In ultras, the second half
says more than VO2max.
On 17 July I ran KBL: 110 km, about 4,400 m of climbing, 18.6 hours. The last 10 km were the fastest stretch of the whole race. Runalyze gave that run an effective VO2max of 27.6 — the value of someone who has only just started running. Instead of writing it off as an error, I checked what the number actually measures. Since then I no longer judge my ultra form by VO2max.
Three pillars we have known for 35 years
Joyner described the classic endurance model in 1991. Maximal oxygen uptake, running economy and the largest fraction of VO2max you can sustain — these three explain most of the differences between athletes over 5 km, 10 km and the marathon.
The model rests on an assumption that went unquestioned for a long time: that these three values are fixed. That a VO2max measured on fresh legs still describes you in the twentieth hour. It does not.
What breaks along the way
Andrew Jones published a paper in The Journal of Physiology whose title sums up the problem: The fourth dimension (online 2023, in print 2024). His argument: the three classic determinants are not static — they deteriorate during prolonged exercise, and at different rates in different people.
Two runners with identical lab results are no longer identical after eight hours. Jones proposes treating this resilience as a fourth, independent determinant of endurance performance: you can have a high VO2max and poor resilience, or the other way round.
Durability — a term worth knowing
Two years earlier Maunder and colleagues named the same mechanism in Sports Medicine: durability, the resistance to deterioration of physiological characteristics during prolonged exercise. In 2025 Hunter and colleagues (with Maunder and Jones) described in Experimental Physiology how to measure durability and which methodological traps to watch for.
And this is the point for a runner: the trace of durability is visible in your own watch file. No lab needed.
How to see it in your own file
The simplest method in the literature is decoupling: split the effort into a first and second half and compare how much work you get per heartbeat — power to heart rate, or pace to heart rate if you have no power meter. If the second half needs a higher heart rate for the same work, your physiology is deteriorating.
There is one caveat in the mountains: pace lies. The same kilometre uphill and downhill is completely different work, so pace to heart rate on a mountain course says more about the profile than about you. Power is far more honest here.
My numbers
KBL, 110 km. In the second half I got 2.7% more power out of every heartbeat than in the first. To be fair, here is how: power fell from 192 to 178 W, but heart rate fell further — from 123 to 111. That was not speeding up; it was working at an ever lower heart rate. Pace to heart rate from the same run shows +1.4%, a slight drift the other way.
A night training run, 47 km and 2,000 m of climbing, 7 hours 38 minutes. Average heart rate 119, 66% of maximum. The final stretch was the fastest of the session. Runalyze’s effective VO2max: 28.1.
Now the picture over several years. On road races my VO2max has stayed roughly where it was: 50.0 in a 2023 marathon, 46.7 and 47.8 in half marathons in 2025 and 2026. Over the same years my second-half slowdown in mountain races, adjusted for climbing, fell from 20.4% in 2023 to 9.6% at KBL — with 15.1% and 12.5% in between. That is 15 races from my watch files.
The model does not lie. It simply measures something other than what that race demanded.
What this means for training
If durability is a separate determinant, improving VO2max need not improve it. The literature lists candidates: fat oxidation capacity, glycogen stores, muscle resistance to damage, economy maintained despite fatigue. Jones says plainly that the mechanisms are unresolved.
In my case one common claim did not hold: more kilometres did not buy more durability. My smallest slowdown came in the year I ran half the weekly volume I ran two years earlier. I do not know what worked — the candidates are pacing discipline early on, mixing running with walking, and plain experience. This is one person, not a rule.
A practical takeaway if you judge your form from an app: if your VO2max and race predictions are falling while you finish long runs better and better — do not panic. Look at the second half, not at the number on the first screen.
What I do not know
1. What a lower heart rate in the second half means. At KBL power per heartbeat rose because heart rate fell faster than power. That may be fitness, but it may also be fatigue that stops the heart rate from climbing. I cannot settle that from the file alone.
2. What happens beyond a full day. Durability research is mostly about the marathon and cycling — hours, not days. My measurements reach 23 hours. In two weeks I run ŁUT 150 and will try to learn more — but that will be one case report, not proof.
3. How much of it is trainable and how much innate. In my series 2026 is a single race — a point, not a line.
Test it on your own data — 14 days free
After every long run UltraFlow tells you plainly whether your durability held and compares the second half with the first — instead of leaving you with an average and a VO2max. 14 days free, card at sign-up, cancel any time.
Sources
Joyner M.J. (1991). Modeling: optimal marathon performance on the basis of physiological factors. Journal of Applied Physiology, 70(2), 683–687. doi
Maunder E., Seiler S., Mildenhall M.J., Kilding A.E., Plews D.J. (2021). The Importance of ‘Durability’ in the Physiological Profiling of Endurance Athletes. Sports Medicine, 51(8), 1619–1628. doi
Jones A.M. (2024). The fourth dimension: physiological resilience as an independent determinant of endurance exercise performance. The Journal of Physiology, 602(17), 4113–4128 (online 2023). doi
Hunter B., Maunder E., Jones A.M., Gallo G., Muniz-Pumares D. (2025). Durability as an index of endurance exercise performance: Methodological considerations. Experimental Physiology, 110(11), 1612–1624. doi