How 330 Arvo Users Train: 82,000 Sets Analyzed (2026)

Methodology

Data was queried directly from the Arvo production database on 2026-05-05. The window spans 2025-11-25 → 2026-05-05 (23 weeks). A "valid set" is a logged set with `skipped = false`, positive `weight_actual`, and `reps_actual` between 1 and 30 (cardio-style high-rep sets are excluded from the strength stats). 1RM estimates use the Epley formula (`weight × (1 + reps/30)`) restricted to working sets in the 1-12 rep range — the band where Epley is best validated.

All percentile and average numbers in this article are computed per-user-per-lift first, then aggregated across users — so a single high-volume lifter cannot skew the average upward. Female samples below n=15 are reported only when explicitly noted; below that threshold the noise dominates the signal.

Big-3 strength ratios (male sample)

The classic strength reference (bench, back squat, conventional deadlift) measured as best-logged 1RM divided by bodyweight. These are working-set extrapolations, not all-out 1RM tests, so they sit a few percent below true single-rep maxes.

Note: female samples are too small in this 5-month window for a meaningful gender breakdown of the big-3. The next quarterly snapshot should clear the n=15 threshold for at least squat and bench.

Volume per workout and per week

The median Arvo session is 14 working sets in 67 minutes — about 4.8 minutes per set including rest. At 3.4 workouts per week that adds up to ~48 sets weekly across all muscle groups.

Average reps per set is 10.4 — squarely in the hypertrophy range (5-15) where set count, not load, drives most growth (Schoenfeld 2017). The session-length data corroborates: lifters consistently choose 60-75 minute sessions over 2-hour marathons, even when the training plan would allow it.

Effort: how close to failure people actually train

Among sets where lifters logged RIR (Reps In Reserve), 78% sit at RIR 0-2 — meaning they are within two reps of muscular failure. 4,070 sets were taken to outright failure (RIR 0); 15,520 sets were 1-2 reps shy of failure; only 5,491 sets were "easy" (RIR 3-5).

The mean RIR across the dataset is 1.89 — a clear signal that the average Arvo user trains hard, not "junk volume" hard. This matches the literature on hypertrophy proximity-to-failure (Helms et al. 2018: 0-3 RIR for compound work, 0-1 RIR for isolation).

Training frequency

Active users (those with ≥ 4 completed workouts in the window) average 3.4 sessions per week — landing exactly in the 3-4×/week sweet spot for natural intermediate hypertrophy that meta-analyses converge on (Schoenfeld 2016, 2× per muscle minimum). The dataset includes upper/lower 4-day, push/pull/legs 6-day, and full-body 3-day splits in roughly equal proportions.

Progression: how fast top-set weight actually moves

For user-exercise pairs tracked over ≥ 4 weeks (n=22), the median top-set weight progression is +4.1% over 4.6 weeks — about 0.9% per week. That is consistent with intermediate-lifter linear progression (Rippetoe's "Practical Programming" estimates 1-2 lbs/session for intermediates, which converts to roughly 0.5-1.5% per week depending on bodyweight class).

The 22-pair sample is small because we required at least 4 distinct calendar weeks of data per exercise. The next snapshot, with the dataset matured to 9-12 months, will give a confident view of this metric.

Most-logged exercises

Out of 2,429 unique exercise names, 221 exercises have at least 100 logged sets — meaning the catalog is broad but the working surface is concentrated. The top 10 cover the classic powerbuilding pattern: barbell big-3 dominate, followed by isolations for back width (lat pulldown), shoulder caps (lateral raise), and arms (triceps extension, bicep curl).

Caveats and what this is not

This is observational data from a self-selected group of users who chose to log their training in Arvo over a 5-month window. It is not a controlled study, the population skews toward intermediate-to-advanced trainees who care enough to track sets, and the female sample is too small for definitive cross-gender claims. The 1RM ratios are Epley extrapolations from working sets, so they will read 5-10% lower than true 1RM tests. Expect this article to be re-published quarterly as the dataset matures.