Maximum Adaptive Volume (MAV): Why Per-Muscle Training Ceilings Beat Generic Volume Rules

You bench-pressed three times last week. The app said that was fine. Today, four days after the last session, your bench top set felt heavier than usual. You cut the session short. Next week you push through, hit the same three sessions, and by Friday your shoulder is aching.

Your training log shows “15 hard chest sets per week.” A textbook MAV table shows 12 to 20 as the adaptive range. You’re inside the range. You’re also injured.

What went wrong is what always goes wrong with population-averaged prescriptions: the range was right in the average sense and wrong for you specifically.

This post is the deep-dive on Maximum Adaptive Volume as a personalized per-muscle ceiling, which is the prescription layer in the adaptive training intelligence stack.

The Volume Landmarks Framework (and Why It’s a Starting Point, Not an Answer)

The modern strength literature uses four volume landmarks, usually credited to Mike Israetel and the Renaissance Periodization team but rooted in prior work on dose-response curves in hypertrophy and strength training.

• MV — Maintenance Volume. The minimum weekly volume per muscle needed to retain current size and strength. Typically 4 to 6 hard sets per week for intermediate lifters.
• MEV — Minimum Effective Volume. The smallest dose that still produces adaptation in trained lifters. Typically 8 to 10 sets per week.
• MAV — Maximum Adaptive Volume. The volume that produces the best adaptation-per-unit-of-recovery. The sweet spot. Typically 12 to 20 sets per week, depending on muscle group and lifter.
• MRV — Maximum Recoverable Volume. The ceiling above which the body stops recovering week-to-week and overreach sets in. Typically 16 to 25+ sets per week.

These landmarks are genuinely useful as a framework. They replaced the “train until you drop” era with a dose-response model. They gave evidence-based programming a shared vocabulary.

What they didn’t do — what no population-averaged framework can do — is tell an individual lifter their specific numbers. The landmarks are ranges. The ranges were measured on populations that skew young, trained, male, and without major life stressors. Generalizing to a broader lifting population introduces error that the research itself doesn’t quantify.

Three specific failure modes recur in the field.

Individual variance within the range. A “12 to 20 sets for chest MAV” range spans a 66% delta. Two lifters both trained for five years can sit at opposite ends. The same textbook prescription — 16 sets — is right-on-the-dot for one and a recipe for chronic shoulder irritation for the other.

Population bias in the research. Most volume landmark research was conducted on trained young adults. Extrapolating to women, older adults, returning lifters, lifters under high life stress, or lifters running other sports on top is an extrapolation the original papers don’t endorse. In practice the extrapolations work at population scale and fail individually.

Static numbers in dynamic lives. MAV in April — well-rested, light work load — is not the same as MAV in November when a job change, a new baby, or ongoing illness has shifted recovery capacity. A table-driven prescription is blind to that. The lifter either undertrains in April or overtrains in November, or more often both in sequence.

The fix is not to refine the table. It is to learn the number from the lifter’s own history.

What a Personalized MAV Actually Is

A personalized MAV, at its core, is a probability distribution over weekly hard-set counts. The distribution’s width reflects how much evidence the system has about you.

The distribution is built from two sources, combined using Bayesian methods.

The prior. The starting belief before the system has your data. A weak, wide prior centered on the published population median for each muscle group. For chest, that might be centered near 13 to 14 hard sets with a standard deviation wide enough to cover the textbook range on both sides. The prior’s purpose isn’t to be correct — it’s to be less wrong than zero information. A lifter with no history gets the population median as their provisional number, with wide enough uncertainty that the prescription is conservative.

The likelihood. Every week of training logs updates the belief. The update process asks: was this week’s volume tolerated or not? Tolerance is measured against post-training markers — HRV rebound to baseline, sleep quality, subjective fatigue, ability to hit the next session’s top set. A week with good tolerance pulls the estimate up. A week with poor tolerance pulls it down. The magnitude of the pull depends on how decisive the evidence is.

The result — the posterior — is the system’s current best belief about your MAV. It has a central tendency (median) and a width (credible interval). As weeks of evidence accumulate, the width shrinks. A lifter with 20 weeks of clean data has a narrow posterior. A lifter with 6 weeks of messy data (illness, missed sessions, big RPE swings) has a wider one.

The lower confidence bound — the conservative floor of that credible interval — is what gets used as the prescription. Typically the 10th or 25th percentile, depending on how conservative the system is tuned. That choice matters more than it sounds like it should:

• Prescribing at the posterior median is a 50/50 bet. Half your weeks the prescription will be below your true MAV; half above.
• Prescribing at the lower percentile (say, p10) biases the bet. Your body will handle the prescribed volume with high probability.
• The cost of the bias is a small amount of unused adaptive capacity. You might be able to handle 15 sets, and the system prescribes 13. Over a training year, that’s a modest amount of progress left on the table.
• The benefit of the bias is that weeks which land at the prescribed volume almost never break the lifter. Injuries and multi-week overreach come from chronic overshooting, not from occasional undershooting.

If that framing feels familiar, it’s because it’s the same principle that shows up in personalized thresholds versus population norms: use your own distribution, surface the uncertainty, and bias toward trustworthy output. The MAV application is the training-load version.

Per Muscle, Not Per Workout

A subtle but important implementation choice: MAV is estimated per muscle group, not per session.

The reason is recovery economy. Chest and back have different recovery curves. Quads and shoulders have different response dynamics. A week of 18 chest sets plus 14 back sets is not physiologically equivalent to a week of 14 chest plus 18 back, even though the totals are identical.

An adaptive system tracks:

• Chest. Pec-major pressing sets. Near-failure sets of flat/incline/decline bench, dumbbell presses, flyes count.
• Back. Separates vertical pull (lats) from horizontal pull (rhomboids, mid-back). Pullups vs rows load different fibers.
• Legs. Separates quad-dominant (squats, leg press, leg extensions) from hip-dominant (RDLs, hip thrusts, hamstring curls). These have independent MAVs.
• Shoulders. Side delts, rear delts, front delts. Most lifters overvolume the front and undervolume the rear.
• Arms. Biceps and triceps. Overlap with compound pull and push work is counted.

The exact taxonomy varies by platform. What matters is that each distinct muscle group has its own posterior, updated from its own direct and indirect training weeks. When you do a heavy bench session, triceps posterior gets partial credit. When you do a pullup session, biceps get partial credit. The weights are not 1:1, but they’re not 0 either.

The other piece of this: within-week scheduling matters. Sixteen chest sets spread across Monday, Wednesday, and Friday with 48-hour gaps is different from 16 chest sets in one Saturday marathon. A well-designed MAV estimator accounts for this via the session-level recovery half-life model, which we cover in per-muscle recovery half-life.

What “Tolerated” Means, Operationally

The update step needs a definition of “the week was tolerated.” Not every system defines this the same way, and the definition is load-bearing for the whole model.

A reasonable set of tolerance signals:

HRV rebound. After a hard training week, HRV typically dips and then recovers. A week was tolerated if HRV is back to the personal p50 or above by the end of the rest day. A week was not tolerated if HRV stays suppressed across the deload.

Resting heart rate. Elevated RHR is a reliable post-training stress marker. A week was tolerated if RHR is back to baseline after the rest day.

Top-set performance. Did next week’s scheduled top set match the plan? Did the lifter hit reps at the planned load and RPE, or did they fall short?

Subjective fatigue. A one-question “how fatigued are you” scale logged daily. Stable or recovering across the week is a tolerance signal.

Injury and symptom flags. Any new joint pain, tendon irritation, or unusual soreness is a negative tolerance signal even when the other markers look fine.

The signals are combined, not averaged. A single strong negative — a persistent HRV depression or new tendon pain — should outweigh positive performance signals. The asymmetry is deliberate: the cost of a false positive (calling a week tolerated when it wasn’t) is injury risk or multi-week overreach. The cost of a false negative (calling a week not-tolerated when it was) is one slightly lower week of prescription. Bias toward the safer direction.

An implementation note: these signals overlap heavily with readiness score inputs. The difference is timeframe. Readiness looks at yesterday to today. MAV tolerance looks at a whole week to the following rest day.

How the Prescription Gets Delivered

A personalized MAV lands in the lifter’s planning surface as a range, not a single number.

For each muscle on a weekly cadence, the system shows:

• Lower bound (p10 to p25). The prescription floor. This is the volume the system is confident the lifter can recover from, given current posterior and recovery state.
• Median (p50). The “current best guess” of the adaptive sweet spot. Useful as context, not as prescription.
• Upper bound (p75 to p90). The ceiling before MRV risk climbs. Volume above this number has historically coincided with missed lifts, elevated injury flags, or chronic HRV depression.
• Current week-to-date. Real-time tally of hard sets completed on this muscle so far this week.

The lifter sees “Chest: 4 of 13 (11–15)” and understands they’ve done 4 hard sets, the floor is 13, the band is 11 to 15. If they have two more chest sessions planned that will total 10 more sets, the system flags that the plan overshoots the band. The lifter can either accept the flag — maybe they’re peaking intentionally — or adjust.

The same display extends to other signals the lifter benefits from seeing together. ACWR shows as a trend band. Monotony and strain show as within-week structure flags. All three are context for the MAV prescription, not competing prescriptions.

The interaction model: MAV tells you how much per muscle. ACWR tells you how fast the total is rising. Monotony tells you whether the shape of the week is recoverable. The lifter reads all three and plans accordingly.

How MAV Updates With Your Life

A static MAV would be almost as wrong as a static table. The posterior has to track context shifts.

The system widens the posterior — expresses more uncertainty — when:

• Life stress markers shift. A change in sleep patterns, consistent RHR elevation, or a self-reported life-event flag (new job, new baby, travel-heavy month) widens the band because the recovery capacity has likely shifted.
• Nutrition shifts. A sustained caloric deficit or a flagged underfueling pattern should widen the posterior. Low energy availability lengthens recovery, and the MAV prior should shift down until evidence confirms otherwise.
• Training structure changes. Switching programs, shifting frequency, or changing accessory work patterns introduces uncertainty that the prior structure doesn’t cover. Widen, then retighten as evidence arrives.
• Age markers. Over long time horizons (years), MAV trends slowly downward for most lifters over 35, faster for lifters over 50. A well-calibrated system tracks the trend and adjusts the prior rather than pretending it’s stationary.

The system tightens the posterior — expresses more confidence — when:

• Multiple weeks at the prescribed volume are clean. If the prescription has been landing in the green band with good tolerance markers for 6 to 8 consecutive weeks, the system shrinks the credible interval and the lower bound can creep up.
• Life markers are stable. No significant HRV disturbance, no RHR drift, no flagged stressors. Stable context means narrower intervals.

The effect over a long training history: a lifter with two years of consistent logging has very tight MAV posteriors. They know within a couple of sets what their chest, back, and leg ceilings are in a normal month. When a stressful month arrives, the bands widen appropriately. When normal service resumes, they re-tighten. The number tracks reality instead of pretending reality is static.

A Worked Example: Three Months of Chest MAV

Consider a lifter starting from scratch with the system.

Month 1. Prior is centered at population median, 13 sets/week, with wide uncertainty (sigma covering roughly 8 to 18). First week: 12 hard chest sets, good tolerance markers. Second week: 14 sets, good tolerance. Third week: 16 sets, marginal tolerance (HRV took an extra day to rebound). Fourth week: 12 sets (deload).

By end of month 1 the posterior has shifted slightly. Median is around 13. Lower bound (p10) is around 10. Upper bound (p90) is around 16. Sigma is tighter than the prior but still wide — the system has only four data points.

Month 2. Prescription floor bumps to 12. The lifter runs weeks at 13, 14, 13, and 15 sets with good tolerance. The posterior tightens. By end of month 2, median is 14, lower bound 12, upper bound 16.

Month 3. The lifter starts a cut for aesthetic reasons — 250 kcal deficit, steady weight loss. The system notes the caloric-deficit flag and widens the posterior slightly to reflect that recovery capacity may be shifting. Prescription floor drops from 12 to 11 as a precaution. Week by week, tolerance markers stay clean despite the cut — the lifter is eating enough protein, sleeping well, and the deficit is modest. The posterior re-narrows over three weeks. By end of month 3, median is 14, lower bound 12, upper bound 15.

Then in week 12 the lifter catches a cold. One session is skipped, and the following two weeks’ tolerance markers are elevated across the board — RHR up, HRV down, subjective fatigue up. The posterior widens. Prescription floor drops to 10. Week 13’s chest work is limited to 10 sets at moderate RPE. Week 14, tolerance returns to baseline, the lifter feels recovered, and the floor climbs back to 11.

Over three months, the lifter’s MAV moved from a population-median starting point to a tight personalized band with an explicit response to a life event. No table could have produced that. A static rule could not have adjusted to the illness. An auto-regulated program running on RPE alone would have caught some of it but not all — and would have left the lifter guessing at the magnitude of the reduction.

That is the MAV value proposition: a training prescription that is specific to your muscles, your history, and your current state, with explicit uncertainty that tracks the confidence the system actually has.

What MAV Is Not

A few things worth saying plainly to avoid overclaiming.

Not a replacement for a coach. MAV answers “how much per muscle this week.” It does not answer “which exercises,” “what technique,” “how to peak for a meet,” or “is your shoulder pain a cue to change form.” Those are coach questions. The number is a volume ceiling, not a program.

Not a guarantee of zero injury. Training volume is one injury input. Technique, mobility, cumulative tendon load, sleep, nutrition, and simple bad luck also matter. A well-calibrated MAV reduces injury risk compared to either population-averaged prescriptions or no prescription, but does not eliminate it.

Not a substitute for progressive overload. MAV is a ceiling on volume, not a plan for how to push load upward over time. Progressive overload happens within the volume band through per-lift load progression.

Not a magic number. The lower-bound prescription is deliberately conservative. Running above the floor (into the credible interval’s upper region) is sometimes appropriate — a peaking block, a short overreach before a planned deload, a test week. The system should let you do that and flag it, rather than rigidly blocking it.

What to Ask a Platform That Claims to Do This

If a platform says it delivers personalized training volume targets, the questions worth asking:

• Is the number per-muscle or global? A single weekly-volume number is not enough.
• Is the prescription from the lower bound or the median? If from the median, you’re on a 50/50 bet.
• Does the platform show you the credible interval? Or just the point estimate? A number without uncertainty is a guess rendered as a measurement.
• How long before the posterior tightens? If the answer is “immediately,” the platform is using a very strong prior — likely the population median — and calling it personalized.
• Does the prescription update with life events? Nutrition flags, illness, stress, sleep changes?
• What signals count as “tolerance”? If it’s only RPE, the system inherits RPE drift. Objective markers (HRV rebound, RHR return-to-baseline, top-set performance) should also be in the mix.

Most consumer platforms fail at least half of these questions. That is not a reason to dismiss them. It is a reason to know what you’re getting: a trend line, not a model.

---

In Summary

A personalized Maximum Adaptive Volume is:

• Per-muscle, not global
• Bayesian with a weak prior centered on population medians
• Updated weekly from objective tolerance markers, not just self-reported RPE
• Prescribed at the lower bound of the credible interval, not the median
• Responsive to life events (stress, nutrition, illness, age)
• Shown to the lifter as a band, not a single number

This is the prescription layer in the adaptive training stack. It is what replaces generic MEV/MRV tables with a number you can actually train on. Around it sit the recovery-half-life model, the RPE calibration layer, the per-lift progression rates, and the ACWR/monotony/strain context alarms. Together they are the difference between a dashboard and a coach-adjacent tool.

For more: per-muscle recovery half-lives, RPE calibration, ACWR context. Omnio’s adaptive training feature lives at /features/adaptive-training.