GOOD LEGS LAB / CYCLING LOAD · VERSION 0.1
What your riding
means for
the gym.
A big riding week changes the question you bring to the gym. Can your ride data help us make better strength-training decisions? Here is how we propose to find out.
These are 42 made-up days, with no gaps in the records. Both lines start at 40 points/day. The Recent line reacts faster when the riding increases. They show riding history, not fitness or recovery.
01 / THE QUESTION
Before the gym,
look at the riding.
What did you ride?
How did it affect you?
Two hours of steady riding and a short, hard session need not leave you feeling the same the next day. We want a useful record of the riding, then a fair test of what it can tell us about the next session.
Cycling Load is our proposed way of combining recorded power and ride time into points. Training State tracks those points over time. Both describe past riding. We still need to test whether they help predict how a rider will feel.
- 01 / RIDEWhat was recorded?
Record the power and ride time, including any gaps. Keep the original power pattern and the total work done at the pedals.
- 02 / LOADA load score.
Give harder riding more weight when calculating points. Keep the detail too: the score cannot describe the whole ride.
- 03 / HISTORYYour riding history.
Track how riding has changed recently and how it compares with the longer history.
- 04 / RESPONSEWhat happens next?
Check whether the information improves a forecast made before the next ride. Test training changes separately.
02 / ONE NUMBER ISN’T THE RIDE
Same score?
Look closer.
Try the three comparisons below. Each uses a critical power (CP) of 300 W: a power reference estimated from hard efforts over different durations. For these examples, that reference stays fixed. The rides are made up to show the maths, not sessions to follow.
MADE-UP RIDE COMPARISON / CP 300 W
Same work, different intensity.
120 min at 150 W
50Load points- Mechanical work
- 1,080 kJ
- Work above CP
- 0 kJ
72 min at 250 W
83.33Load points- Mechanical work
- 1,080 kJ
- Work above CP
- 0 kJ
Both rides involve 1,080 kJ of work at the pedals, and neither goes above the 300 W CP reference. The shorter, harder ride receives more points.
Made-up rides, real arithmetic. Points, work and work above CP describe different things. Do not add them together or use these examples as workouts.
See all three ride comparisons
Same work: 120 minutes at 150 W gives 50 points; 72 minutes at 250 W gives 83.33. Both give 1,080 kJ mechanical work and zero excess above CP.
Same points: four hours at 150 W and one hour at 300 W each give 100 points. Their mechanical work is 2,160 and 1,080 kJ respectively; both have zero excess above CP.
Same excess: ten minutes at 360 W and twenty minutes at 330 W each give 36 kJ excess above CP. Their points are 24 and 40.33; their mechanical work is 216 and 396 kJ.
The order of the effort matters too. Switching between 400 and 200 W every second gives the same two-minute totals as a minute at each power. The score cannot distinguish the patterns. That does not mean your body responds to them in the same way.
How we calculate load points
For each recorded interval, divide power by CP, square that number and multiply by the interval’s length in seconds. Add the results, then multiply by 100 / 3,600. CP must be an estimate recorded before the ride. Squaring gives high power more weight; we need to test whether that weighting is useful.
The scale is set so that one mathematical hour at CP gives 100 points. That is a choice of units, not a claim that you can hold CP for an hour or a suggestion to try.
Mechanical work is power multiplied by time, added across the ride and divided by 1,000 to give kilojoules. Work above CP counts only the power above that reference. It does not tell you how much W′ — the work capacity used in the critical-power model — remains.
CP and functional threshold power (FTP) are estimated differently and can differ. The calculation requires CP; replacing it with FTP changes its meaning. Check the power recording and the reference estimate. Calculate from the recorded intervals: averaging power first can change the answer.
03 / TRAINING STATE
Has your riding
changed?
Recent follows changes in riding more quickly. Base changes more slowly. Both average your daily load points, with older rides gradually carrying less weight. The proposed seven- and 42-day settings control how quickly the lines react.
Both lines use points per day. Seven and 42 days describe how quickly the averages respond; they are not cut-offs that discard every older ride. They are also not recovery times. A higher line does not, by itself, mean fitter, more tired or less ready to train.
A missing ride is not a rest day. If only half a ride was recorded, we have only part of its load. The history needs to show those gaps rather than treat them as zero.
The reference matters. Raise CP from 300 to 330 W and the same ride receives 17.36% fewer points, while its mechanical work stays unchanged. The line can move because the reference changed, even though you did exactly the same riding.
How the two history lines are calculated
history today = (1 − α) × history yesterday + α × known daily points
Each day’s history must use only records available on that day. We must show where it started and any earlier gaps, rather than fill unknown days with zero. The chart uses complete made-up records and a starting value of 40 points/day for both lines.
04 / THE STRENGTH QUESTION
What did you
actually lift?
Does a detailed gym log help us predict how your next ride will feel?
Our first proposed test asks riders how hard a familiar cycling effort feels the next day. After an easy warm-up, they ride for six minutes at a fixed power and report their whole-body effort in the final minute. This measures one cycling response; it is not a strength test or an all-purpose readiness score.
We would compare two forecasts. Both know the rider’s cycling history, basic gym duration and effort, and relevant background information. One also knows what the rider actually lifted: exercises, weights, sets, repetitions, timing and whether the work was familiar. Both must use information available at the same time.
This asks whether logging gym detail adds anything beyond knowing that a rider lifted. The detailed forecast also has to beat the rider’s usual response and other sensible, simple forecasts. We test it on later rides that neither model has seen.
Predicting the next ride better would be useful progress. It would not show that moving a gym session, reducing sets or changing a weight helps. We would need a separate study comparing a specific training change with a credible alternative, with suitable safety oversight.
The study is still being planned. There are no participant results, and it is not approved, registered or recruiting. Joining the app waiting list does not enrol you in research.
Read the related proposed Recovery Curve protocol ↗05 / WHAT WE NEED TO TEST
How would we know
it helps?
A calculation can be correct and still tell a rider very little. We need to answer four different questions.
Check the maths.
Does the calculation give the right answer? Check units, power intervals, reference changes and incomplete recordings. Agree the definition before comparing versions.
Check the measurement.
Can we measure the response consistently? Check the recording quality, repeat the familiar task and make sure the process is practical for riders. Participant research needs approval and safeguards before it starts.
Predict a later ride.
Can it predict a later ride? Give both models the same information deadline and test on later observations they have not seen. Compare their errors with simple forecasts and report how uncertain the answers are.
Test the training change.
Does acting on it help? Test a defined change to strength planning against an alternative, then measure the outcomes riders care about.
Version 0.1 is a proposal. We still need to test whether it tells us more than the cycling load scores riders already use.
06 / PRIMARY RESEARCH
The research
we’re drawing on.
These studies help us decide what to ask and how to test it. They do not establish that our load formula works or that it should change your strength sessions.
- Sanders et al. (2017) — Methods of monitoring training load and their relationships to changes in fitness and performance in competitive road cyclistsWe need to compare our score fairly with load methods cyclists already use.
- Bossi et al. (2020) — Optimizing interval training through power-output variation within the work intervalsThe pattern of hard and easy efforts can matter, even at the same average power.
- Karsten et al. (2021) — Relationship between the critical power test and a 20-min functional threshold power test in cyclingCP and FTP can differ. A formula designed for one should not quietly use the other.
- Hamilton et al. (2024) — Durability of the moderate-to-heavy-intensity transition is related to the effects of prolonged exercise on severe-intensity performanceA long ride can change how an athlete responds to later efforts. The test we choose matters.
Based on Good Legs Cycling Load and Training State, version 0.1, dated 8 October 2026. The accompanying analytics handover informed the background. Internal implementation documents are not published here.
Written by Good Legs. Questions or corrections: tony@
07 / THE QUESTIONS IN YOUR WEEK
What does this mean
for your gym work?
Can a cycling load score tell me whether to lift?
Not on its own. The score describes recorded riding; it cannot currently tell you which exercises, sets or weights to choose today. We first need to test what it predicts, then test whether acting on that prediction improves training.
If you want help arranging gym days, start with your key rides and the days you can lift. That guide helps with the calendar; it does not assess recovery.
Do two rides with the same load points have the same effect?
Not necessarily. At CP 300 W, four hours at 150 W and one hour at 300 W both give 100 points. They involve 2,160 and 1,080 kJ of work respectively, and the score leaves out the order of the effort. Equal points cannot establish how either ride affects you.
Compare the rides or explore the separate question of same-day session priorities.
Can I use FTP instead of critical power?
This formula uses a CP estimate recorded before the ride. CP and FTP come from different test procedures and can differ. Swapping them changes the points, even when the ride stays the same. This page shows examples rather than calculating your own load.
Read the research comparing CP and a 20-minute FTP test, then inspect the reference-change example.
Why record strength sets, repetitions and load alongside cycling?
They record what you actually did in the gym. We want to know whether that detail improves a forecast beyond ride history and basic gym duration and effort. The first proposed test asks how hard a familiar fixed-power ride feels the next day. It does not measure strength performance or prove that an automatic gym adjustment helps.
See how we would compare the forecasts. We have no participant results yet.
These answers describe version 0.1. How we use research in the Lab and our practical guides.