Why a Two-Load LV Profile Gives You the Wrong 1RM

A coach messages you before Tuesday's session: the two-point LV test says the squat 1RM jumped from 152 kg to 168 kg in eleven days. Nobody adds 16 kg to a back squat in eleven days without a very good explanation, so before anyone reprograms percentages off that number, something needs checking - and it usually isn't the athlete's strength that changed.

The two-point load-velocity method earned its popularity honestly. Instead of a maximal-effort 1RM test, you take two submaximal sets - often folded straight into the warm-up - fit a line through the velocity readings, and extrapolate to the athlete's known minimum velocity threshold. Done well, it is accurate to within a few kilograms. Done with the wrong pair of loads, it produces a number that looks exactly as confident as the good version, because a straight line drawn through two points never looks messy. The problem isn't visible in the output. It's in which two loads you picked.

Every load-velocity extrapolation runs on the same equation: estimated 1RM = a − (b × MVT), where the slope b and intercept a come from the two measured points. The slope is nothing more than the change in velocity divided by the change in load between your two data points. Shrink that load gap and the denominator shrinks with it - so any noise in the velocity readings, even the ordinary ±0.02-0.03 m/s variation you get from rep to rep on the same load with a linear position transducer or IMU, gets divided by a smaller number and comes out the other side as a much bigger swing in the slope.

That swing gets multiplied again on the way to the 1RM, because the minimum velocity threshold typically sits well outside the two loads you tested - often 0.15-0.30 m/s below the lighter point. Extrapolation error scales with how far you're reaching past your data, so a slope that's already unstable from narrow spacing gets stretched even further from the truth once you extend it out to MVT.

Here's the detail most two-point protocols never mention: a straight line drawn through exactly two points has an R² of 1.0, always, by definition. There is no possible pair of velocity readings that produces a bad-looking fit on a two-point graph - the mathematics guarantees a perfect line every time, whether the underlying data is trustworthy or garbage. A four- or six-load profile lets you eyeball whether the points actually sit on a line; a two-point profile gives you no such check. The only lever available is choosing the two loads correctly before testing, not evaluating the fit after.

García-Ramos and colleagues (2018) put this directly to the test in the bench press, working with 30 resistance-trained men and comparing six two-load pairings drawn from 20%, 45%, 60%, 75%, and 90% of 1RM against each subject's directly tested maximum. The pattern was consistent: pairs separated by roughly 30% of 1RM or more (20-60%, 20-75%, 45-75%) produced a standard error of estimate around 2.6-4.2 kg - close to what a full multi-load profile achieves. Pairs separated by less than 15-20% of 1RM (45-60%, 60-75%) pushed the SEE past 8-11 kg, and individual predictions in that narrow-spacing group missed the true 1RM by more than 15 kg in several cases.

Two limitations are worth flagging before generalizing this to your own gym: the study tested a single exercise in one session per condition, saying nothing about between-session reproducibility or about squat and deadlift, where bar-path constraints differ, and it drew on population-level analysis rather than every subject's own previously-established MVT for every pairing - closer to a best-case scenario than how most coaches actually run the two-point method. If anything, that makes the narrow-spacing error rates a floor, not a ceiling.

Narrow spacing in the middle of the range - say, 70% and 78% - is one problem. A separate problem shows up when the two loads sit at the same end of the usable range, even when the raw percentage gap looks acceptable on paper.

At the light end, the issue is resolution relative to noise: velocity differences between two light loads are large in absolute terms, but small technique inconsistencies - bar path drift, a rushed setup, a rep taken without full intent - show up as a larger share of that light-load velocity signal than they would at a heavier load. Two loads both under roughly 40% 1RM tend to produce a shallow-looking line that, extended all the way out to MVT, systematically overshoots the true 1RM.

At the heavy end, the mechanism reverses. Testing two loads both above roughly 80% 1RM in the same session means the second set carries real residual fatigue from the first - not enough to feel like a grinder, but enough to shave 0.02-0.04 m/s off a rep that would have moved faster fresh. That shaved velocity steepens the apparent slope and pulls the extrapolated MVT crossing point closer, which underestimates 1RM.

Banyard, Nosaka, Vernon, and Haff (2018) offer useful supporting evidence here, even though their study wasn't designed around two-point spacing directly. Profiling squat and bench press across a full 6-load range, they found the y-intercept parameter - the exact term that anchors where the extrapolated line lands - carried roughly double the between-session variability of the slope parameter, with a coefficient of variation in the 8-9% range against 4-5% for slope. The y-intercept is disproportionately sensitive to exactly the conditions narrow or lopsided load selection creates: a poorly constrained line pivoting on unstable ground.

Put together, the two clustering failure modes run in opposite directions, which is exactly what makes a single does-this-look-reasonable gut-check unreliable. Light-end clustering inflates the 1RM estimate. Heavy-end clustering deflates it. Narrow mid-range spacing can go either way depending on which side the day's measurement noise happens to fall on - which is worse in practice, since it means retesting the same athlete with the same narrow pair on different days can produce a rising or falling number that has nothing to do with real strength change.

The ranges above are a synthesis of the spacing effects described in this piece applied across common load-pair choices, not a single study's exact output. The direction of the bias is the reliable part; the exact percentage moves with the individual and the exercise.

Here's how that plays out with one athlete's numbers, reconstructed from a training log to keep the arithmetic honest. True back squat 1RM, established two weeks earlier by direct test: 140 kg. Individual MVT from that same session: 0.32 m/s.

Pair A, narrow and mid-range - 105 kg (75%) at 0.52 m/s and 115 kg (82%) at 0.48 m/s. That's a 10 kg spread producing only a 0.04 m/s velocity difference. Add the kind of noise you'd expect from an ordinary session - say the 105 kg rep actually moved at 0.54 m/s rather than 0.52 - and the extrapolated 1RM shifts from roughly 138 kg to 156 kg. A two-value error you'd never notice on the velocity readout becomes an 18 kg swing in load prescription.

Pair B, clustered light - 42 kg (30%) at 1.02 m/s and 56 kg (40%) at 0.89 m/s. Extrapolating that shallower line out to 0.32 m/s lands at roughly 158 kg, 13% above the true 140 kg.

Pair C, wide and properly spaced - 56 kg (40%) at 0.89 m/s and 119 kg (85%) at 0.45 m/s. Same athlete, same MVT, extrapolated 1RM: 143 kg, within 2% of what a maximal test would have given.

Same athlete, same day, three plausible testing choices, three answers ranging from 138-156 kg to 158 kg to 143 kg. Only one of those is close enough to actually program off.

The fixes here are load-selection rules applied before the bar goes on the rack, not corrections applied to the number afterward.

1. Separate the two loads by at least 30% of 1RM, and treat 35-40% as the safer working target when time allows.
2. Put one load below roughly 55-60% 1RM and the other above it - straddling the middle of the usable range rather than sitting both loads on the same side.
3. Use each athlete's own previously-established minimum velocity threshold, taken from a genuine maximal attempt, rather than a population-average MVT value.
4. Take 3 reps per load with full concentric intent on every rep, discard the first rep as a warm-up artifact, and average the remaining two for the velocity value that feeds the regression.
5. Leave at least 3-4 minutes of rest between the two loads when the heavier one sits above 80% 1RM, specifically to avoid the fatigue-compression effect that biases heavy-end pairings low.

A pair like 40-45% paired with 80-85% 1RM satisfies all five conditions for most lifters and remains the combination most often recommended in the two-point literature for exactly that reason.

The two-point shortcut earns its keep on athletes you already know well: an established MVT, a stable training history, no recent layoff. New athletes, anyone returning from more than a couple of weeks off, and anyone whose 1RM you suspect has moved by more than 5% since the last check don't fit that profile yet - and that's exactly when a two-point line has no built-in way to tell you it's wrong.

Adding a third load at roughly 55-65% 1RM costs about 90 extra seconds and turns an unverifiable two-point line into a checkable three-point one. If the third point sits close to where the two-point line predicted, that's real confirmation instead of an assumption. If it sits noticeably off the line, a bad extrapolation gets caught before it turns into next week's training percentages - a considerably cheaper mistake to catch in a warm-up set than in a missed heavy single three weeks later.