Rising Velocity Mid-Block: Real Gain or Artifact?

A coach on our beta list sent over a load-velocity export last month: back squat at a fixed 100kg had been climbing steadily for six weeks - 0.58 m/s in week 1, 0.65 m/s by week 6, a 12% rise at the exact same bar weight. His question was blunt - does he bump the training max for the next block, or is something else going on? The athlete hadn't tested a new 1RM in that window, hadn't dropped body weight, hadn't changed anything he could point to except showing up and doing the work.

That question comes up constantly once a team starts tracking velocity across a block instead of just on testing day. The honest answer is that a rising number at a fixed load can mean three completely different things, and only one of them is the strength gain everyone hopes for. It can be a real increase in force-generating capacity. It can be the athlete simply getting better at the movement - tighter bracing, a cleaner bar path, less wasted motion out of the hole - which raises velocity without raising the strength ceiling underneath it. Or it can be nothing more than the day-to-day noise every velocity measurement carries.

Reprogram off the wrong one and you either hold back an athlete who's genuinely stronger, or load someone into weights their nervous system isn't ready to move yet. This piece covers what the research says about each source, and the checklist I run before a rising trend touches anyone's next block.

Start with the one everyone assumes is happening - real neuromuscular adaptation. When an athlete's maximum voluntary force output goes up, a fixed absolute load - say that 100kg squat - now represents a smaller percentage of their new 1RM, and velocity at that load rises accordingly. This is the mechanism behind every load-velocity profile ever built, and it's genuine. The catch: it isn't the only mechanism producing the same signature on a velocity graph.

The second driver is technical efficiency, and it gets underweighted constantly. A lifter who tightens their brace, shortens the bar path, or improves rebound timing out of the bottom will move a submaximal load faster - sometimes 5-10% faster - without any change in how much force their muscles can generate. This shows up hardest in athletes newer to a lift, anyone who just had a cue corrected, or anyone returning to a movement after a layoff. The bar speed is real. The strength gain it implies is not, or not to the degree the number suggests.

The third driver is the one coaches trust least and should probably trust more - measurement variability. Bar speed at a given load varies session to session even when nothing about the athlete has changed - sleep, warm-up length, time of day, a bar loaded a kilogram light because a bumper plate rounds down. None of that is device malfunction. It's the normal noise band every velocity tool carries, large enough that one session's reading above the trendline isn't, on its own, evidence of anything.

The overlap here is exactly the problem. A 6% rise could be any one of the three, or a mix of all three at once, and the number alone won't tell you which.

Pareja-Blanco and colleagues (2017) tracked resistance-trained men through an eight-week squat block, comparing groups training with either a 15% or a 30% within-set velocity-loss threshold. Both groups saw mean velocity at fixed loads rise significantly - but the lower-velocity-loss group's gain tracked far more closely with a retested 1RM (effect size around 0.9) than the higher-loss group, whose velocity numbers moved further than their tested max did. The authors' own framing: velocity change at a fixed load is a reasonable proxy for strength change only when the stimulus hasn't also altered technical execution or fatigue in ways that move bar speed on their own. Limitation - squat only, in already-trained men, over one block length, so a less-trained population could look very different.

Courel-Ibanez and colleagues (2019) put the measurement-noise question to a direct test, comparing five velocity-measuring technologies - linear position transducers, an IMU sensor, and several camera and app-based systems - across repeated sessions on the same lifters and loads. Repeatability varied by device and load - the most consistent tools held a coefficient of variation around 3-5% at moderate loads, while camera and app-based systems drifted toward 8-9% below 50% 1RM, where bar acceleration is fastest and harder to sample cleanly. Their conclusion was blunt - a velocity change smaller than a device's own established CV at that load zone isn't a finding, it's the tool's noise floor. Limitation - a lab-controlled comparison run by technicians, so field conditions with athletes self-administering warm-ups likely push real-world variability higher still.

Banyard, Nosaka, and Haff (2017) add the technical-efficiency piece almost as a side finding - in their back squat load-velocity reliability study, the very first testing session an athlete completed under a new protocol consistently produced lower velocities at matched loads than every session after it, even with no training in between. The authors attributed this to athletes getting more comfortable with the setup and pacing of the protocol, and the gap closed almost entirely by the second or third exposure - a reminder that any trend built from an athlete's first few sessions on a new lift, rack, or device will show an early rise that has nothing to do with strength.

Here's the sequence I run any time a fixed-load velocity trend climbs enough to tempt a program change.

Check the whole profile, not one point. Pull velocity at every load you've tested that block, not just the one you happen to track daily. A real strength gain shifts the entire load-velocity line - light, moderate, and heavy loads all move up roughly together. A rise isolated to one load, especially the load an athlete has practiced most often, points toward technique rather than a raised ceiling.

Require a sustained trend, not a good week. Given the 3-9% noise band above, one strong session proves nothing. Look for the rise to hold across at least three sessions with variance settling down, not widening. A trend still bouncing 5-6% session to session hasn't separated itself from noise.

Watch the technical markers alongside the number. If bar-path deviation is shrinking and rep-to-rep velocity variance within a set is tightening at the same time the average is rising, that's the fingerprint of technical efficiency, not raw strength. If those technical markers were already stable and stay stable while velocity keeps climbing, that's a better sign of something changing underneath.

Control the obvious confounds before you trust anything. Same time of day, same warm-up length, no new stimulant in the mix, plates actually weighed rather than assumed - a rounded 20kg bumper plate is sometimes 19.3kg, and that alone can produce a velocity change that looks like a training effect. It sounds pedantic until it's the actual explanation, which happens more often than anyone expects.

Confirm with a direct test before you reprogram. When a trend clears the first four checks, schedule a near-max attempt rather than extrapolating off the submaximal trend line alone. A genuine strength gain moves a heavy single at a speed consistent with the new implied capacity. A single that grinds nowhere near that speed means technique or noise did more of the work than adaptation did.

For the mechanics of turning a velocity reading into an estimated max, this guide to calculating 1RM from velocity covers the math; for the flip side of this problem, our piece on why most lifters overestimate their 1RM covers other common sources of an inflated number.

The coach from the opening of this piece agreed to let us walk through his athlete's data in detail. The lifter, a 26-year-old with about four years of training history, had his squat setup corrected mid-block - the coach cued a tighter brace and straighter bar path after noticing the bar drift forward out of the hole on video. Velocity at a fixed 100kg was tracked twice weekly across the eight-week block that followed.

Weeks 1 through 3 showed the steepest climb - 0.58 m/s up to 0.63 m/s, right alongside a drop in bar-path deviation from 4.1cm to 2.4cm as the new bracing cue took hold. That's the technical-efficiency signature almost exactly as described above - fast improvement, tightly correlated with a technical marker, front-loaded early in the exposure to the corrected cue.

Weeks 4 through 8 told a different story. Bar-path deviation held flat around 2.2-2.5cm - the technique fix had stuck and stopped improving - yet velocity kept climbing, from 0.63 m/s to 0.67 m/s. With the technical variable held steady, that remaining rise had nowhere else to come from except an actual increase in force capacity. A heavy single in week 8 at 148kg, six kilograms above his pre-block max, moved at 0.17 m/s - within the expected range for a genuine near-max attempt, not the grinding speed you'd see if the number were inflated by noise alone.

Split out that way, the 16% total rise breaks into roughly two pieces - half from cleaning up the bar path in the first three weeks, half from genuine strength adaptation once technique had already stabilized. Programmed as if the whole 16% were strength, the next block's top sets would have opened light for week 4 and overshot by week 8. Splitting the two components let the coach load weeks 4 through 8 more aggressively than the raw trendline alone would have justified.