October 10, 2026 · 12 min read

1.3 km Short of the Summit — How a Failed Ride Built WattLog.pro's Power Guide

On 19 August 2026 I rode solo from home to Góra Świętej Anny — St. Anne's Mountain — and back. 144 km, flat for most of the way, with the climb to the sanctuary as the point of the whole day. I never rode it. At km 72.4, about 1.3 km from the top, I turned around. The legs were gone, and the 72 km home were the longest kilometres of the year.

I build WattLog.pro, a training platform that reads power, heart rate and cadence, so I had every second of that ride on record. This is what the data said — and how it turned into a new feature.

Two mistakes, one ride

Looking back, I made two mistakes. Only one of them happened on the bike.

I started tired

WattLog.pro tracks training load with the classic fitness–fatigue model: CTL is "fitness", a 42-day average of daily TSS; ATL is "fatigue", a 7-day average; TSB is the difference between them. The idea goes back to Banister's model, which describes performance as fitness minus fatigue [1].

On the morning of the ride my CTL was 30 and my TSB was −15. In the seven days before it I had collected 386 TSS, almost as much as the ride itself, and my CTL had climbed from 21 to 30 in eleven days. Then I went out for a ride worth 400 TSS — thirteen times my average training day. That is not a pacing problem. It is a planning problem. The next part is about the mistake that was a pacing problem.

I rode the average right and the distribution wrong

Here is the interesting part. For the next attempt, Power Guide plans this route at a normalized power of 120–148 W, depending on the strategy. In August I rode it at 128 W, right in that range. On average I rode about the way a plan would have told me to.

The average was not the problem. The variability index was: 1.46, meaning normalized power was 46% above average power. A plan ridden evenly comes out near 1.0. WattLog.pro's Power by grade card places every power reading on the grade at that point of the route:

On the flat I was a calm endurance rider. Every time the road tilted up I rode above threshold: 36 minutes of climbing at an average of 106% of FTP. Over the whole ride I spent 52 minutes above FTP, 32 of them before the turnaround, and started 67 separate efforts above FTP lasting at least 30 seconds.

Under the hill

This is the whole ride: the route profile on top, and below it the power while pedalling for every half kilometre. Orange bars are stretches ridden above my FTP at the time; the white line is the plan Power Guide gives for this route today.

The approach to the sanctuary starts at km 70.9. I rode the hardest three minutes of that stretch at 192 W, 110% of FTP, and my heart rate reached 175 bpm — the highest of the whole ride. Then I stopped and turned around. The way home shows what was left:

For seven hours the efficiency factor (normalized power per heartbeat) stayed between 0.86 and 0.97. In the eighth hour it fell to 0.72: 20% less power for the same effort from the heart. In the last 24 km I spent three minutes above FTP. There was nothing left to spend.

What I was missing

"How hard should I ride the climbs?" — that was the question I could not answer on the day. Pacing research points one way: on variable terrain, riding a bit harder uphill and into headwinds, and easier downhill and with tailwinds, is faster than a constant effort [3, 4]. The key words are "a bit". The models vary power by a modest margin around the average, not by 40%.

I had nothing on the bars that turned that into a number. So I built it.

Power Guide

Power Guide (the Power plan card in the app) gives every stretch of a route a target power. You pick a strategy, the route is split into terrain sections, each section gets a target, and a physics model — gravity, rolling resistance and air drag [5] — turns the targets into speed, time and load for the whole ride.

Sections, not segments

A route is stored as 50 m segments, each with its own grade. Following them directly would change the target every few seconds. Power Guide groups them into grade bands (descent below −3%, flat up to 2%, then climbs in bands of 2–4%, 4–6%, 6–8% and 8%+), merges anything shorter than 400 m into the neighbour closest in grade, and folds a flat shorter than 1 km between two climbs into one climb. A climb that eases for 300 m is ridden as one climb, at the easier of its two targets. Without that, the cockpit would announce "Flat 147 W" five times in a row where nothing changes.

Three strategies

Two constraints shape these numbers:

• Descents get a floor, not "coast". Routes in WattLog.pro are ridden on a smart trainer in SIM mode, where the speed comes from the trainer. Stop pedalling and the virtual rider stops, even downhill. So a descent asks for an easy spin.
• Climbs are capped by your own power curve. The plan works out how long a climb takes at its target, looks up your best power for that duration in the last 90 days, and lowers the target if it asks for more than the strategy's share of it. The plan never asks for four minutes harder than you have ridden four minutes lately.

Finally WattLog.pro compares the plan's normalized power with your best power for its whole duration. If the plan asks for more, it says so instead of quietly trimming it — an FTP set too high and a lack of long hard rides look the same from here.

The plan for St. Anne's Mountain

This is how Power Guide shows the route with my current FTP of 184 W and the "Easy finish" strategy — 19 sections: 8 climbs, 2 descents and long flats, one of them 46.7 km:

The line that matters is Climbs. Even the most aggressive strategy, with an FTP 9 W higher than I had in August, never asks for more than 162 W on these climbs. This is how I rode them:

Climbs longer than 0.4 km; mean power over the section, as in the post-ride audit. Left out: km 71.7–72.4, where I stopped and turned around.

During the ride

The chosen plan rides along. Above the route cockpit, one line shows the current section, its target, the distance left and whether you are on target: within ±5% (never tighter than ±5 W) of a 5-second average, because a verdict that flips with every pedal stroke tells you nothing. 300 m before the next section it shows what is coming. This is how the approach to St. Anne would look:

Targets switch by distance, not time: the climb starts at km 51.9 however fast you got there.

After the ride: the pacing audit

When the ride is saved, WattLog.pro compares ridden power with the plan, section by section. Sections of at least two minutes ridden 15% or more over or under target get a sentence. If an over-hard section is followed by one ridden at 90% of target or less, a second sentence appears: the cost of going too hard, made visible.

I ran the August ride through that audit against the "Easy finish" plan. It is the same function WattLog.pro runs after a route ride, fed with the moments I reached each section:

Four climbs longer than two minutes, four times too hard, four times followed by a drop. The whole ride in four sentences.

The audit also marks almost every flat as too easy. Partly that's true, and partly it's riding outdoors: traffic lights, junctions and the stop at the turnaround count as zero watts. The audit was built for trainer rides, where there are none.

What it doesn't do

What I'd take from it

I came home with a TSS of 400 and the feeling that the distance was the problem. The data said otherwise: the distance was fine; I rode the climbs like intervals, and I started the day already tired. Power Guide answers the question I couldn't answer at km 51: how hard, exactly, on this climb. For the second attempt at St. Anne's Mountain, the plan says 130–132 W on the approach to the sanctuary. Last time I rode its hardest three minutes at 192.

References

1. Calvert TW, Banister EW, Savage MV, Bach T. A systems model of the effects of training on physical performance. IEEE Trans Syst Man Cybern. 1976;SMC-6(2):94–102.
2. Skiba PF, Chidnok W, Vanhatalo A, Jones AM. Modeling the expenditure and reconstitution of work capacity above critical power. Med Sci Sports Exerc. 2012;44(8):1526–1532. doi:10.1249/MSS.0b013e3182517a80
3. Swain DP. A model for optimizing cycling performance by varying power on hills and in wind. Med Sci Sports Exerc. 1997;29(8):1104–1108. full text
4. 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):1001–1009. doi:10.1080/02640410600944709
5. Martin JC, Milliken DL, Cobb JE, McFadden KL, Coggan AR. Validation of a mathematical model for road cycling power. J Appl Biomech. 1998;14(3):276–291. doi:10.1123/jab.14.3.276

← All blog posts