Counting sets for "shoulders" is why your rear delts lag

Most training logs let you count sets per muscle group. You did 18 sets of “shoulders” this week, which sounds thorough, and your rear delts are still nowhere.

The number was never wrong. It was just too coarse to be useful.

Muscle groups are not single units

The deltoid has three heads with different origins, different lines of pull and — critically — different exercises that actually load them:

• Anterior — shoulder flexion. Loaded by every pressing movement you already do.
• Lateral — abduction in the scapular plane. Loaded by raises and upright rows.
• Posterior — horizontal abduction and external rotation. Loaded by rear-delt flyes, face pulls and wide-grip rows.

Now recount those 18 sets. If your week was bench, incline, overhead press and a few lateral raises, the split might be 14 anterior, 4 lateral, 0 posterior. “18 sets of shoulders” and “zero sets of posterior delts” are both true statements about the same week.

The same problem exists everywhere:

• Biceps — the long head crosses the shoulder joint and gets far more stretch-mediated work in an incline curl than in a preacher curl, which biases the short head.
• Triceps — the long head is the only one crossing the shoulder, so overhead extensions hit it in a way pushdowns do not.
• Hamstrings — knee-flexion work (leg curls) and hip-extension work (RDLs) are not interchangeable, and the biceps femoris short head only does the former.
• Pecs — the clavicular head needs some incline in the movement to take a real share of the load.

Fractional counting beats binary counting

The usual fix is to tag each exercise with the muscles it “works”, then count a full set for each. That overcounts badly: a barbell bench press is not one full set for anterior delts and one full set for triceps and one for pecs.

The more honest approach is fractional. Each exercise contributes a fraction of a set to each head it loads, based on how much that head actually contributes to the movement. A bench press might count as 1.0 to the sternal pec, 0.5 to the anterior delt, 0.4 to the triceps, and 0.0 to everything else. Add those fractions across your week and you get a per-head volume number that means something.

It also fixes the double-counting that makes people think they are doing 30 sets of arms a week when they are doing 12 direct sets and a lot of pressing.

Target bands, not target numbers

Once you have per-head weekly volume, the obvious question is how much is enough. The research here supports a range rather than a single figure — roughly 10 to 20 hard sets per muscle per week for most trained lifters chasing hypertrophy, with meaningful individual variation and diminishing returns past the top of the range.

Bands are the right shape for this because both ends carry information:

• Under the band — the head is undertrained and is likely your limiting factor.
• Inside the band — you are in the productive range; consistency matters more than adding sets.
• Over the band — you may be spending recovery you would rather put somewhere else.

The useful output is not “you did 14 sets.” It is “your posterior delts are at 2 and your anterior delts are at 19, and that gap is why your shoulders look narrow from the side.”

How to fix a lagging head

Once you can see it, the fix is boring and effective:

1. Add direct work, early in the session. A head that only ever gets trained as a synergist at the end of a workout is being trained with your worst effort.
2. Pick the exercise that actually loads it. Face pulls and reverse pec-deck for posterior delts. Overhead extensions for the triceps long head. Incline curls for the biceps long head. Leg curls and RDLs, not one or the other.
3. Give it two exposures a week. Splitting the same volume across two sessions is generally at least as effective as one, and it is easier to do well.
4. Re-check in three weeks. The point of measuring is that you can tell whether the fix worked.

Seeing it without doing the arithmetic

Doing fractional counting by hand is possible and nobody does it, because it means maintaining an activation table for every exercise you own and re-adding it every week.

Momentum does it automatically. Every logged set is distributed across the heads that exercise actually loads, summed weekly, and scored against per-head target bands — then drawn on an anatomical figure so a lagging head is something you notice rather than something you calculate. Heads that fall below their band get flagged directly.