Eleven athletes now have both tests, treadmill and forest, at 4 mmol/L. Names are codes A1–A11, shuffled so they do not follow speed.
Treadmill speed sets the broad level. It explains about 79% of the spread in forest speed, so roughly 21% is something else: terrain, technique, rough-ground running. That 21% is a point estimate. With only 11 people it could plausibly be a lot bigger or a lot smaller.
What the leftover looks like
Forest speeds in the group vary by 1.12 km/h (standard deviation). Once treadmill speed is accounted for, the leftover is 0.53 km/h. In the forest that is about ±15–17 s/km. Forest speed here runs from 58% to 68% of treadmill speed, averaging 63%.
Forest speed relative to the treadmill line
Click an athlete. Bars are km/h faster or slower in the forest than the treadmill line predicts.
Leftover scatter after the fit is 0.53 km/h, about ±15–17 s/km. Forest speed in this group is 58–68% of treadmill speed, averaging 63%.
The residuals run from +0.81 km/h (A5) to −1.01 km/h (A4).
How sure is 0.79?
Two usual methods, and they agree. Fisher's z assumes the speeds are roughly normal. The bootstrap redraws the 11 athletes over and over. One person can swing a redraw, so both are shown.
90% confidence interval for R²
Fisher z0.48–0.93
Bootstrap0.54–0.95
The dot is the estimate, R² = 0.79. With only 11 athletes the range is wide, but it stays well above zero.
At 90% confidence, treadmill speed explains somewhere between about half and a bit over 90% of forest speed. If R stayed near 0.89, the planned sample of 25 athletes would pull the lower bound up to about 0.6 which is much better.
Some of the leftover is just noise. Each 4 mmol/L speed comes from five points, and a few fits were not perfect, so those values probably carry ±0.2–0.4 km/h of uncertainty on their own.
A treadmill test at the IFK orienteering club gym.
It's October, and so far 15 tests have been completed successfully.
Treadmill tests
14
Forest tests
1
Retests needed
0
Summary
Fourteen treadmill tests (C1) are complete, and the first forest test (C2) went well. So far no session has needed to be retested or redone, and the measurements have been of good quality across the board and have gone as planned.
A First Look at the Treadmill Data
To give a sense of what each test produces, below is one anonymised treadmill test. Lactate stays low and fairly flat through the first three stages, then rises sharply over the last two — the classic shape the study relies on to locate the threshold. A simple linear interpolation between the last two stages puts this athlete's speed at 4 mmol/L at roughly 17.7 km/h. The final analysis will use the polynomial curve fit described in the proposal, so this number is illustrative only.
The speeds shown are the adjusted values from the treadmill calibration described in the previous update. In this test they differ from the displayed speed by up to 0.3 km/h, which is exactly the kind of gap that would otherwise blur comparisons between the two treadmills.
Athlete A — Treadmill lactate curve
Blood lactate at the end of each 5-minute stage, plotted against calibrated treadmill speed. The square marks the speed at 4 mmol/L, linearly interpolated between the last two stages.
Heart rate
Perceived exertion
Show treadmill data
Stage
Displayed (km/h)
Adjusted (km/h)
Lactate (mmol/L)
HR (BPM)
RPE
1
13.0
13.23
0.8
147
9
2
15.0
15.10
1.3
157
13
3
16.5
16.46
1.4
173
16
4
17.5
17.34
2.7
178
18
5
18.5
18.21
5.6
187
19
Lactate, heart rate, and RPE recorded at the end of each stage. Adjusted speed is the displayed speed corrected using the treadmill calibration.
Challenges from the Forest Test
The marked forest loop used for the C2 intervals.
The forest condition has proven to be the harder of the two to standardise. On a treadmill, interval duration is fixed and speed is the variable. In the forest the loop itself is fixed, so the time each runner takes to complete it varies widely with ability. To keep intervals close to the intended ~5 minutes, I have to estimate, for each athlete, how much running distance they would cover in roughly that time.
In practice, this means running different versions of the loop in the first interval (the slowest) and in the last interval, so that interval times stay roughly equivalent. The first forest test shows what this looks like. Reps 1–3 were made up of two short sections and one long section, while reps 4–5 swapped one short section for a second long one. To compare the reps on equal terms, each is adjusted down to a single loop.
Athlete B — What each forest rep was made of
Short section (240 m)
Long section (445 m)
Sections run in each rep (not in running order). Times show the raw rep time, then the time adjusted to a single loop.
Athlete B — Forest lactate by rep
Blood lactate after each rep, plotted against speed from the adjusted single-loop pace and connected in rep order.Show forest data
Rep
Short / Long
Time
Adj. time
Distance (m)
GPS (m)
Adj. GPS (m)
Pace
Adj. pace
Lactate
RPE
HR
1
2 / 1
5:35
4:15
925
943
702
6:02
6:12
0.8
12
137
2
2 / 1
5:12
3:45
925
958
712
5:37
5:28
1.3
15
152
3
2 / 1
4:25
3:21
925
948
704
4:46
4:53
3.3
16.5
166
4
1 / 2
5:42
3:24
1,130
1,169
708
5:02
4:57
4.1
18
171
5
1 / 2
5:20
3:14
1,130
1,157
702
4:43
4:43
7.1
19.5
174
Loop distance is 685 m (short section 240 m, long section 445 m). For reps longer than one full loop, the time is adjusted down to a single loop by averaging the times of the duplicated section, and the GPS distances of the short and long sections are averaged then added together. Adjusted pace is the adjusted time divided by the loop distance. Pace in min/km; lactate in mmol/L; RPE on the Borg 6–20 scale; HR in BPM.
That solves the timing problem but introduces another one. Because the loop is not uniform, some runners end up spending proportionally more of an interval in the tougher sections of the terrain than others. In this test, rep 4 — the first rep with the extra long section — was slightly slower than rep 3 on adjusted pace (4:57 vs 4:53 min/km), yet lactate, heart rate, and RPE all rose. From one test it is impossible to say how much of that is the change in terrain and how much is simply fatigue building across reps, and that ambiguity is a potential source of bias, since terrain cost is exactly what the secondary research question sets out to measure.
There are two options I am currently weighing:
Fixed loop sets. Use a fixed set of loop versions for men and a slightly shorter fixed set for women, so that every athlete within a group covers the same terrain.
Adjust in analysis. Keep tailoring the loop to the athlete and account for the differences in terrain afterwards, when comparing C1 and C2.
Thank You
I am grateful for how many people have stepped up for these tests am thankful to every participant who has been a part of the study so far. Keep an eye out for the next update!
