Aerobic metrics¶
Training load and fitness tell you how much work you did. These four numbers describe how the work went — whether your aerobic engine is getting stronger, how steady the ride was, and how much anaerobic capacity you spent along the way.
They appear in the Aerobic response card on each ride, and the efficiency trend is charted on the Power view.
Rides only — and training rides at that
The card is shown on cycling activities only. Every figure in it is defined against cycling power, so on a run, a swim or a gym session there is nothing for it to say.
It is also hidden on rides you have labelled commute. Stops, traffic and a loaded bike make these numbers describe the journey rather than your aerobic fitness. Remove the label on the activity page if you want the card back for a particular ride.
The four metrics¶
| Metric | What it tells you |
|---|---|
| Efficiency factor | Weighted power per heartbeat (W/bpm). How much power your heart is buying you. |
| Variability index | Weighted power ÷ average power. How steady the ride was: 1.00 is metronomic, above 1.10 means surging. |
| Aerobic decoupling | How far your power-to-heart-rate ratio drifted from the first half of the ride to the second. |
| W′ balance | A per-second stream showing how much of your anaerobic capacity was left at each moment. |
Efficiency factor¶
This is the one to watch across a base phase. Divide the ride's weighted power by its average heart rate and you get how many watts each heartbeat is buying you.
If that number climbs over weeks while your training load stays flat, you are getting aerobically fitter in a way the fitness curve can't show you: the fitness number would look identical whether you rode those hours well or badly.
The Power view charts your efficiency factor over time, one point per steady endurance ride. Interval sessions, very short rides and rides missing power or heart rate are left out, so every point on the chart is comparable with the others.
Compare like with like
Efficiency factor moves with the kind of riding you do. A tempo ride and a recovery spin will not produce the same number, and neither is "better". The trend across similar rides is the signal; a single ride's value on its own is not.
Variability index¶
Weighted power divided by average power — a one-number summary of how punchy the ride was.
- Around 1.00 — steady state. A time trial or a disciplined endurance ride.
- 1.00–1.10 — normal rolling terrain and traffic.
- Above 1.10 — intervals, a hilly race, or a lot of stopping and starting.
openkoutsi already used this internally to categorize your workouts; now you can see it. It is also what decides whether a decoupling figure is worth showing.
Aerobic decoupling¶
Split the ride in half. Work out the power-to-heart-rate ratio for each half. If the second half's ratio is lower, your heart rate crept up while your power didn't — your aerobic system was fading. That drift, as a percentage, is aerobic decoupling.
- Under 5% — generally considered good aerobic durability.
- Above 5% — you faded. Over a long ride that may simply mean you rode past what your endurance currently supports.
A negative number means the second half was the more efficient one, which usually means you started conservatively and warmed into the ride.
Stops, and which part of the ride is measured¶
Half of a ride is a strange thing to talk about once you have stopped in the middle of it. Stand around for an hour at lunch and your heart rate is back where it started; comparing what came before against what came after measures the lunch, not your durability.
So the figure is taken over the longest continuous block of the ride:
- A short stop changes nothing. Anything under ten minutes — a traffic light, a puncture, a coffee — is ridden through. Your heart rate is back where it was within a couple of minutes of rolling again, so the ride either side of it is still one ride. A five-minute stop on a seven-hour ride is measured straight across.
- A long stop divides the ride, and the longest piece is the one measured. The card then tells you how much of the ride the figure covers — "measured over 4h 12m of 7h 03m" — so a number over part of your day is never presented as the whole of it.
- A stop is time with nothing recorded on it, whichever way your head unit writes it down: paused so that nothing at all is logged, or left running while the power meter — which stops broadcasting when the cranks stop — goes quiet and the strap keeps counting. Neither is a fault, and neither costs you the figure any more.
A heart-rate strap that drops out while you are riding is a different thing: the watts have no pulse to be paired against, and that still shows as recordings that don't line up.
When openkoutsi won't show you a number¶
A decoupling figure computed over a hard interval session is noise, and presenting noise as if it meant something is worse than showing nothing. So the number is only stored when it can be trusted. Otherwise the card tells you why it is missing:
| Reason shown | What it means |
|---|---|
| The ride was too short | Decoupling needs roughly an hour of steady riding. |
| Stops broke the ride up | The ride was long enough, but no continuous stretch of it lasted the hour the measurement needs — stop-start riding with long breaks in it. |
| No power data | Both power and heart rate are required. |
| No heart-rate data | As above. |
| The heart-rate data is unusable | A flat trace, for example. |
| The recordings don't line up | Power and heart rate have to be compared moment for moment; a heart-rate strap dropping out repeatedly while you ride makes that unreliable. |
| This was an interval session | The measurement would describe the intervals, not your durability. Judged on variability index together with how hard the ride was and how long it lasted — nobody rides intervals for three hours — so a long ride made variable by descents, junctions and stops still gets its figure. |
| The two halves were ridden differently | A ramp or a negative split produces a big number that reflects your pacing, not your durability. |
Heart-rate drift is not purely a fitness signal
Heat, dehydration, caffeine, altitude, illness and poor sleep all push heart rate up over a long ride. A hot summer ride can show high decoupling from an athlete whose fitness is fine.
Read the number in the context of the day you had. A trend across many similar rides in similar conditions is meaningful; one ride in a heatwave is not.
W′ balance¶
Above your critical power you are spending a finite anaerobic reserve. Below it, you are refilling it. W′ (pronounced "W prime") is the size of that reserve in kilojoules, and W′ balance tracks how much of it you had left at every second of the ride.
Turn it on in the stream chart on any activity to see the story of a hard session: the reserve draining through each effort, and how much it recovered in the gaps before the next one. A ride where the balance hits zero and stays there looks very different from one where you kept topping it back up — even if both have the same training load.
The critical power and W′ used are estimated from your own power bests as they stood on the day of that ride, and stored with the activity. That means an old ride keeps the W′ story it actually had at the time, instead of being quietly rewritten every time your power curve improves.
Needs enough power history
Estimating critical power takes a spread of hard efforts from two to twenty minutes. Until you have those — and until they produce a believable pair of numbers, which a diet of purely steady riding does not — openkoutsi shows no W′ balance rather than guessing at a reserve size. A W′ curve built on an invented number would look convincing and mean nothing.
Needs 1-second recording
W′ balance is calculated second by second, so it needs your head unit set to 1-second (not "smart") recording. On a file recorded at a lower rate the arithmetic would be wrong by the sampling ratio and the error would grow across the ride, so openkoutsi leaves the stream out instead.
Getting these on older activities¶
Efficiency factor and variability index are calculated on the spot from data every activity already has, so they show up on your whole history immediately.
Aerobic decoupling and W′ balance are derived from the per-second data streams when an activity is processed. Rides that were uploaded before this feature existed pick them up when you reprocess the activity from its detail page.
A reprocess re-reads the recording, it does not re-make it
Reprocessing recomputes from the streams the activity was stored with, so a ride keeps the recording it arrived with. Rides uploaded before openkoutsi put every stream on a shared clock — where a dropout leaves a marked hole rather than silently shortening the channel — keep the older shape, and on those a long strap dropout genuinely does leave power and heart rate out of step. Upload the ride again (or re-sync it from your provider) if you want it read on the shared clock.
What the AI coach sees¶
When you ask for an AI analysis of a ride, efficiency factor, variability index and the decoupling figure are included in the summary sent to the model — or, if decoupling was not measured, the reason why, so the coach doesn't invent one. See Your data & AI for the full list of what is sent.