← Back to the blog

How much should my rider’s power vary in a hilly time trial?

PacingUpdated 6 min readBorja Alfaraz

2026 Tour time trial

37 s

What the demo plan would have lost riding at constant power with the same average power (401 W).

Calculated with the SuperRecord engine on the Évian–Thonon time trial (26 km, 488 m of climbing). Almost all of the time is lost on the climb.

That a variable power split beats constant power on a hilly course has been studied for almost thirty years1,2. The coach’s question is a different one: how much should power vary, will my rider tolerate it, and how do I brief it so they can execute it with a number on their back. This article answers that with a real case, the 2026 Tour time trial in the SuperRecord demo.

What decision has to be made?

Three numbers by section, not one for the whole event: power on the climbs, power on the fast sections (descents and tailwind) and power everywhere else. The practical question is how far they move away from the average. Too little and you leave time behind; too much and the rider reaches the fast part with empty legs, or cannot execute the plan.

The case: the Évian–Thonon time trial at the 2026 Tour

The demo reproduces the 26 km time trial with the rider’s CdA, weight and the weather on the day (a 13 km/h north-north-easterly at 15:00). The SuperRecord plan for that 32:19 looks like this:

  • Climb, km 0 to 11: 428 W on average, peaking at 441 W on the 5 % ramps.
  • Descent, km 11 to 17: 329 W. Above 80 km/h, pushing harder buys almost no time.
  • Flat finish, km 17 to 26: 387 W, a little more on the short rises and less on the gentle downhill drags.
Variable planConstant, same average
3003504004500510152025KilometrePower (W)441 W320 W
Figure 1. Kilometre-by-kilometre power plan from the SuperRecord demo (2026 Tour time trial) against riding a constant 401 W. Same average power; the variable plan is 37 seconds faster.

At the same 401 W average, riding constant would have cost 37 seconds. Where the time goes is what matters to the coach: 50 seconds are lost on the climb and only 13 are won back on the descent and the finish. The time is where the rider is slow.

How much variation do you need to win that time?

You do not need the full plan to take a good share of it. With the same 401 W average on the same course:

SplitPower rangeTime lost
Constant401 W throughout+37 s
±5 % variation374 to 413 W+22 s
±10 % variation350 to 427 W+9 s
SuperRecord plan320 to 441 Wreference

Table 1. Time lost against the demo plan at the same average power. Calculated with the SuperRecord engine on the 2026 Tour time trial.

Most of the gain comes cheaply: ±10 % already leaves only 9 seconds on the table. With riders on a simulated hilly time trial, a ±5 % split following the gradient was faster than constant power at the same perceived exertion, but two of the seven could not hold it in the second half3. And the variation has to be in the right place: varying power without following the gradient makes the time worse, not better4.

Coach’s rule. Start with a small, well-placed range (around ±5 %), on the climbs and not just anywhere. Widen it towards ±10 % only once the rider has rehearsed it in training and reaches the finish with legs left.

What changes in the training plan?

  • Rehearse the transitions: going from 440 W on the ramp to 330 W on the descent does not happen by itself. Use intervals that link a climb above target with fast flat riding straight after.
  • Tolerance above the average: if the plan asks for 10 % more for 15 minutes, that duration has to be trained, not just the average power.
  • Position on the ramps: if the rider stands or leaves the extensions on every ramp, the plan loses part of its edge. Train the climb in the time-trial position.

How do you take it into the race?

  1. Few numbers: three or four targets by block, not twenty-six. The kilometre-by-kilometre plan is for the coach; the rider needs the blocks.
  2. Screen by section: load the plan on the head unit so the target changes on its own as each block begins.
  3. A range, not a number: “420 to 440 on the climb” is easier to execute than “430”.
  4. Review afterwards: compare each block’s time with the forecast. Time lost on the climb means too little power there; time lost at the finish means too much on the climb.

Frequently asked questions

Does this apply to a flat time trial?

It does with wind. With no climbs and no wind, the gain from varying is small. With a headwind one way, the logic is the same: more power where the rider is slow (see out-and-back time trial in the wind).

What if my rider blows up on the climb?

Then the climbing ceiling was above their reserve. How to set it is in what power ceiling to set on climbs.

You can see the full plan, section by section, in the demo on the Tour time trial.

References

  1. Swain DP. A model for optimizing cycling performance by varying power on hills and in wind. Med Sci Sports Exerc. 1997;29(8):1104-8. doi:10.1097/00005768-199708000-00017
  2. Atkinson G, Peacock O, Passfield L. Variable versus constant power strategies during cycling time-trials: prediction of time savings using an up-to-date mathematical model. J Sports Sci. 2007;25(9):1001-9. doi:10.1080/02640410600944709
  3. Atkinson G, Peacock O, Law M. Acceptability of power variation during a simulated hilly time trial. Int J Sports Med. 2007;28(2):157-63. doi:10.1055/s-2006-924209
  4. Wells MS, Marwood S. Effects of power variation on cycle performance during simulated hilly time-trials. Eur J Sport Sci. 2016;16(8):912-8. doi:10.1080/17461391.2016.1156162

Related articles