most positions. Conversely, the two tests performed with the HI Trainer do not seem to be good predictors of selection, except perhaps for offensive linemen.

APPLICATIONS IN SPORT

We hope this article will help federation officials and coaches target physical tests offering better predictability for the selection of an elite team, thereby facilitating the evaluation process by coaches, who will be able to focus on the test(s) most useful for selecting elite club players. Players aspiring to be selected can also adapt their preparation based on tests deemed statistically relevant during the selection camps. Overall, this research could contribute to Football Québec’s elite program and that of similar football institutions across North America by offering a scientific basis for an improved process of selecting junior elite players. An indirect result may be to benefit the on-field performances of present teams as well. Finally, the study may also help improve selection practices for competitions at the pre-university, university and professional levels.

ACKNOWLEDGEMENTS

The authors wish to thank Football Québec for providing the raw data that made this study possible.

REFERENCES

  1. Andrew, M., Potgieter J.C. & Grobbelaar H.W. (2007). Sport psychological skill levels and related psychosocial factors that distinguish between rugby union players of different participation levels. South African Journal for Research in Sport, Physical Education and Recreation, 29(1), 1-14.
  2. Baechle, T.R. & Earle R.W. (2008). Essentials of strength training and conditioning. Human kinetics.
  3. Buckthorpe, M., Morris J., Folland J.P. (2012). Validity of vertical jump measurement devices. Journal of Sports Sciences, 30(1), 63-69.
  4. Dobbs, C.W., Gill N.D., Smart D.J., McGuigan M.R. (2015). Relationship Between Vertical and Horizontal Jump Variables and Muscular Performance in Athletes. The Journal of Strength & Conditioning Research, 29(3), 661-671.
  5. Fitzgerald, C.F., Jensen R.L. (2018) A Comparison of the National Football League’s Annual National Football League Combine 1999-2000 to 2015-2016 The Journal of Strength & Conditioning Research, 34(3), 771-781.
  6. Football Canada. (2019). Football Canada Cup. http://footballcanada.com/championships/football-canada-cup/.
  7. Football Québec. (2017). Modèle du développement à long terme de l’athlète en football au Québec. http://footballquebec.com/wpcontent/uploads/2017/06/MDLTA2017.pdf.
  8. Ghigiarelli, J.J. (2011). Combine Performance Descriptors and Predictors of Recruit Ranking for the Top High School Football Recruits from 2001 to 2009: Differences between Position Groups. The Journal of Strength & Conditioning Research, 25(5), 1193-1203.
  9. Gillen, Z.M., Shoemaker M.E., McKay B.D. & Cramer J.T. Performance Differences between National Football League and High School American Football Combine Participants. Research quarterly for exercise and sport, 90(2), 227-233.
  10. Johnston, K., Wattie N., Schorer J. & Baker J. Talent Identification in Sport: A Systematic Review. Sports Medicine, 48(1), 97-109.
  11. Kuzmits, F.E. & Adams A.J. (2008) The NFL Combine: Does It Predict Performance in the National Football League? The Journal of Strength & Conditioning Research 22(6), 1721-1727.
  12. Langley, J.G. & Chetlin R.D. (2017) Test Re-Test Reliability of Four Versions of the 3-Cone Test in Non-Athletic Men. Journal of sports science & medicine, 16(1), 44-52.
  13. Lyons, B.D., Hoffman B.J., Michel J.W., Williams K.J. (2011) On the Predictive Efficiency of Past Performance and Physical Ability: The Case of the National Football League. Human Performance, 24(2), 158-172.
  14. Mann, J.B., Ivey P.J., Brechue W.F. & Mayhew J.L.(2015) Validity and reliability of hand and electronic timing for 40-yd sprint in college football players. The Journal of Strength & Conditioning Research, 29(6), 1509-1514.
  15. Mayhew, J. L., Houser, J. J., Briney, B. B., Williams, T. B., Piper, F. C., & Brechue, W. F. (2010). Comparison between hand and electronic timing of 40-yd dash performance in college football players. The Journal of Strength & Conditioning Research, 24(2), 447-451.
  16. McGee, K.J., Burkett L.N. (2003) The National Football League combine: a reliable predictor of draft status? The Journal of Strength & Conditioning Research, 17(1), 6-11.
  17. McKie, G.L., Islam H., Townsend L.K., Howe G.J. & Hazell T.J. (2018). Establishing a practical treadmill sprint as an alternative to the Wingate anaerobic test. Measurement in Physical Education and Exercise Science, 22(2), 138-144.
  18. Meylan, C.M.P., Nosaka K., Green J., Cronin J.B. Temporal and kinetic analysis of unilateral jumping in the vertical, horizontal, and lateral directions. Journal of sports sciences, 28(5), 545-554.
  19. Moore, M. Scouting the Elite, Unbreakable Football Player. The Physician and sportsmedicine, 9(2), 131-135.
  20. Réseau du sport étudiant du Québec. (2017) Rapport annuel 16/17. https://www.rseqmontreal.com/fichiersUpload/fichiers/20180928114826-rapport-annuel-2017.pdf
  21. Robbins, D.W. (2010). The National Football League (NFL) combine: Does normalized data better predict performance in the NFL draft? The Journal of Strength & Conditioning Research, 24(11), 2888-2899.
  22. Sierer, S.P., Battaglini C.L., Mihalik J.P., Shields E.W., Tomasini N.T. (2008). The National Football League Combine: performance differences between drafted and nondrafted players entering the 2004 and 2005 drafts. The Journal of Strength & Conditioning Research, 22(1), 6-12.
  23. Stewart, P.F., Turner A.N. & Miller S.C. (2014). Reliability, factorial validity, and interrelationships of five commonly used change of direction speed tests. Scandinavian journal of medicine & science in sports, 24(3), 500-506.
  24. U-Sports. (2019). Vanier Cup History. https://presto-en.usports.ca/sports/fball/history/vanier_cup_history.
  25. Young, W., Macdonald C., Heggen T., Fitzpatrick J. (1997). An evaluation of the specificity, validity and reliability of jumping tests. The Journal of sports medicine and physical fitness, 37(4), 240-245.

Print Friendly, PDF & Email