AMRAP pacing: what the research supports

Pacing is anticipatory, not reactive

The idea is called teleoanticipation: work rate during heavy exercise is regulated by a feedback system that takes the finishing point into account, rather than simply responding to how bad things feel right now.¹

You already know this is true from experience. You do not run the last minute of a twenty-minute AMRAP the way you ran minute twelve, and nothing physiological changed at minute nineteen to permit it. What changed is that the end became close enough to spend against.

The direct evidence comes from lying to people about the clock. When researchers manipulated the time feedback athletes could see, without their knowledge, pacing strategy changed.⁶ If effort were governed purely by the state of the muscle, a false clock would make no difference. It does.

The central governor is a model, and it is disputed

The best-known account built on teleoanticipation is the central governor model, which proposes the brain regulates performance by modulating how much muscle it recruits, using the sensation of fatigue to keep you inside your physiological capacity.²

It is also a model plenty of physiologists want retired. A 2009 rebuttal in the same journal family went through the supporting claims and argued the case had not been made.³ The argument has continued since.

Take from that what is safe to take: effort in a self-paced workout is regulated centrally and in anticipation. Whether there is a discrete "governor" doing it is a live question, and any article that hands you the mechanism without mentioning the argument is selling a tidier story than exists.

Is even pacing actually best?

This is where I have to be careful, because the answer is less convenient than the coaching consensus.

The standard reference defines the pacing strategies — all-out, positive, even, negative, parabolic, variable — and reports that very short events (30 seconds or less) suit an all-out approach, while events beyond about two minutes tend to produce better times when the pace is distributed more evenly.⁴ That is the basis for most pacing advice you will read, including most of mine.

But a 2025 systematic review and network meta-analysis comparing imposed pacing strategies found no universally superior one, reported that comparative effects remain unclear because of conflicting findings and methodological variability, and concluded that self-pacing has real value as a flexible, context-adaptive strategy.⁵

So the honest position is narrower than "even pacing wins": pacing distribution matters, an even split is a reasonable default for anything past a couple of minutes, and imposing a strategy on an athlete is not reliably better than letting a trained one self-pace. If you have been told the science settles this, it does not.

Where our Pace Variance Index sits

AMRAP With Friends computes a number we call the Pace Variance Index from your round times:

(slowest round − fastest round) ÷ average round × 100

Except on the shortest missions — five minutes and under — the first round is excluded, for the reason above: an opening round run with full reserves and no feedback would otherwise mark honest pacing as a collapse.¹ On a five-minute clock every round counts, because with only a handful of rounds in the workout, dropping one discards too much of the evidence.

Why pacing matters more in an AMRAP

Two features of the format make the pacing decision unusually costly. There is no prescribed rest, so every recovery you take is paid for out of your own score. And the endpoint is fixed in time rather than in work, so unlike a race you cannot finish early by going faster — you can only do more.

The instrumentation says the intensity is high throughout and does not discriminate much. In a direct comparison, roughly 4-minute and roughly 17-minute sessions both sat near 90% of maximum heart rate with no significant difference between them.⁷ In a 20-minute AMRAP, mean heart rate was around 171 bpm while oxygen uptake was only about 64% of maximum.⁸

Both of those say the same thing for the athlete: your heart rate will be high whatever you do, so it cannot tell you whether your pacing was any good. Round times can. That is the whole argument for measuring them, and it is covered further onenergy systems and the time cap.

What the research does not show

• That any one pacing strategy is best. The most recent synthesis says the opposite — conflicting findings, unclear comparative effects.⁵
• That the central governor model is correct.It is a contested model, not an established mechanism.³
• That our variance bands predict anything.They are unvalidated coaching thresholds. Said again here because it is the single easiest thing on this page to misread.
• That pacing research on cyclists and runners transfers cleanly to AMRAP.Most of this literature is continuous single-mode exercise. An AMRAP is multi-movement, self-scaled and interrupted by transitions. The principles should carry; the specifics have not been tested here.

Related questions

Measure your own splits

Whatever the literature settles on, the useful move is the same: log every round so you can see the distribution instead of guessing at it. Type your round times below and this runs the same functions the app runs — including the band thresholds this page has just told you are unvalidated.

The practical version of this page isthe pacing guide.