I caught the discussion about off- and on-ice speed testing on the podcast. Great stuff. A couple quick things I've seen from a numbers perspective.
-In the last full cycle we were with the women's team (Aug 15-Worlds '16), the correlation between VJ height and on-ice sprint time was -.59. It was -.49 between VJ and the on-ice agility test, and .62 between the on-ice sprint and agility tests. All of these are "moderate" correlations.
-A couple years ago, I looked at off-ice 10-yard dash times in June (the only camp we tested this) and then on-ice sprint times in March (the camp where times should be the fastest), and the correlation was around .4. This is quite a bit lower, in large part because we never tested those two things at the same camp.
-In Devan's group, the correlation between the 10-yard sprint on land and the on-ice sprint was 0.45, and between VJ and on-ice sprint -.49.
-At our development camp, the correlation between 10-yard sprint and goal to blue was .33, between VJ and goal to blue -.33.
All of these tests were done at different times of the annual cycle, which will influence the strength of the relationship. Our development camp was the first week in July so some players had not skated much at all heading into the camp. Devan's group may have the cleanest data set since they've been sprinting regularly throughout the season. They're also done with different populations, but the numbers are pretty consistent with published research in this area and generally tell a similar story.
There is a relationship between off- and on-ice tests, but it's not as strong and direct as we may think. To understand the discrepancy, I think it's helpful to take a step back and think about what qualities would contribute to someone being a fast runner AND a fast skater, and then try to identify what qualities may contribute more to one than the other. The image below is from a quick presentation I gave to the USA coaching staff, and just provides a superficial illustration of similarities and differences in qualities contributing to skating and running speed.
In short, skating is skill, but so is running and so is jumping, and the movement requirements are different enough that we'd expect to see some differences.
I also think we may see stronger relationships between a longer off-ice sprint test and the goal to blue. Even though the ground contact times of a 10-yard sprint are more similar to skating than sprints of longer than 10 yards (see below), a bad start in a 10-yard test that lasts less than 2.0 seconds doesn't provide a lot of time for the athlete to "make up" for it. And since most people are only doing 2-3 testing reps, the variability in someone's performance could cloud the relationship.
Ground contact times and stride rates during different types of skating strides.
Ground contact times and stride rates during different phases of acceleration.
You were right to point out that comparing rankings without looking at the actual spread of the numbers can be misleading. With Devan's group, I used magnitude based inferences to categorize players into groups based on how far their scores were away from the group average. In the image below, small differences above and below the mean were all labeled “average”. Moderate and large differences above and below the mean were clustered together to limit group numbers and labeled as “above average” and “below average”, respectively, and very large differences above and below the mean were labeled as “great” and “bad”. This helps bucket players relative to their actual score, instead of just ranking them. Several of the players fell in the same category across the two sprint tests, but a lot didn't. In these cases, I just posed basic questions to Devan that may explain why they didn't.
In most cases, the questions were right (e.g. if a player had a better on-ice than off-ice sprint score it's because they were either a great skater or a horrific runner). There were really only a couple players that didn't fit what we'd expect.