In the sport of ice hockey, skating is an essential skill that requires a combination of speed, strength, power, and balance. Developing each player's skating ability is an integral component of a hockey team's practice and training programs. Expense and a lack of available ice time limits sufficient on-ice training for many teams. Consequently, on-ice training is frequently supplemented with off-ice training and conditioning before and during the hockey season. To use off-ice performance measures for evaluative purposes or to aid in designing efficient and effective off-ice training programs, it is important that sports scientists and coaches understand the relationship of specific off-ice performance to on-ice skating performance.
Previous researchers have reported an association of several off-ice performance measures with on-ice performance. Off-ice sprint time is reported to be a predictor of both forward skating speed and game performance (1,4,8,12). Mascaro et al. (10) found vertical jump as the best predictor of skating speed. Lastly, balance ability is also reported as a predictor of skating speed (1). Although the relationship of these off-ice tests to forward skating speed is well reported, the relationship of off-ice tests to on-ice crossover and turning performance has not been studied in depth. This relationship is important as crossovers and glide turns are performed frequently during hockey games. Examining 12 National Hockey League (NHL) forwards, Bracko et al. (3) analyzed the percentage of time the forwards spent in various skating patterns. He reported that during games, players spent 72.1% of their on-ice time performing crossovers or glide turns. To evaluate these specific skating skills, the relationship of off-ice measures to a cornering S test has been studied (4,8). The cornering S test includes crossovers to both the right and left within a single trial. We are not aware of any studies examining crossover skating performance and turning in a single direction. Thus, in addition to forward skating, we sought to examine the performance of players skating designated courses consisting of turns and crossovers exclusively to the right or to the left and compare performance on these skating tests with a series of select off-ice performance measures.
The purpose of this study was to examine the relationship of off-ice performance measures with on-ice turning, crossover, and forward skating performance in high school male hockey players. Furthermore, we sought to evaluate if off-ice measures could predict right and left turning and crossover performance in addition to forward on-ice sprinting. We hypothesized that off-ice sprint speed, jumping ability, and balance would be associated with and predict all on-ice performance measures and secondly, that side-to-side difference in off-ice jumping and balance tests would be associated with and predict directional on-ice performance.
Experimental Approach to the Problem
The study was designed to investigate the relationship of off-ice performance measures (independent variables) with on-ice performance measures (dependent variables) in male high school hockey players. We used a within-subjects, repeated measures design to test the null hypothesis that off-ice performance measures would not be associated with on-ice performance. All the measures were analyzed to find correlation and predictive values. Off-ice performance measures were specifically selected to represent and assess dynamic balance, explosive power, and speed. The 3 on-ice tests were designed to evaluate right and left turning performance and forward skate speed as measures of skating ability. Data collection from each subject occurred on a single day. Testing was conducted approximately 2 weeks after the completion of the high school hockey season. All the subjects were members of a high school team and had participated in off-ice training in addition to on-ice practices throughout the season. The subjects completed off-ice performance measures followed by on-ice testing with time allotted for warm-up before both the off-ice and on-ice testing. The subjects were instructed to prepare for the testing session as they would a typical hockey practice session. Verbal encouragement was provided by examiners during the testing procedures. Specifically, the study design attempts to address the practical question of how off-ice performance measures can predict performance of on-ice forward skating, turning, and crossover performance.
Forty male high school hockey players (mean ± SD age = 16.3 ± 1.1 years; height = 177.9 ± 6.8 cm; weight = 72.5 ± 8.9 kg; and years of playing high school hockey = 1.9 ± 1.1 years) were recruited from local high schools in Rochester, MN. A sample size of 40 participants provides 73% power to detect correlation coefficients at a magnitude of 0.40 that are significantly different from a correlation coefficient of 0, or 92% power to detect correlation coefficients of 0.50 that are significantly different from 0. Forty participants provides approximately 75% power to detect an R2 value of 0.25 or greater (effect size f2 = 0.33) for testing 5 predictor variables' combined contributions to variance in on-ice skating performance at α = 0.05.
The subjects were competitive high school hockey players who were members of their school's varsity or junior varsity teams during the current high school hockey season. Participant skill level ranged from those who may not play beyond the junior varsity level to that of division 1 college recruits. The subjects were free of lower extremity conditions that would limit their performance on test procedures. Exclusionary criteria included any past surgery for knee or ankle instability. The subjects were given a written description of study procedures and potential risks. Study procedures were approved by the Mayo Foundation Institutional Review Board, Mayo Clinic, Rochester, MN. All testing procedures and risks were explained, and the participants and their parent or legal guardian provided written consent before any testing.
All off-ice tests were conducted before on-ice testing. The subjects wore running shoes, athletic shorts, and shirts for off-ice testing. Upon arrival at the ice arena, the subjects completed an intake questionnaire on age, handedness, shooting preference (right vs. left), leg dominance (leg used to kick a ball), directional preference for on-ice crossovers and turning, and number of years playing high school hockey. Height and weight were measured. Leg length was measured from the most prominent aspect of the anterior superior iliac spine to the most prominent aspect of the medial malleolus. With the exception of balance testing, the order of off-ice single-leg performance measures was randomized. Tests were completed by a single examiner at each testing station. The subjects were allowed time for warm-up and practice attempts before performing the actual tests. For each specific jump test, the subjects performed 3 trials with the best attempt recorded.
Off-ice testing included a series of horizontal hops and vertical jumps. The testing surface was a firm, rubberized floor. Horizontal hop tests included a double limb forward hop (broad jump), a horizontal single limb forward hop on the right, a single limb forward hop on left, a right limb to left limb lateral hop, and a left limb to right limb lateral hop. The subjects completed the double limb horizontal hop first, followed by all single limb hops consecutively on each leg. For forward horizontal hops, the subjects started with their toes behind a start line and jump distance was measured from the start line to the posterior aspect of the most posterior aspect of the heel. A standard metal tape measure secured to the floor was used to record jump distance. For the double limb horizontal hop, the subjects stood with both feet behind the starting line and jumped forward as far as possible. Arm swing was allowed. To qualify as a successful attempt, the subjects had to maintain balance for 2 seconds upon landing. For single limb horizontal hop tests, the subjects stood on a single limb and jumped forward landing on the same limb. Again, the subjects were required to maintain balance for 2 seconds upon landing. Jumps were repeated if the opposite foot touched the ground or if balance was lost. Bolgla and Keskula (2) reported an intraclass correlation coefficient (ICC) of 0.96 and a standard error of measurement (SEM) of 4.56 cm for the single limb horizontal hop test. To perform lateral hop tests, the subjects stood on 1 limb with the medial border of their foot behind the start line. The subjects jumped laterally as far as possible, landing on the opposite limb maintaining balance for a minimum of 2 seconds. Hop distance was measured from the starting line to the medial aspect of the landing foot. The jump was repeated if the subject touched the ground with the nonlanding foot, or if the landing foot moved upon landing. Each subject also performed a series of 3 vertical jumps including a double limb vertical jump and a single limb vertical jump on each leg. The double limb jump was completed first, followed by single limb jumps. Jump performance was recorded via a Vertec (Sports Imports, Columbus, OH, USA) jump measurement system. To set target vanes, the lowest vane on the Vertec was set to the maximal reach of the subject while standing with both feet on the ground. The subject then performed a maximum effort double limb vertical jump from a stationary position. An approach step was not allowed. Arm swing was not restricted. For the single limb vertical jumps, a similar procedure was used. The subjects started from a single limb stance but were allowed to land on both feet. Both double and single limb vertical jump performance using the Vertec measurement device are considered reliable (5,6). Burr (5) reported an ICC of 0.98 for the double limb vertical jump in male NHL players. McElveen et al. (11) reported an ICC range of 0.93–0.97 for the single limb vertical jump in physically active young adults.
Dynamic balance testing used the modified star excursion balance test using the Y Balance Test device (Functional Movement Systems, Inc., Chatham, VA, USA). Before study participation, the subjects watched a video demonstrating the test procedures. After verbal instructions, the subjects performed 3 practice trials on each limb for each test maneuver. The testing sequence used was based on previous research (9,13). Per protocol, the subject placed the stance foot on the stationary center platform. The other limb moved a reach indicator in 1 of 3 designated directions. To successfully complete each test maneuver or “reach,” the subject was required to return his limb to the start position next to the stance limb. The test sequence alternated from left limb tests to right limb test for motions identified as anterior reach, posteromedial reach, and posterolateral reach. Direction of reach was referenced to the stance limb. A trial was rejected if the subject contacted the ground with the “reaching” limb, if the reached indicator was flicked to gain additional distance, or if the reach indicator was used for support. The examiner recorded the best of 3 trials for each individual test from the Y balance device scale. Reach distance was normalized to leg length. A composite score of the total distance of the 3 directions for each leg was also recorded. Plisky (13) reported intrarater reliability using the Y Balance device ranged from an ICC of 0.85 to 0.90 (SEM 2.01–5.84 cm), whereas the interrater reliability ranged from an ICC of 0.99 to 1.00 (SEM 0.68–3.3l cm).
Forty-yard sprint testing was completed on an indoor track with lines designating the start and finish points. Before testing, the subjects were allowed time to jog and sprint to warm-up. When the subject reported he was ready for testing, the subject positioned himself behind the start line. Three investigators were positioned at the finish with hand-held timers (Survivor III, Accusplit, Livermore, CA, USA). Timing began upon initiation of forward movement by the subject at the start line and stopped when the subject crossed the finish line. The subjects completed a single attempt. The average of the 3 investigators' displayed times was recorded.
The subjects wore full equipment and carried hockey sticks for on-ice testing. The subjects were allowed approximately 10 minutes on ice to warm-up. The order of right or left turning and crossover tests was randomized. All the tests were performed once. A trial was repeated if a subject fell or lost balance during the test. As with off-ice sprint testing, an average time from 3 timers was recorded. The investigators started timing at the subject's first forward movement at the start line and stopped when the subject crossed the finish line. To minimize inconsistency in ice surface, a maximum of 10 skaters were tested with the ice resurfaced before a subsequent test group. For the forward skating sprint, the subjects started behind the goal line and skated through the far blue line, covering a distance of 34.5 m (Figure 1). The investigators were positioned at the blue line for timing.
For the short radius turn course, a marker cone was placed on each blue line an equal distance from the side boards. Starting behind the center red line, the subject skated toward and made a sharp turn around the first cone, then skated toward and made a sharp turn around the second cone, and finished by crossing the center line (Figure 2). The course was exclusively left or right turns. The subjects performed a single trial in each direction. The investigators were positioned at the center red line for timing. Cones were moved between the subject's tests to maintain a consistent ice surface for all the subjects.
Four marker cones were placed on each end-zone face off circle equidistant from one another to mark the crossover turn course (Figure 3). The subjects started this skating test positioned behind the blue line near the side boards. The subjects skated a designated course around both face off circles and finished by crossing the blue line near its center point. The course required either all right or all left crossover turns. Each subject skated 1 trial in each direction. The investigators were positioned at the blue line. Both ends of the ice were used, so not >5 skaters performed the test on a single course.
The dependent variables analyzed were the on-ice skating tests, which included forward skate time, right crossover time, left crossover time, right short radius time, and left short radius time. Seventeen independent variables analyzed included 5 hop tests (double limb hop, single limb anterior hop on the right, single limb anterior hop on the left, and right to left limb lateral hop, left to right limb lateral hop), 3 vertical jump tests (double limb jump, single limb jump on the right, and single limb jump on the left), 8 dynamic balance tests (right anterior reach, left anterior reach, right posteromedial reach, left posteromedial reach, right posterolateral reach, left posterolateral reach, and a composite score for the 3 right and 3 left balance tests), and a 40-yd sprint.
Descriptive and inferential statistics were analyzed with SPSS 15.0 software (SPSS Inc., Chicago, IL, USA). Paired t-tests were used to examine differences between the right and left on-ice turning performance. Pearson correlation coefficients were used to examine the relationship of off-ice to on-ice measurements. We used a principal components factor analysis with varimax rotation to examine the number of factors with eigen values exceeding 1.0 that contributed to variance in the on-ice and off-ice performance measures. The factor analysis was used to reduce the number of observed variables to a fewer number of representative unobserved variables by identifying patterns of related observed variables. Parameters with correlation loadings of ≥0.50 on any given factor were interpreted as being significant contributors to the factor. Once the factor analysis was completed, we then further analyzed data with hierarchical multiple regression to examine the predictive relationship of multiple off-ice performance variables to the on-ice performance measures. Interpretation of the factor analysis dictated our selection of predictor variables to enter into the regression analyses. Separate regression analyses were conducted for each on-ice performance measure. Predictor variables entered into the hierarchical regressions included (a) sprint time, (b) double limb anterior hop distance, (c) right and left posterolateral balance scores, (d) double limb vertical jump height, and (e) right and left anterior balance scores. All inferential statistics were conducted with α = 0.05 to limit the potential for type 1 error.
Forty subjects were recruited for this study. Thirty-eight subjects completed testing and were included in the final analysis. One subject withdrew during testing and a second did not have the required equipment to complete on-ice testing.
For on-ice tests, no significant differences were found between right and left short radius turning performances or right and left crossover performances, even when considering shooting handedness, leg dominance, and turning preference. Pearson correlations are presented in Table 1. Five off-ice variables correlated significantly with all on-ice performance measures. These variables included sprint time, lateral bound right to left, double limb horizontal hop, balance on right in posterolateral direction, and the composite balance performance on the right. Results of the factor analysis are presented in Table 2. Five factors were identified. Factor 1 was identified as posterior balance measures, factor 2 timed performance measures, factor 3 horizontal hop measures, factor 4 vertical jump measures, and factor 5 anterior balance measures. For the hierarchical regression analysis, sprint time represented the factor timed performance measures, double limb forward hop represented horizontal hop measures, right and left posterolateral dynamic balance measures represented the factor of posterior balance measures, double limb vertical jump represented vertical jump measures, and lastly, right and left anterior balance measures represented the anterior balance factor. Using these variables in the hierarchical regression, off-ice sprint time was the only significant predictor of on-ice skating performance measures, accounting for 65.4% of the variability in forward skate time (F1,36 = 68.061, p < 0.001), 30.8% of the variability in right crossover time (F1,36 = 15.991, p < 0.001), 37.9% of the variability in left crossover time (F1,36 = 21.989, p < 0.001), 21.8% of the variability in right short radius time (F1,36 = 10.021, p = 0.003), and 45.0% of the variability in left short radius time (F1,36 = 29.479, p < 0.001). The contribution of all 5 factors in explaining the variability in on-ice performance along with regression equations is presented in Table 3.
Although our data support the hypothesis that measures of off-ice sprint speed, jumping, and balance ability would be associated with skating performance, only off-ice sprint time predicted on-ice performance measures. Our hypothesis that jumping and balance measures would predict on-ice performance was not supported. Likewise, we were unable to support our hypothesis that side-to-side difference in off-ice jumping and balance tests would predict directional on-ice performance
Because the associations of off-ice measures vary across specific on-ice measures, it is necessary to examine individually the tasks of forward skating, crossover skating, and short radius skating. In assessing forward skate time, the strongest association was with sprint time (r = 0.81, p < 0.01). All jumping tests were moderately correlated with forward skating with r values ranging from −0.40 to −0.58. The negative correlation reflects greater jump height is associated with faster skating times. Of all jumping tasks, the single-leg vertical jump tests had the 2 strongest correlations with r = −0.55 (p < 0.01) for the right limb jump, and r = −0.58 (p < 0.01) for the left limb. The correlation for double limb vertical jump was similar with r = −0.51 (p < 0.01). We included single limb right to left and left to right lateral hops to reflect the lateral component of the hockey stride. The correlations of the lateral jumps with forward skating time had r values of −0.454 (p < 0.01) for the right to left hop, and −0.496 (p < 0.01) for the left to right. Of the 6 unilateral Y balance tests, only the right posterolateral balance (r = −0.565, p < 0.01) and left posterolateral balance (r = −0.356, p = 0.03) tests were significantly associated with forward skating time (r = 0.565, p < 0.01). Both were also associated with crossovers and short radius turns to the direction opposite of the stance leg. Although this maneuver is similar to a crossover stride of the “inside” leg, it does not replicate the stride used in forward skating, The posteromedial balance test, which would be similar to the forward skating stride, was not significantly associated with forward skating performance. Behm (1) also reported a mixed association of balance with skating. He reported a wobble board balance test significantly correlated with skating speed in junior level hockey players ages 16–25. Further analysis however showed that balance correlated with skating speed only in players under 19. Although comparison of ages between studies is possible, comparison of skill levels between studies is not.
Although our correlation analysis revealed multiple associations between forward skating performance and the off-ice variables, our regression found the only significant off-ice predictor variable for forward skating was sprint time. The other 4 factors representing hopping, jumping, and balance did not significantly account for variability in skating tasks. Previous investigators have reported otherwise. Farlinger et al. (8) reported that the 30-m sprint time and the 3 hop jump distance were the greatest predictors of 35-m on-ice skating performance in male hockey players age 15–22 with the 30-m sprint accounting for 46% of the variability in skating performance. Vertical jump is also reported to predict skating performance (7,10). Mascaro et al. (10) conducted off-ice tests that consisted of a 40-yd dash, standing long jump, vertical jump, and isokinetic quadriceps and hamstring testing, and compared their performance with a 54.9-m skating sprint in professional hockey players. Other than isokinetic testing, the best predictor of skating speed was the vertical jump test. Although we found a significant correlation between double limb hop and vertical jump with forward skating speed, these variables did not predict forward skating performance above and beyond that accounted for by sprinting speed. Differences in the abilities of players between the studies may partially explain the inconsistency between studies.
Unique to this investigation was the examination of single direction turning and cornering. Most studies of on-ice turns or cornering use skating courses that require turns in both directions (4,8). We used single direction turning and cornering within a single trial to examine our hypothesis that side-to-side differences in unilateral off-ice tests would be associated with differences in directional on-ice performance. We were unable to support this hypothesis because no significant difference in directional on-ice performances was detected. This finding might be explained by the experience level of our study group. To illustrate, when specific skating characteristics of 12 professional ice hockey forwards were examined, total occurrences for left crossovers was 20.2 and 17.7% for right crossovers (3). The frequency of left gliding turns was 17.8 and 16.4% for gliding turns to the right (3). Of 12 skating characteristics identified and observed, crossovers and gliding turns had the greatest frequency among all the characteristics. Although the professional game is different than the high school game in many respects, the frequency of performing crossovers and gliding turns is likely similar. Thus, our high school hockey players may have developed proficiency in crossover and turning ability in both directions because of the frequency of performing these skills in games and practices over their years of playing hockey.
As with forward skating, off-ice sprint time predicted crossover performance. Thirty-one percent of the variance in right crossover time and 38% of the variance in left crossover time is explained by off-ice sprint time. Given sprint time accounted for 65% of the variance seen in forward skate time, the inclusion of forward skating segments in the crossover course may contribute to this finding. Consistent with this possibility, Farlinger et al. (8) suggested that the cornering S test used in their study did not differ enough from straight sprint skating to measure agility.
The last on-ice measures examined were right and left short radius turn courses. Off-ice tests that correlated the greatest with both right and left short radius tests were sprint time and double limb hop. The only other off-ice measure significantly correlated with both directions was right limb posterolateral balance on the Y balance test. Off-ice sprint time predicted 22% of the variability in the short radius right turn course and 45% of the variability in the left turn course. As with the crossover course, there was a considerable and necessary forward skating component to this course which may factor into our results.
Regression equations (Table 3) can be used to predict skating performance. For example, skating sprint speed may be estimated using the 40-yd sprint time. The regression equation for the 34.5-m skating sprint is ŷ = 1.916 + 0.602 (sprint time). Thus, for every 1-second difference in the 40-yd sprint time, there will be approximately a 0.6-second difference in the 34.5-m on-ice sprint time. Based on the regression equation, a player who ran the 40-yd dash in 4.8 seconds would project to have a time of 4.81 seconds on the 34.5-m forward skate test, whereas a time of 5.8 seconds in the 40-yd sprint would have project to a time of 5.41 seconds for the on-ice sprint. Similar estimates can be calculated for other skating tasks based on the 40-yd sprint time entered into the regression equation for each specific skating task.
Including only male skaters limits the ability to generalize findings to female skaters. Similar findings, however, have been reported in female hockey players. Bracko et al. (4) tested female players (ages 8–16). On-ice tests including a 6.10-m acceleration and 47.85-m speed test and a cornering S test were compared with off-ice tests. Off-ice tests included a 40-yd sprint, vertical jump, and timed sit-ups and push-ups. Consistent with our results, the 40-yd sprint was the strongest predictor of skating speed.
There are additional limitations to this study. All the participants were male high school hockey players, which resulted in narrow range of data and limits generalizablility. A wider range of skill levels may provide greater insight into predictor variables. It should also be noted that physiologic measures are only 1 component of a successful hockey player. Lastly, as the design of this study was not longitudinal, we are unable to determine if changes in off-ice measures over time would result in on-ice performance changes.
Off-ice training is an essential component of conditioning and preseason preparation for many hockey teams. Off-ice tests may also be used as metrics to evaluate players. Based on measures specifically used in our study, the 40-yd sprint is the best predictor of skating performance. Based on the regression equation, for every 1-second difference in the 40-yd sprint time, there will be approximately a 0.6-second difference in the 34.5-m on-ice sprint. The 40-yd sprint predicts forward skating performance to a greater degree than it does crossover or tuning performance. Although our findings suggest that at the high school level, faster runners are typically faster skaters, we cannot say that improving sprint time will result in a faster skater. Longitudinal repeated measures studies are needed to examine if improvements in off-ice sprinting result in improved speed on ice.
This project was funded by the Mayo Clinic Orthopedic Research Review Committee, Mayo Clinic, Rochester, MN. The authors would like to thank Carol Best, Dan Gaz, and Justin Deeg for their assistance. They do not have any professional affiliation with any manufacturers of equipment used in this study. The results of this study do not constitute endorsement of the products by the authors or the National Strength and Conditioning Association.