Your station splits are training instructions

After a race, most athletes know roughly how it went. The SkiErg took something out of them, the wall balls landed on heavy legs. What they usually cannot answer is the question that would actually change their training: which station cost me the most, relative to athletes who finished where I finished? A four-minute sled push is fast or slow depending entirely on the company you keep; without a reference population, a raw split is a diary entry, not an instruction.

That reference population exists. The free HYROX Station Analyzer at athlete.tools takes your eight station splits and scores each one as a percentile against real race results in your format and division (the men’s open benchmark set alone holds over 176,000 results), shows the median and top-ten-percent time per station, and names your biggest opportunity outright. We do not make that tool and have no stake in it; it is simply the cleanest answer we have found to the which-station question.

Read the output the blunt way: the station sitting furthest below your cohort’s median usually deserves the next block of training time more than the one you are already good at. Improving a strength feels better; improving the outlier moves the finish time.

Knowing the station is step one. The training question is structure, not volume: "do more SkiErg" is not a plan. The decisions that move the number are the length of each work interval, the rest between efforts, how many rounds you can hold before form degrades, and above all whether the station arrives fresh or on legs that already ran. In the race it arrives exactly once, after a kilometre of running. Training it in isolation at full effort is a different stimulus from training it with prior load attached, and only one of those looks like race day.

This is a structure the app was built to hold. Put a short run or row in the block ahead of the target station so it never arrives fresh, then repeat the block in rounds with the rest programmed on purpose. A typical targeted session: 800 metres of running, three minutes hard on the SkiErg, two minutes of rest, four rounds. The builder prices the session while you assemble it, and voice cues announce every transition, so the phone stays on the floor where it belongs.

Then close the loop with what the session recorded: per-round splits, heart rate across each round, time in zone. If round one is clearly faster than round four, the work interval is too long or the rest too short for the pace you chose. Reading session analytics walks through exactly how to make that call.

The loop, compressed: race data tells you what to target, structure decides how to train it, and session data tells you whether it is working. Run it between races and the next set of splits stops being a surprise.

Related reading: why your race was slower than your sim, on the one part of race day no gym session can rehearse.