How to Test Bobsled Push-Start Power Without a Push Track: A 15m/30m Weighted Sled Protocol

You've got two turf lanes behind the weight room and a sled with plates bolted on it. What you don't have is 40 meters of banked ice at a two-degree pitch, so pushing that sled and clocking a 30-meter split has become the go-to way to test push-start power in the off-season. The number that comes back looks like data — a time to the hundredth of a second — but it's really measuring how hard someone can move a specific mass across a patch of turf, at a friction coefficient nowhere near what steel runners see on ice. A push athlete's turf time can drop six-tenths across a block while their real push-start capacity barely moves, or the reverse happens: flat sled numbers hide a genuine strength gain the ice would reward and the rubber track never will.

This guide covers a field protocol built around that mismatch: instead of trusting one loaded sled time as a stand-in for ice performance, run the sled at three or four loads, clock splits at 15m and 30m for each, and fit a load-velocity profile across them. The resulting numbers — percent velocity decrement at a given load, and the slope connecting load to that decrement — describe how an athlete produces force in a way far less tangled up with whatever friction sits under the sled that day, which is the transferable signal a program without ice access can actually work with.

That gap between turf and ice is exactly why USA Bobsled and Skeleton's own combine testing has never tried to replicate ice conditions directly — it tests dry-land proxies instead and checks them against real outcomes. Tomasevicz, Ransone, and Bach (2020) analyzed 11 combine variables — sprints at 15m, 30m, and 60m, a flying 30m, a standing broad jump, a shot toss, a squat, a power clean, body mass, and two dry-land bobsled pushes — across 75 athletes spanning two Olympic qualification cycles (2009 and 2013). Their principal component analysis found a single component explaining 56.9% of the variance in the pool, built almost entirely from the four sprint tests plus the two dry-land pushes, not the strength lifts. That component separated Olympians from non-Olympians and National Team members from the rest.

The study's own limitation is worth sitting with: none of those variables happen on ice, and the analysis is correlational — it shows which proxies track real selection outcomes over two cycles, not that any one causes better on-ice pushing. That's the model worth borrowing. A proxy doesn't need to replicate ice friction to be useful; it needs a checkable relationship to the force and velocity an athlete actually produces, and a single sled time on turf doesn't give you that on its own — the profiling step below does.

Cahill, Oliver, Cronin, Clark, Cross, and Lloyd (2021) built the method this protocol borrows, applying it to sled pushing rather than sprinting or towing. Ninety male athletes (16.9 ± 0.9 years) ran one unresisted sprint and three loaded sled-push sprints of increasing resistance, with maximal velocity captured by radar on every run. Plotting load against velocity produced an individual load-velocity relationship for each athlete, and from it a percent velocity decrement, or %Vdec, at any given load relative to that athlete's own unresisted number.

The reason this beats a raw loaded time on turf comes down to what %Vdec cancels out. The loaded and unloaded runs happen on the same surface, in the same footwear, the same day — so the friction coefficient of that surface acts on both numbers in roughly the same direction and mostly cancels out of the ratio. It doesn't cancel perfectly, since sled friction isn't purely proportional to velocity, but a ratio built this way travels across surfaces far better than a raw 30-meter time, which is what makes it usable when ice simply isn't available.

The gear list here is short on purpose — this protocol is built for programs without a push track, not ones trying to build a lab.

What You Need

• A weighted push sled with a known frame mass (typically 15-25 kg empty) and plate posts in known increments, ideally 2.5-5 kg jumps.
• Dual-beam timing gates or a laser/radar unit at 0m, 15m, and 30m. A hip-mounted IMU or GPS wearable works as an alternative and removes gate-alignment error entirely.
• 35-40 meters of flat, unobstructed surface — turf, rubberized track, or a gym floor with sled runners all work, but the surface and sled must stay identical across a training block. Switching mid-block changes the friction coefficient enough to distort a trend line.
• A calibrated scale for same-day body mass, since every load here is set as a percentage of body mass and that number shifts week to week.

Run this across a single dedicated session, not squeezed onto the end of a lift. Full recovery between loaded efforts matters more here than in a conditioning circuit — a tired third rep just adds noise to the profile.

Session Steps

1. Warm up generally, then run 2 potentiation sled pushes at a light load to groove the start position and confirm the sled tracks straight.
2. Run the empty-sled condition (frame mass only) over 30m with a split at 15m. Two trials, keep the faster one, and record this as the baseline velocity rather than a true unresisted sprint, since the frame mass never fully disappears in most gyms.
3. Run three loaded conditions, targeting roughly 20-25%, 45-50%, and 70-75% of body mass as starting points. Two trials per load, best trial kept, minimum 3 minutes recovery between loads and 4-5 minutes before the heaviest one.
4. Record the 0-15m and 0-30m splits to the nearest 0.01s, then derive the 15-30m interval by subtraction. Keep the two intervals separate rather than averaging — a bobsled's own first timing eye sits at 15m, and averaging washes out exactly the split that matters most.
5. Discard any trial with a false start, a slipped foot, or a sled that drifted off line — a wandering sled adds resistance the load-velocity math doesn't account for.
6. Calculate %Vdec at each load, per interval: %Vdec = (1 − loaded velocity ÷ empty-sled velocity) × 100.

The 20-25%, 45-50%, and 70-75% starting loads above are just starting points. Cahill et al. (2021) found large between-athlete variation in the load actually needed to hit a given decrement, so treat the table below as a first pass to dial in over two pilot sessions before trusting the numbers.

Two caveats before setting targets off this table. The sample was 90 high-school-age male athletes, not trained push athletes, so the percentages are a starting range rather than a sport-specific standard — a national-level athlete may sit well outside these bands and still be normal for their own strength level. And this is a single-session reliability study; it validates the Vdec method itself, not a claim that a given %Vdec predicts push-start time on ice.

Grounding the interpretation in what these two studies actually found — and where they stop short — keeps a coaching cue from overreaching.

Cahill et al. (2021) is the direct source for the load ranges above, and its central finding is really about variation: the load needed to cause a 50% Vdec ranged from 45% to 85% of body mass across the confidence interval, wide enough that two athletes of similar bodyweight can need meaningfully different loads to train the same point on their curve. Its limitation is the population — high-school-age athletes in a general sprint context, not push specialists — so it confirms the method is reliable without confirming any particular %Vdec target transfers to faster on-ice pushing.

Tomasevicz, Ransone, and Bach (2020) found that dry-land sprint and push variables, not strength lifts, carried the signal separating Olympians and National Team athletes from the rest of the pool (PC1 explaining 56.9% of variance). That is a strong argument for prioritizing sprint- and push-based field testing over gym numbers when time is limited. Its limitation is that the analysis is retrospective and correlational across two qualification cycles for one federation combine, which does not establish that improving those variables causes better selection outcomes elsewhere.

The patterns below combine both studies directional findings with what shows up across repeated sled-push sessions — a practical field guide, not a validated diagnostic scale.

The full profile takes 25-30 minutes with warm-up and recovery built in — too long for every session.

• Every 2 weeks in-season: run it as its own short session rather than tacking it onto a lift, since a fatigued nervous system flattens the load-velocity relationship.
• After any sled or surface change: re-run the empty-sled baseline first, since a new frame mass or surface grip shifts the reference point the whole profile is built on.
• Monthly in the off-season: re-profile fully as body mass and strength shift — a load that produced a 50% Vdec in October can land closer to 40% by January with no change to the protocol itself.
• Before a competition block: taper loaded volume but keep one light-load session to confirm the 0-15m pattern hasn't drifted, since that interval maps most directly to a bobsled's own first timing checkpoint.

Key References

• Tomasevicz, C. L., Ransone, J. W., & Bach, C. W. (2020). Predicting Bobsled Pushing Ability From Various Combine Testing Events. Journal of Strength and Conditioning Research, 34(9), 2618-2626.
• Cahill, M. J., Oliver, J. L., Cronin, J. B., Clark, K. P., Cross, M. R., & Lloyd, R. S. (2021). Sled-Push Load-Velocity Profiling and Implications for Sprint Training Prescription in Young Athletes. Journal of Strength and Conditioning Research, 35(11), 3084-3089.
• Cross, M. R., Brughelli, M., Samozino, P., Brown, S. R., & Morin, J-B. (2017). Optimal Loading for Maximizing Power During Sled-Resisted Sprinting. International Journal of Sports Physiology and Performance, 12(8), 1069-1077.