Home / Guides / Exercise Muscle Mapping

How Ægir Iron Maps Every Exercise to the Muscles It Trains

Every per-muscle number in Ægir Iron starts with one question about each exercise: which muscles does a set of it train, and how much? The answers sit under the program analyzer, the Body Map, the weekly volume read, the exercise cards and the program generator. This page explains how we built those answers for more than 900 strength exercises, and the checks a new exercise passes before it counts.

What one exercise's entry holds

Each strength exercise carries a weight between 0 and 1 for every muscle group it trains, out of sixteen groups: chest, shoulders, biceps, triceps, forearms, abs, obliques, traps, lats, lower back, glutes, quads, hamstrings, calves, adductors and neck. A weight of 1 is a prime mover, the muscle the lift is built around. Lower weights are helpers. A set credits each muscle by its weight, which is called fractional set counting. A 2026 meta-regression found this method fit the research better than counting only the muscle an exercise is named after (Pelland et al., 2026; the idea was set out earlier by Schoenfeld et al., 2019).

The same weights drive everything that reads muscles, so the screens agree with each other:

• the program analyzer (see See What a Program Trains Before You Run It) and the generator's accessory picks;
• the Body Map and weekly sets per muscle for what you log;
• the Primary and Secondary muscles and the anatomical figure on each exercise's card, where Primary means a weight of 0.7 or more;
• the Strength Level heat map.

Mobility and cardio rows carry no weights and credit no muscle.

Where the weights came from

The first version came from rules keyed on movement names. A press got chest, triceps and front delts, and a pull got lats and biceps. That is a reasonable start, and it goes wrong in predictable ways. A wall sit matched the core rule because its name contains "l sit". Handstand push-ups picked up a full chest credit from the push-up rule. One carry rule gave a yoke walk a full forearm credit, although a yoke has nothing to grip.

So in late September 2026 we reviewed the whole catalog one movement family at a time, in seventeen batches:

• the back squat
• the bench press family
• incline and close-grip presses
• flys and dips
• the Olympic lifts
• the deadlift family
• overhead presses
• rows, pull-ups and pulldowns
• squats
• shoulder isolations
• arms
• hip and hamstring isolations
• calves and neck
• calisthenics holds
• plyometrics
• strongman events and carries
• the core

For each family, the mechanics were checked against the published research, and a weight moved only where the evidence pointed one way. Three kinds of evidence carry the most weight:

• Electromyography (EMG) shows which muscles are working hard during a lift and how two variations compare. For example, two studies found less triceps activity with dumbbells than with a barbell or Smith machine, with the same chest and front delts (Saeterbakken et al., 2011; Farias et al., 2017), so dumbbell presses credit the triceps a little less than the barbell bench does.
• Training studies with muscle imaging show what a lift grows over weeks, which is closer to the question a volume count asks. After twelve weeks of leg press, MRI showed the gluteus maximus and adductor magnus growing more than the quadriceps as a whole (Kinoshita et al., 2026), so the leg press credits glutes and adductors as well as quads.
• Where studies disagree, the existing weight stays. Two studies of squat stance width measured different adductor muscles and reached different answers, so the sumo squat kept its base adductor share instead of taking the larger one.

Some questions have no study behind them. We found no published data on the deltoids or hamstrings in the snatch or the clean, for example. In those cases the weight is a written coaching judgement, recorded as ours rather than presented as a finding.

A second reviewer, and one rule for helpers

When the seventeen batches were done, an independent pass read every source they had added against its abstract. It also ran the rules over the whole catalog before and after the changes, so that any exercise whose numbers had moved could be checked by hand. It found that the batches had used three different standards for when a helper muscle earns a share. Under one standard the back squat's hamstrings were zeroed for not growing in a training study, while under another the hip thrust's kept their share on the same kind of evidence.

The fix was one written rule, now applied to every exercise:

• A muscle that moves the load takes a helper share whenever it measurably works.
• A muscle that only holds the body in position takes one when it works hard or grew in training.
• A training study that found no growth lowers a share but never zeroes it.
• A share is set relative to whatever limits the lift. The grip ends a farmer's walk, so the back and hips are helpers there. A yoke isn't limited by grip, so they become prime movers.

Applying the rule moved one whole family: squats now credit the hamstrings 0.3. The same pass corrected a handful of smaller shares and several study summaries that had been read too generously.

The sources, in aggregate

We don't publish a weight-by-weight table here, because it would be long and mostly of interest to us. Here is the evidence base by the numbers:

922	Strength exercises in the catalog today, out of 1,008 in all (the rest are mobility and cardio, which train no muscle group). Every one has its own reviewed entry: 914 when the review closed, and each exercise added since went through the checks below. The automated checks require that no strength exercise is left unreviewed.
117	Published sources in the app's bibliography for the muscle model: 108 studies, 8 reviews and a meta-analysis.
80	Of those, cited directly on individual exercises' review records.
28	Written coaching conventions that cover the gaps the research leaves, each labelled as our judgement.

Every source is listed with its link on the research page. Where a study's full text wasn't open and only the abstract could be read, the record says so.

What a new exercise goes through

The catalog grows as people ask for movements, so the review also has to hold for exercises added after it. When a new exercise is added in the app's catalog manager, its muscles are set in the same form, in one of three ways:

• Same muscles as another exercise. This is the usual case: a new bar, grip, tempo or implement on a movement that is already mapped. The new row copies that movement's reviewed weights and its sources along with them.
• Custom weights, with the source. A weight for each muscle, plus the study's DOI or link and a sentence on what it found. The form refuses to save custom weights without a source.
• Leave it to the rules. The form shows what the name-based rules would give the new exercise, and warns when no rule recognises the name and the exercise would get only its category's generic muscles.

An exercise that has already been reviewed refuses a change that disagrees with its review. Changing it means changing the review, with the study that justifies it. A readiness checklist on each new row shows what is still missing, and a data health screen reads the live catalog for exercises with no muscles, no prime mover, only the generic category muscles, or a setting that contradicts a review.

Settings made in the catalog manager take effect in the app at once. They are then moved into the code, where custom weights get their study entered in the bibliography and become a pinned review record of their own.

What the automated checks catch

Every change to the app runs a test suite that reads the whole catalog. It fails when:

• a strength exercise trains no muscle group, or has no prime mover;
• a reviewed exercise scores anything other than what its review recorded;
• a review cites a source that isn't in the bibliography;
• an unreviewed strength exercise appears;
• a muscle setting made in the catalog manager hasn't been moved into the code.

The checks are tested themselves: each one carries a case it must catch, so a check that has stopped finding anything fails instead of passing.

What the weights don't claim

A weight is a credit for counting volume. It isn't a measurement of how hard a muscle works in your body on a given day. It has three limits:

• The amounts are ours. The research decides which muscles a lift credits and which direction a weight moves. The exact number is a coaching judgement anchored to that research. Our helper weights run a little more generous than the flat half-set the meta-regression used for indirect work, so a deadlift set counts as about 4.9 effective sets here against about 4 under that method.
• EMG is not growth. A muscle can be very active in a lift without growing much from it. That is why training studies adjust the size of a share, and why the helper rule treats them differently from activity.
• One shoulder region. The front, side and rear delts share a single shoulders group, so studies about which head a raise works can't move a weight. Splitting them is on our list.

Common questions

How does Ægir Iron decide which muscles an exercise works? Each of more than 900 strength exercises has a weight from 0 to 1 for every muscle group it trains, reviewed one movement family at a time against published EMG and muscle-growth studies, and then checked again by an independent second pass. Where the research is silent, the weight is a written coaching judgement, labelled as ours.

Why does a deadlift count toward my traps and quads? Because they work during it. Ægir Iron credits a set to every muscle it trains, weighted by involvement. A conventional deadlift counts fully for the glutes, most of a set for the hamstrings and lower back, and part of a set for the traps, quads, forearms and adductors.

What happens when a new exercise is added? It either copies the reviewed weights of the movement it varies, or gets its own weights with a cited study (custom weights can't be saved without one). Automated checks then fail if it trains no muscle, has no prime mover, or contradicts an existing review.

See it on your own training →

References

1. Pelland JC, Remmert JF, Robinson ZP, Hinson SR, Zourdos MC. The resistance training dose response: meta-regressions exploring the effects of weekly volume and frequency on muscle hypertrophy and strength gains. Sports Med. 2026;56(2):481-505.
2. Schoenfeld BJ, Grgic J, Haun C, Itagaki T, Helms ER. Calculating set-volume for the limb muscles with the performance of multi-joint exercises: implications for resistance training prescription. Sports (Basel). 2019;7(7):177.
3. Saeterbakken AH, van den Tillaar R, Fimland MS. A comparison of muscle activity and 1-RM strength of three chest-press exercises with different stability requirements. J Sports Sci. 2011;29(5):533-538.
4. Kinoshita M, Maeo S, Kobayashi Y, et al. Hypertrophic effects of single- versus multi-joint exercise: a direct comparison between knee extension and leg press. Med Sci Sports Exerc. 2026;58(7):1566-1580.

The full list of sources behind the muscle model is on the research page.

Keep reading: The program muscle analyzer · Sets per muscle per week · Research and sources