Dose and Response: The Two Signals That Predict Your Race Pace (We Tested 35,906 Runs)
A viral post said your Parkrun 5k comes down to two numbers: training load and speed at a given heart rate. We tested it on 35,906 runs from 479 athletes. The structure held, and the catch is the most useful part.
How Do You Measure Fitness Without a Lab?
A post from Brownlee has been going around, and it framed a problem better than we usually hear it. How do you measure running fitness from real-world training, without a treadmill test and a mask?
The issue is noise. Every run is different. A different route, different weather, a different effort, a different night of sleep behind it. Compare one session to the next and you learn almost nothing. What you want is a standardized effort that lots of people repeat in the same place at full effort. Parkrun is about as close as free-living running gets. Same 5k, same course, every Saturday.
The claim was that most of what predicts your Saturday time sits in two numbers. Your dose, meaning how much training you have banked (Chronic Training Load, the 42-day rolling load). And your response, meaning how efficiently your body handles it, measured as your speed at a given heart rate. Neither is strong alone. Combined they roughly double what you can explain, and they barely overlap, because one is volume and the other is efficiency.
We cannot pull Parkrun results. But we have a lot of runs sitting next to a lot of training-load history, so we ran the same idea on our own data.
What We Did
We pulled 35,906 runs that carried both heart rate and pace, from 479 athletes on athletedata. For each athlete we used their hard efforts as a stand-in for a race-pace test, since we have no actual 5k time-trial set. To keep it fair, we scored each effort against that athlete's own distance-expected pace, so a short fast run did not automatically look better than a long one. Then we lined each effort up against two things as of that day: the athlete's CTL, and their recent speed at a given heart rate from easier runs.
Everything is within-athlete. We are asking "you versus your own normal," not comparing fast runners to slow ones. That distinction matters. Cross-person averages mostly measure who someone is, not whether a signal tracks their fitness.
Each Signal Alone Is Weak
Within an athlete, training load correlated about r = 0.13 with hard-run pace. Recent efficiency, about the same. In plain terms, each one explains under 2% of how fast a given hard run goes. Hand us only your CTL, or only your efficiency, and we could barely guess your pace on any single day.
This is the part people miss when they get excited about a single metric. CTL is not your fitness. Efficiency is not your fitness. Each is one noisy slice of it.
They Barely Overlap, Which Is the Point
The two signals correlated only about 0.16 with each other within an athlete. Brownlee reported 0.11, the same neighborhood. That low number is good news. It means fitness-from-volume and fitness-from-efficiency are mostly independent, so each one carries information the other misses.
That is also why efficiency factor (your pace per heartbeat) has stuck around as an aerobic-fitness marker for years. As TrainingPeaks describes it, a rising EF means more pace for the same heart rate, which is a different question from how much you have been training.
Together They Nearly Double, and It Holds for Almost Everyone
Stack the two and the combined predictor nearly doubled the explained variance, 1.8 times the better single signal. More convincing than the average: the additive model beat the best single signal for 94 of our 96 athletes. Brownlee found it won for all of his. Two datasets collected completely differently, same conclusion. When a result replicates like that, it is worth believing.
The Catch Is the Real Lesson
Here is the number we cannot round away. Even combined, the two signals explain only about 3% of how fast any single run goes. For one run, that is almost nothing.
That looks like the finding falling apart. It is the opposite. It is the reason Parkrun is so useful. Any single run is buried in noise, a hill here, a headwind there, a route you have never run before, legs that are flat for no clear reason. The fitness signal is real, but in one session the noise is much bigger. A fixed 5k on the same course at full effort scrapes most of that noise away, which is why it reads your fitness so cleanly.
You probably do not run a weekly Parkrun. But you have hundreds of your own runs, and across enough of them, dose and response point the right way. The trend is the signal. The single day is mostly noise.
What To Actually Do With This
Track both numbers and watch the trend over weeks, not the reading on any one day.
- Is your CTL drifting up over a training block? That is your dose, the aerobic base you are building.
- Is your pace at a given heart rate improving over the same stretch? That is your response, your body getting more out of that base.
If both are rising, you are getting fitter in two independent ways and your race pace almost certainly follows. If one is flat, you know which half to work on. A big base with no efficiency is an engine that has not learned to run fast yet. Sharp efficiency with no volume is a sharp engine with no tank behind it. This is the same reason training load tracking and heart-rate-based fitness reads tell you more together than either does alone.
And if you do have a Parkrun or any repeatable hard effort, use it. It is the cleanest read you will get without a lab. The two wrist numbers explain it. The fixed course is what lets you see them.
athletedata connects your training and recovery data and reads your own dose and response over time, so you can see your fitness trend instead of guessing from one run. Get started here.
