LT1 and LT2 Calculator Cycling: Percentage vs Step Test
An LT1 and LT2 calculator for cycling either scales your FTP or fits step-test data. On a 250 W rider the LT1 answers span 37.5 watts.
An LT1 and LT2 calculator for cycling does one of two different jobs, and the two are not interchangeable. One kind takes your FTP and multiplies it by a fixed percentage. The other takes blood lactate values you measured stage by stage in a step test, fits a curve, and reads the two thresholds off that curve. The first needs one number from you and answers instantly. The second needs a test.
Both get sold as calculators. Only one of them has seen anything about you beyond a single watt figure, and the gap between them is wider than most riders expect.
Two tools, one name
| Percentage estimator | Curve fit from step-test data | |
|---|---|---|
| Input | FTP, or HRmax / threshold heart rate | Power and blood lactate per stage, plus a resting sample |
| Output | Two watt figures scaled from your FTP | LT1 and LT2 interpolated from your own curve |
| Time cost | Seconds | Meter, strips, a protocol, roughly 30–45 minutes |
| Reasonable use | A placeholder, or a sanity check on a number you already have | A baseline you intend to retest against |
| Fails when | You are not the rider the percentage was derived from, and nothing in the input can tell you whether you are | The protocol collected too few stages, or stages too wide, around the threshold you care about |
The percentage path is enough when you have no meter and no near-term plan to get one, when you want a rough ceiling for one ride this weekend, or when you are checking whether a fitted number that came back looks out of place. It stops being enough for any figure you intend to repeat, compare, or defend against a retest six weeks later.
What the percentage rules cost you, in watts
The published anchors do not agree with each other. For LT1 you will find figures circulating between roughly 60% and 75% of FTP; for LT2, between roughly 90% and 100%. Before trusting either output, check which percentage your calculator applies and whether it says where the figure came from.
Run that disagreement through arithmetic on a rider with a 250 W FTP:
| Threshold | Low anchor | High anchor | Spread |
|---|---|---|---|
| LT1 | 60% → 150 W | 75% → 187.5 W | 37.5 W |
| LT2 | 90% → 225 W | 100% → 250 W | 25 W |
Thirty-seven watts is not a rounding difference. It is the whole distance between two tools you could open in adjacent browser tabs, both described as an LT1 and LT2 calculator for cycling.
The spread also scales with you. Both bands are fixed slices of FTP — fifteen percentage points wide for LT1, ten for LT2 — so a 200 W rider gets bands of 30 W and 20 W, and a 350 W rider gets 52.5 W and 35 W. Getting stronger widens the estimate in absolute terms, and nothing you do in training narrows it, because nothing about you enters the calculation. Feed the same FTP from two different riders into the same estimator and it returns the same two numbers. That is not a flaw in the tool. That is what a percentage is.
Which produces the awkward part: the threshold that percentages resolve worst is the one most riders open the calculator for. The LT2 anchors cluster within ten points of each other, so the arithmetic cannot spread them far. The LT1 anchors sit twenty-five to forty points below FTP, and the same disagreement expressed in percent turns into more watts — on exactly the number riders reach for when deciding how easy an easy ride has to be.
If you already have step-test values sitting in a notebook, the percentage path is the wrong one to be on. Free online lactate test analysis covers what happens to those values instead.
Heart rate estimates inherit an error you may never have measured
The heart-rate versions work identically with a different denominator: a percentage of HRmax, or a percentage of a threshold heart rate. What matters is what sits under that denominator.
Worked example, fictional rider. An age formula puts HRmax at 185 bpm. A maximal effort actually recorded on the road would have shown 193. An 85% anchor built on the formula returns 157 bpm; built on the recorded figure it returns 164. Seven beats, invisible in the output, because the calculator prints one number and no error bar.
Two recording habits make heart rate worth pairing with lactate values at all.
- Take HR at the same point in every stage, and write down which point. End of stage is the common convention. Heart rate at a fixed power is rarely identical at minute one and minute three, so a stage average and an end-of-stage reading are two different series and should not be mixed.
- Log whether FTP and HRmax were measured or derived, and when. An HRmax from three seasons ago and one recorded last month are different inputs. A calculator cannot tell them apart.
Does your step test resolve the answer at all?
A curve fit cannot recover what the protocol never collected. Two things decide that: how wide the stages are, and how many of them fall on each side of the threshold you care about.
Take a common cycling grid — start 100 W, 30 W increments, 3 minutes per stage, one sample per stage:
100 · 130 · 160 · 190 · 220 · 250 · 280
Finishing the 250 W stage takes 18 minutes of riding plus the pauses for sampling. If LT2 lands near the top of that grid, the fit has five or six points beneath it and is well constrained. If LT1 lands somewhere near 170 W, it has three samples below it, and one of those is the opening stage.
Now the shortcut that quietly ruins the low end: starting at 160 W to save two stages and two strips. LT2 is untouched. LT1 now has one point below it, sometimes none, and the fitted low end is a line drawn through a single measurement and a resting value. The output still prints to the watt.
Count the stages that will fall below your expected LT1 before you pick the starting power. Fewer than two, and the bottom of your curve is an assumption wearing a decimal point.
Narrower stages buy resolution and cost time. A 20 W grid from 120 W reaches 260 W in eight stages and 24 minutes. A 30 W grid from 100 W reaches 280 W in seven stages and 21 minutes, but before interpolation it can only place a threshold to the nearest 30 W. Neither is wrong; they are different trades, and which one you want depends on whether LT1 or LT2 is the number you came for.
Stage duration is a separate lever from stage width. Shorter stages fit more workloads into a session; longer stages give each sample more time at a steady workload before it is taken. That changes what the sample represents, so hold it constant across tests — a 3-minute grid and a 5-minute grid do not produce comparable datasets. Budget one strip per stage, one for the resting sample, and one or two spares for readings you want to repeat.
The full input-to-output path is set out in calculate lactate curve online.
The same dataset, three methods, three answers
Fitting is not a single operation. LT2 gets placed by fixed-value conventions such as a 4 mmol/L crossing, by baseline-plus-offset conventions, and by geometric ones such as Dmax and its modified variants. LT1 is generally placed by locating the first clear departure from the resting level, and implementations define "clear departure" differently. These are not roundings of one another. They are separate definitions that happen to share a label.
One rule follows, and it holds whichever tool you use: run your own dataset through more than one method once, and treat the spread you see as your error bar. Then choose a method and stay on it. If you later switch, re-read the old dataset with the new method before comparing anything — a 12 W gain between two tests means nothing if one was read with Dmax and the other at 4 mmol/L.
Turning fitted thresholds into zone boundaries is a separate step, covered in the lactate threshold calculator for LT1, LT2 and zones.
The autumn comparability trap
Tests taken either side of a season change are the ones most likely to mislead. A test ridden outdoors in warm weather and a test ridden indoors in a cool room in November differ in airflow, riding position, cadence, gradient and power-meter calibration before any physiology enters the comparison. If those two are your baseline and your retest, part of the difference you read is the setup.
For a block that starts on the trainer in October or November, the cheap fix is to make the indoor test the baseline. Run it in the conditions you will retest in, and record room temperature, fan setting, trainer or ergometer model, and the calibration step alongside the data. One line on the sheet.
What to have on the sheet before you open any calculator
- Power per stage and stage duration, as ridden rather than as planned
- Blood lactate per stage in mmol/L, plus one resting sample taken before the warm-up
- Heart rate at a defined point in each stage, with that point named
- Meter model and strip lot — a change in either starts a new series
- Whether FTP and HRmax were measured or estimated, and on what date
- Indoor or outdoor, room temperature, fan, trainer or ergometer model, calibration step
- Any stage cut short or repeated, flagged rather than deleted
All of that is record-keeping and arithmetic. What the resulting numbers should change in your training, and anything touching health, is a conversation for a coach or physician who can see the rest of your situation. The top stages of a step test, like any recorded HRmax, are near-maximal efforts: if you have a known heart condition, get chest pain, dizziness or unusual breathlessness on exertion, or are returning from a long break, get medical clearance before you ride one.
If you are not drawing blood
Non-invasive routes to an LT1-like number exist and get discussed constantly: the talk test, ventilatory landmarks, DFA a1 derived from heart-rate variability. None of them produces the input a lactate calculator consumes. They estimate something adjacent, by a different mechanism, with their own failure modes.
The workable compromise is a filing rule rather than a verdict. Keep each method in its own series, and do not plot them on a single trend line. An LT1 from a talk test and an LT1 from a fitted curve share a label and a rider. That is the whole of what they share.
The number no calculator returns
Run the same protocol twice inside a fortnight, in matched conditions, before you treat any watt figure as a baseline. The gap between those two runs is the floor on the change you can detect afterwards — and it is the one figure no calculator reports, because it does not exist until you have generated it yourself.