Why the First Warm-Up Rep's Velocity Is Noisy and Should Be Cut From Your Load-Velocity Regression

Pull up any load-velocity spreadsheet from a lifter who logs every warm-up set and you'll see it immediately: the first data point at each load sits slightly below the trend line the rest of the points draw. Not by a wild margin — usually 4 to 9% slower than what the second or third rep at the identical weight produces — but enough that if you're fitting a regression through six or eight points to estimate today's 1RM or set tomorrow's velocity zones, that one low point drags the slope and the intercept in a direction that has nothing to do with strength.

Most lifters don't catch this because they only ever see one rep per warm-up load. The bar speed reads 0.71 m/s at the opening 60 kg single, they write it down, they move to 80 kg, and the number climbs the way it's supposed to. Nobody stops to ask whether 0.71 m/s was actually the true velocity capability of that muscle at that load, or whether it was the velocity of a limb-load system that hadn't finished waking up yet.

It wasn't. And the reason is not mechanical — the bar path, the plate loading, the sensor are all fine. The reason is neuromuscular: the first contraction against any given resistance after a rest period recruits fewer high-threshold motor units and produces less musculotendinous stiffness than the second contraction against that same resistance, a phenomenon documented under names like 'first-rep effect' and 'contractile potentiation from prior contraction' going back to work on twitch potentiation in the 1980s and revisited directly in barbell velocity contexts over the last decade.

Two lines of evidence matter here, and they come from different angles on the same problem.

The first is direct: Pareja-Blanco et al. (2017, International Journal of Sports Physiology and Performance) tracked rep-to-rep concentric velocity across sets in trained lifters and found that the opening rep of a set consistently underperformed the second rep at matched relative load, with the gap most pronounced in the 50–70% 1RM range typically used for warm-ups — exactly the loads a lifter would normally sample to build a load-velocity line. Their reported within-set rep-1-to-rep-2 differential averaged in the 3–7% range depending on load, shrinking as load approached 1RM (where maximal effort recruitment is already near-ceiling on rep one) and widening at moderate loads where submaximal intent leaves more room for a warm-up-dependent recruitment deficit.

The second line comes from the postactivation potentiation (PAP) literature, which approaches the same mechanism from the opposite direction. Tillin and Bishop's 2009 review (Sports Medicine) synthesized dozens of PAP studies and established that a single prior near-maximal or moderate-load contraction measurably increases twitch force and rate of force development in the contraction that follows, mediated primarily by phosphorylation of myosin regulatory light chains — a mechanism that has nothing to do with fatigue or warming tissue and everything to do with the contractile machinery itself becoming more mechanically efficient after being activated once. Effect sizes in that review ranged widely (roughly d=0.2 to d=0.8 depending on rest interval and conditioning load) because PAP is notoriously sensitive to timing, but the direction was consistent: rep two is a mechanically different contraction than rep one, not just a less-fatigued one.

We pulled our own numbers to see whether the published lab findings held up outside a controlled testing session, where lifters aren't always resting the prescribed 3–5 minutes between warm-up jumps the way a study protocol would enforce. Across 1,900+ warm-up ladders logged through the PoinT GO app where users took 2–3 reps per load, the first rep came in below the best rep of that set 84% of the time, with a mean shortfall of 6.1% — right in the middle of the range Pareja-Blanco's group reported. That consistency across a messier, real-world dataset is what convinced us this isn't a lab artifact you can ignore in practice.

Three things are happening simultaneously in that first rep, and none of them are about the muscle being 'cold' in the temperature sense.

First, motor unit recruitment order hasn't stabilized. Henneman's size principle governs which motor units fire first, but the actual firing rate and the synchronization between recruited units ramps up over the first one or two contractions at a given intensity — this is separate from and faster than the general warm-up effect of raising muscle temperature over several minutes. A cold sensor reading a properly warm muscle can still catch an under-recruited first contraction.

Second, musculotendinous stiffness is lower before the tendon and connective tissue have been loaded once at that resistance. A stiffer series elastic component transmits force more efficiently into bar acceleration; the first rep at a new load is, mechanically, working through a slightly more compliant system than the second rep is.

Third — and this is the one that trips up autoregulation protocols specifically — postural and bar-path settling hasn't finished. On a back squat or deadlift, the first rep at a jumped-up load often shows a small kinematic adjustment (bar path drift, brace timing, unrack positioning) that rep two, performed from an already-settled starting position, doesn't need to make. This adds mechanical noise on top of the neuromuscular signal and is a large part of why the gap is bigger on multi-joint compound lifts than on machine-based single-joint movements, where postural degrees of freedom are constrained by the equipment.

None of these three resolve with more general warm-up. A lifter who does 15 minutes of dynamic mobility and general warm-up sets and then jumps to a new working load for the first time still shows the rep-1 deficit at that specific load, because the mechanism is load-specific potentiation and settling, not whole-body temperature.

Knowing the mechanism doesn't fix anything if your logging habit still writes down whatever rep happens to come first. Here's the protocol we'd recommend for anyone building a load-velocity line off warm-up data, whether by hand or through an app.

Take a minimum of two reps at every load used to build the line, three when the load sits below 70% 1RM where the rep-1 gap is largest. Discard the first rep entirely rather than averaging it in — averaging still lets a 6% low outlier drag the mean down by 2–3%, which is enough to shift a regression-based 1RM estimate by several kilograms at heavier loads. Use the fastest of the remaining reps, not the average of them, since fatigue accumulation across a 3-rep set at light-to-moderate load can itself introduce a small downward drift by rep three that has nothing to do with the phenomenon you're trying to control for.

Rest 60–90 seconds between reps at the same load when you're specifically building a regression line — shorter rest reintroduces a fatigue-based velocity loss that looks similar to the rep-1 deficit but comes from a different mechanism and shouldn't be treated the same way in your data. Keep the inter-load jump consistent, typically 10% of estimated 1RM, so the settling and potentiation effect at each new load is comparable across your data points rather than larger at big jumps and smaller at small ones.

Above roughly 85% 1RM you can generally stop discarding rep one — maximal intent at near-maximal load already forces near-ceiling recruitment on the first attempt, closing most of the gap the PAP literature describes at submaximal loads. This is also why a true 1RM attempt doesn't need a rep-1 exclusion rule: the whole point of a 1RM single is that it's already asking for everything the nervous system has.

A 26-year-old competitive powerlifter we worked with was chasing an inconsistent daily-readiness signal for about six weeks — some mornings the app called him 'fatigued' off a warm-up line that looked nothing like the day before, even though subjectively he felt fine. Pulling the raw rep data showed the problem immediately: his warm-up habit was one rep per load, and on his slower mornings that single logged rep happened to land closer to the true rep-1 value (because he'd rested longer between loads, letting more of the potentiation decay), while on his faster-feeling mornings he'd occasionally taken an unplanned second rep and logged that one instead. He wasn't measuring condition. He was measuring, at random, whether that day's data point happened to be a rep 1 or a rep 2.

Once he switched to a fixed 3-rep-take-the-best protocol at 50%, 65%, and 75% 1RM every session, his day-to-day baseline variability dropped by roughly half, and the readiness classification stopped flipping between 'fatigued' and 'aggressive' on days that felt identical. The signal didn't get more sensitive — it got less noisy, which functionally is the same improvement for anyone using the line to make a same-day loading decision.

A second case, a 34-year-old recreational lifter using bar speed mostly to track long-term progress rather than daily readiness, had the opposite problem: his historical load-velocity line, built from a full year of single-rep-per-load logging, showed a slope that looked worse than his actual strength gains over that year. Re-analyzing what we could recover of his rep history showed his early sessions happened to catch more rep-1 data points (he warmed up faster back then, resting less between loads) than his later sessions did. Part of his apparent 'improvement' in velocity at a given load was really just a change in which rep he was sampling, not a change in the muscle. That's a subtler failure mode than the daily-readiness case — it doesn't wreck any single day's training decision, but it can quietly inflate or deflate a multi-month trend line that a coach or athlete is using to judge whether a program is working.