When VBT Velocity Jumps After Caffeine: Reading It Right

An athlete on my roster texted me a screenshot last month: his squat load-velocity profile from that morning's session had spit out a projected 1RM of 212kg, twelve kilograms above anything he'd tested in the previous three months. He'd taken a scoop of a new pre-workout about forty minutes before warming up. Nothing else about his training had changed - same accessory work, same sleep, same body weight on the scale that morning. He wanted to know if he should go test the new number that weekend.

This message shows up in my inbox more often now that pre-workout stacks routinely carry 300-400mg of caffeine and velocity-based training retests have become a normal Tuesday-morning habit. The problem is real, and it's not the athlete misreading his app. Bar speed genuinely increases under caffeine at most training loads, and if your estimated 1RM comes from extrapolating a load-velocity regression line built out of those same loads, that estimate climbs right along with the bar speed - even when the athlete's actual strength ceiling hasn't moved more than a kilogram or two.

Mixing the two up is where programs go sideways: a coach loads the next block off an e1RM that's mostly pharmacology, the athlete misses a heavy single he 'should' have hit on paper, and there's a confidence problem stacked on top of a programming problem. This piece covers what caffeine actually does to bar speed at different relative loads, what two controlled trials found when they isolated the effect by load zone, and the checks I run before a caffeinated session touches anyone's training numbers.

Caffeine's main mechanism relevant to barbell velocity is adenosine receptor antagonism in the central nervous system. Adenosine accumulates during effort and dampens the drive to recruit and fire motor units at high rates; blocking those receptors lets an athlete recruit more of the motor units available to them, fire those units faster, and tolerate more discomfort at a given intensity than they would unstimulated. None of that changes how much contractile tissue is available or how much force a maximally-activated muscle can produce - it changes how close to that existing ceiling the nervous system is willing to push on a given rep.

That distinction matters for how the effect shows up across a load-velocity spectrum. At lighter and moderate loads - the 40-75% 1RM range most LV profiles sample - bar speed is limited mostly by rate of force development and voluntary drive, exactly what caffeine improves. Near a true 1RM, velocity is limited by how much maximal force the muscle can generate at all, a ceiling caffeine doesn't touch acutely. The result is a velocity increase that's real but uneven: bigger at the loads your profile uses to build its regression line, smaller at the loads closest to what you're actually trying to estimate.

Because most load-velocity profiling protocols extrapolate a regression line out to a minimum velocity threshold rather than testing an athlete all the way to true failure every session, an upward shift concentrated in the submaximal points drags the extrapolated intercept - your estimated 1RM - up by more than the athlete's real capacity moved. A profile reading +6% velocity at 60% and 75% loads but only +1-2% at a genuine near-max attempt isn't describing a stronger athlete. It's a regression line tilted by a stimulant.

That gradient is the fingerprint you're looking for. A flat percentage increase across every load zone, including a genuine heavy single, looks more like a real neuromuscular shift. A gradient that's steep at light loads and nearly flat near true max looks like caffeine.

Pallares and colleagues (2013) ran a dose-response trial in resistance-trained men, testing placebo against 3mg/kg, 6mg/kg, and 9mg/kg of caffeine before bench press and full squat sessions performed across a range of loads with a linear position transducer on the bar. Both 3mg/kg and 6mg/kg produced significant increases in mean velocity and power compared with placebo, concentrated at the submaximal loads tested; 9mg/kg added no further performance benefit over 6mg/kg but did add more side effects - elevated heart rate, jitteriness, occasional mild gastrointestinal distress. The takeaway that's held up well since: more caffeine past roughly 6mg/kg buys you side effects, not more bar speed. The limitation worth flagging is that this was a single-session acute design in trained males only, so it says nothing about whether daily caffeine use blunts the effect over time.

Grgic and Mikulic (2017) ran a placebo-controlled crossover using 6mg/kg caffeine before Smith-machine bench press, measuring mean velocity and power across a spread of loads from roughly 25% to 85% 1RM, then testing an actual 1RM attempt in the same session. Mean velocity at the submaximal loads increased with a small-to-moderate effect size, but the directly tested 1RM increased only marginally - a kilogram or two, not consistently significant across their sample. That gap between the submaximal-load velocity gain and the directly-tested maximal-strength gain is exactly the mechanism behind an inflated LV-profile 1RM: the profile 'sees' the submaximal jump and extrapolates it all the way to the top, while a real 1RM attempt shows the ceiling barely moved. The limitation: one exercise, a Smith-machine bench press, in habitual caffeine users, whose blunted baseline sensitivity may understate how large this gap gets in someone who rarely uses caffeine at all.

One more study worth knowing for context: Mora-Rodriguez and colleagues (2012) found caffeine's velocity benefit was noticeably larger in morning sessions than evening ones, essentially reversing part of the normal circadian dip in neuromuscular output. A lifter who normally trains in the evening but tests their LV profile at 7am with a pre-workout in their system is stacking two velocity-boosting factors, not one - which can make an apparent jump look larger than caffeine alone would produce.

Here's the sequence I run before any caffeinated session gets to touch a training number.

Rule 1 - Log the dose and the clock, every time. 'Had a pre-workout' isn't a data point. Write down the milligrams (check the label - many stacks run 300-400mg per scoop) and how many minutes before the first working set it went down. Plasma levels peak around 45-60 minutes post-ingestion and the performance window runs roughly 3-6 hours, so timing matters as much as dose.

Rule 2 - Check the gradient, not just the estimated 1RM. Pull up velocity at your lightest tested load and your heaviest tested load separately. If the light-load number jumped 7% and the heavy-load number jumped 1%, that's the caffeine gradient from the table above, not a strength gain worth programming around.

Rule 3 - Never blend a caffeinated data point into a profile built caffeine-free. If the last four sessions in your regression were fasted or habitual-dose and this one wasn't, pull it out before recalculating the line. Mixing conditions is one of the most common ways a profile drifts for reasons unrelated to training.

Rule 4 - Retest clean within 48 hours before you act on a jump. If a session throws an e1RM meaningfully above trend, don't reprogram off it. Run the same protocol again within a day or two under matched, normal conditions. A caffeine-driven number won't repeat; a real adaptation will.

Rule 5 - Weight the effect down for daily users. Habitual caffeine consumers develop measurable tolerance, and the acute velocity bump shrinks - often by a third or more - within one to two weeks of consistent daily intake. It's a smaller confound for a two-coffees-a-day athlete than for someone caffeine-naive, but rarely zero. When in doubt, standardize: same caffeine status on every profiling day, full stop.

For the mechanics of pulling a number off the regression line, this guide to calculating 1RM from velocity walks through the math; for swings that don't line up with any caffeine or timing explanation, our piece on why most lifters overestimate their 1RM covers the other common sources of drift.

The athlete from the opening of this piece trains squat with a three-point LV profile - 60%, 75%, and 85% of his most recent tested 1RM - plus, on test days, one genuine heavy single to confirm the estimate against reality. His baseline 1RM going into the block was 180kg.

Week 1, no caffeine, fasted morning session: 108kg moved at 0.92 m/s, 135kg at 0.68 m/s, 153kg at 0.47 m/s. The regression extrapolated to an e1RM of 181kg - close enough to his tested number to trust the line.

Week 5, same protocol, but he'd had roughly 300mg of caffeine (about 3.5mg/kg at his body weight) forty minutes before the first working set: 108kg moved at 0.99 m/s, 135kg at 0.73 m/s, 153kg at 0.50 m/s - a 6-8% velocity increase across all three loads. Run through the same regression, that spit out an e1RM of 196kg, a 16kg jump in five weeks. He also attempted a heavy single that day at 190kg, above his old tested max, expecting the profile to hold. The bar moved at 0.15 m/s - a hard, grinding rep that only barely locked out, nowhere near what a genuinely 196kg-capable lifter's bar speed on a 190kg single should look like.

Three days later, same protocol, no caffeine: 108kg at 0.93 m/s, 135kg at 0.69 m/s, 153kg at 0.48 m/s - essentially back to his week-1 numbers, regression landing at 183kg. That's a genuine 2-3kg improvement over five weeks of real training - nowhere near the 16kg the caffeinated session implied. We built his next block's top sets around 183-185kg, not 196kg, and he hit every prescribed rep without a single grinding single.

Programmed off the week-5 number instead, his top sets would have opened at loads he genuinely wasn't ready to move - missed reps, ballooning RPE, an athlete who starts distrusting the whole velocity-tracking process because the numbers 'lied.' They didn't lie. They measured exactly what was in front of them - a bar moving faster because of a stimulant, not a stronger athlete underneath it.