Progressing a gym exercise or a functional session from the log

Keep the exercise, its name, its unit and its rep range for the block. Change exercises at a block change, for pain, or for missing equipment, not for variety. The stronger reason is measurement: a changed scheme cannot be read, as in a squat written x3, x5, x10 in six weeks.

The evidence for keeping is weak: one review concludes, in young men, that some systematic variation seems to help and excessive random rotation may compromise gains, with no effect size and no count of the studies behind it (REF-1487, B+, abstract), and ACSM's 2026 overview cannot determine an effect of varied exercises on strength or size (REF-1476, A).

One scheme stays one block: 3-6 weeks by the house default (SYNTH-42, C+; SYNTH-90, B-). One movement has one name; two names split the history, so ask once and join them.

The next number starts from the last done set and its RPE. Not from memory and not from the plan's old number. When the athlete lifted more than planned at or under the target RPE, the plan follows the athlete. If the log holds a load, the plan carries a load: a zero or "bodyweight" is written only when no load is known, with an RPE and a sentence saying they pick it. No RPE on the last set: repeat the load and ask for one.

Move one axis per exercise per session, and name it. Barbell or machine: reps up inside the range at the target RPE on all sets, then the smallest load step and back to the bottom of the range (double progression); or, on a low-rep strength goal (a 4-8 band or one rep target), add load directly once all sets end at or under the target.

Load and rep progression show no clear difference in trained young adults over 8 weeks (REF-1479, B+, 8-12RM, to failure) or untrained legs over 10 (REF-1480, B+, abstract), so choosing either is fine at moderate reps (inference); the point is to choose and name it.

On a heavy low-rep goal load stays first: REF-1479 does not test that band, REF-1480's abstract gives no rep targets, heavier loads raise strength (REF-1476, A), and REF-1479's 1RM interval does not rule out a load advantage.

Sets move last and at a block change, not session to session (inference): diminishing returns are sharper for strength than for size (REF-1477, A-, abstract, no threshold).

Step size: the smallest step the gym has, commonly 2.5 kg on a barbell and the next dumbbell up; about 5% as a ceiling on a heavy barbell lift is practice, nothing more; on a light lift or a dumbbell the smallest step may be larger, a reason to add reps first (practice). Double progression itself has never been compared with another rule.

Hold when the log says hold. RPE over target or sets not finished: the same load next time, not more.

Twice running, or a clear drop (2 or more reps lost at the same load on most sets, or a falling verdict on a like-for-like scale): one step back, and look for why before changing the programme: pain or illness first (SYNTH-73, SYNTH-36), then sleep, food, the hard ride or run before it, spacing by SYNTH-31. One bad session is no verdict.

At block level the leading groups' first checks are adherence, whether load progressed, sleep and food (REF-1143, B-, muscle growth, transfer is inference), and one before-and-after test cannot separate a low response from noise (REF-441, B+); applying that to a single session is our inference.

Fourteen days or more is a return, not a step up. After 14-27 days: about 10% under the last load or two small steps, whichever is more, about two-thirds of the sets by SYNTH-36's 50-70% of prior volume, target RPE one lower, back at the old load in 2-3 weeks. After 4 weeks or more: about 20% under, the same sets rule, 3-5 weeks back. How far back is practice.

The cut is for tissue as much as lost strength (inference): after 10 weeks off, untrained adults lost 5.4% of leg press 1RM and 3.6% of curl 1RM, and after five weeks of two sessions a week no measure differed significantly from their week-10 level (REF-1488, B+, no endurance training, breaks not due to illness or injury).

Muscle disused for 10 days lost 29% of force and nine sessions restored it (REF-1142, B, n = 11, one leg suspended, not a training break). Strength holds longer than the aerobic side (PHYS-41, established for the order of the losses). Illness or injury: SYNTH-36 and SYNTH-73 decide. The first session back counts as new for SYNTH-31's gap (inference).

For an endurance athlete strength is a supplement: progress on purpose, and hold on purpose. Stay short of failure, about 2-3 reps to spare, SYNTH-43 and SYNTH-44 (practice; target RPE about 7-8 unless the plan says otherwise). Strength gains varied little across estimated reps in reserve (REF-1483, B+, exploratory, estimated, detailed numbers from its preprint).

Pick the one or two lifts the block is for and let the rest hold; do not progress everything at once, and when strength and endurance load rise together, ramp by the slower system (SYNTH-21, practice). Expect progress in a build of two sessions a week: a rep on most sessions, a load step every 1-3 weeks on the main lifts while they are new, slower after.

Do not expect it on one session every 7-10 days, in a heavy endurance block, in the race phase or in a taper.

In season hold the load (SYNTH-43's build load; one session a week kept the gains in REF-800, B, 6 per group), and add a small step only when the held load has read 2 under target for two sessions, to keep the relative load (REF-587, B+, general populations): that keeps strength, and is no build.

In the taper: sets cut, load kept, nothing new in the last 2-3 weeks (SYNTH-34, practice). Flat by design is not a stall; say which it is.

RPE steers, and it is only as good as the set behind it. It is most reliable near failure: error about 1 rep with up to 5 reps left and more than 2 with 7-10 left (REF-1484, B, chest and leg press; on leg press 1 rep or less only up to 3 left), and 0.65 ± 0.78 reps calling 1 or 3 left on the bench (REF-1485, B, trained lifters).

Nothing tests a squat, a hinge, a low-rep set or a station, neither study tests the RPE number itself, and both took sets on to failure after each call, feedback a normal set lacks (inference). So an RPE of 5-6, about 4-5 reps left on the app's reading, is a soft "easy": on a leg press that call was already off by more than a rep.

It licenses one step with reps back at the bottom of the range, not a step at the same reps (hip thrust 110 × 10 at RPE 6, then 115 × 10 at RPE 9). An RPE of 9 after a hard run or ride is more likely a tired day than lost strength: hold, and read it again on a fresh day (inference; no study).

It never explains away pain or illness: ask first, and SYNTH-73 or SYNTH-36 decide. ACSM finds too little evidence to quantify exact reps-in-reserve or RPE targets (REF-1476, A).

Bodyweight: reps to the top of the range, then a harder variation or added load, back at the bottom. Progressive push-up variations raised bench 1RM in 4 weeks (REF-673, B+, 23 moderately trained men), banded push-ups at 6RM raised it like the bench (REF-671, B+), and position-adjusted push-ups did too (REF-672, B); that a harder variation should follow the top of a range is those entries' inference.

A move stuck at the same reps for 3 or more sessions outside a declared hold needs a named change, as pull-ups at 5-6 × 3 for five weeks did.

Stations, carries, ergs, sled, wall balls, lunges: no study tests a progression rule. A PubMed search on 2026-10-30 found none, and a 2018 review found no study of training-load management in CrossFit (REF-1492, B-, search closed 2016).

Practice: one unit per block, metres or seconds on an erg and not both; fix the load and move distance, or fix distance and move load, or fix both and move rest. A carry or lunge states its load in the load field, not in a comment or the name; state once whether it is per hand. Change the ball or the sled and it is a new baseline.

For HYROX the end point is the division's race load at race distance; SYNTH-71 owns the race plan and the week.

A station the athlete cannot do gets a named substitute decided once, by SYNTH-98's logic (nothing that provokes symptoms, then closest to the movement; a transfer, it names no station; PHYS-40, established for specificity), kept under its own name; it is not skipped a third time.

Circuits and mixed sessions compare only with themselves. Repeat the same session as a benchmark every 4-6 weeks (practice) and otherwise do not compare. Reps per minute is a proposal, not a validated rule, and a higher rate can mean more fitness or an easier workout (REF-1493, C+).

Supersets and circuits are time savers: about 37% shorter sessions (a discussion figure, no interval) with no clear difference in volume load (SMD 0.05, −0.48 to 0.57), less in same-muscle pairs (SMD −1.08), higher RPE (SMD 0.77), and 1RM SMD 0.10 (−0.40 to 0.60) on three small trials at 50-80% 1RM (REF-1495, A-); circuits beat control on VO2max and bench 1RM, against no other training (REF-1494, B+, abstract).

Compare a superset session with the last superset session (inference). Not on the heaviest lifts with short rests: REF-1495's rests between supersets ran 1 to 3 min, its one heavy protocol was acute, and its authors say load still needs care for maximal strength, so the cycling guide's about 3 min rests for hard sets stand (SYNTH-43; inference).

Someone new to this kind of session starts with 2-4 weeks at moderate load to learn the lifts (SYNTH-71, SYNTH-75, practice).

Periodisation and deloads: no elaborate scheme. A small to moderate edge for 1RM (Key Findings 7), not determinable in ACSM's overview, nothing tested in endurance athletes. A planned lighter week is optional, and signals can call for one (SYNTH-29): halve the sets, keep the load (SYNTH-29, SYNTH-43). Afterwards go back to the numbers before it. It is not a step and not a verdict.

Read the log honestly. An e1RM is a reading only on low-rep loaded sets; the app has none above 12 reps, and it means nothing on a 20-rep wall ball. A falling verdict after a lighter load with more reps is not a loss. Compare like with like: same exercise, unit, load class and set structure.

Progressing beat holding in the one trial that held load and reps flat, and holding still worked. 55 untrained young women, arms as the unit, 3 × 8-12, 3 days a week, 8 weeks: triceps thickness +0.79 cm (+22.9%) when load rose at the top of the range, +0.41 cm (+11.6%) with the same load and reps, +0.05 cm (+1.8%) with no training (REF-1481, B+, abstract). Thickness of one small muscle, no strength outcome, effort not reported.

Load or reps: no clear difference, and neither is equivalence. 43 randomised, 38 finished, 8 weeks of lower-body work to failure: Smith squat 1RM 2.0 kg toward load (90% CI −2.4 to 7.8), rectus femoris 2.8 mm toward reps (90% CI −0.5 to 5.8), others small; 4 of 21 lost in the reps arm against 1 of 22 (REF-1479, B+). 39 untrained people, one leg each, 10 weeks: leg extension 1RM 52.90 to 69.05 kg vs 51.67 to 66.82 kg, P > 0.05 (REF-1480, B+, abstract).

Real rules exist; none is tested as a rule. "Increased the load whenever it reached the upper limit of the repetition range", step not given (REF-1481). "Aimed to increase load while maintaining this target repetition range" of 8-12RM, or to increase reps per set at the initial load, no step, no trigger (REF-1479).

The repetitions in a weekly failure set set next week's kg, for leg press +2.5 kg at 11-12 reps, +5 at 13-15, none at 8-10, −5 at 6-7 and −7.5 under 5, used in both arms and compared with nothing, and needing a set to failure that SYNTH-43 and SYNTH-44 avoid (REF-1488, B+, Table 1).

ACSM's 2026 stand is thin on progression, and in tension with itself. Across 137 reviews and more than 30,000 participants, load, sets, frequency, order and range moved strength; failure and periodisation did not clearly. Its progression rows read cannot determine for autoregulation, varied exercises and, for strength, periodisation.

The authors write that progression is "not necessary to achieve beneficial outcomes" and overload "likely a requirement only for those seeking continued longer term progress", while the same paragraph calls progression "a requirement to achieve a meaningful benefit" for some inexperienced people on necessarily low loads, a later one makes it a core principle for the untrained, and the abstract says healthy adults should train progressively; no row compares progressing with not (REF-1476, A, mostly inexperienced trainees).

Session RPE as a steer is a hint. Fewer high-intensity sessions, strength 9.68 vs 9.28%, muscle area 6.55 vs 9.65%, in 20 trained young men over 6 weeks; the rule is not in the abstract (REF-1482, B-). Autoregulated methods ranked above fixed percentages, with the clear gap for APRE on the bench and no moderate or large effect on the squat (REF-685, B+).

Closeness to failure: little for strength, some for size. Strength slope 0.003 per estimated rep in reserve (95% CI −0.012 to 0.018), with reps in reserve estimated, modest fit and detailed numbers from the preprint (REF-1483, B+).

Stopping at 20% velocity loss gave similar 1RM gains to 40% with 40% fewer reps (REF-584, B+, 22 young men, squats) and the largest 1RM gains of four thresholds (REF-1085, B+, about 13 per group, moderately trained men). The base reads these as its reason for stopping short; the app has no velocity device, so its field version is reps to spare (SYNTH-43).

Periodisation: a small to moderate edge for 1RM, with a lower estimate in a volume-equated pool of different studies. ES 0.43 (0.27-0.58), 18 mostly older studies (REF-683, A); ES 0.31 (0.04-0.57) with volume equated, and hypertrophy 0.13 (−0.10 to 0.36) (REF-1486, A-, abstract).

Undulating over linear ES 0.31 (0.02-0.61), trained 0.61 (0.00-1.22), untrained 0.06 (−0.20 to 0.31), with no test of that subgroup difference in the abstract (REF-1486); no difference in 17 head-to-head studies (REF-684, A-). None names endurance athletes.

A week off cost lifters strength; a lighter week showed no evidence of a cost. A complete week off at week 5 of 9 gave smaller strength gains (REF-217, B, 39 resistance-trained). Two 2-set weeks, 46 against 56 sets, no evidence of a difference in 19 untrained men, not sized for that (REF-1226, B+). Coaches agree a deload cuts demand, and that practice rests on experience (REF-1228, C).

Functional-training injury data give rates, not causes. 36% reported an injury (95% CI 32-41%), 4.3 per 1000 h (3.35-5.23) and 9.9 (3.3-16.4) in the five prospective cohorts, single studies 0.04 to 18.90 (REF-1489, B+, abstract).

One cohort of 406: people alternating prescribed and scaled loads had higher odds than those always scaled (adjusted OR 3.5, 1.7-7.3), "limited evidence", "scaled" undefined, and the review found no prevention study (REF-1490, B+).

New CrossFit members: 9.5 injured people and 10.6 injuries per 1000 h, 2.66 counting only the 7 injuries not provoked by an old problem, 38% of those who opened the email answered both surveys, compared with experienced people only across studies (REF-1491, B; this cohort may be among REF-1489's five, so the two are not independent).

HYROX athletes: 14.4 per 1,000 h, mostly overuse, knee and lower leg (REF-904, B). None tests a load jump or a progression.