Your training log already ran the frequency study the meta-analyses can't.

Two squat blocks, two frequencies, one answer already in the log

Somewhere in a three-year training export there's a spring block where squats ran twice a week and a fall block where they ran three times. Same lifter, same bar, similar working weights. One block moved the top set. The other moved it slower, or not at all, and the lifter almost certainly never went back to check which.

That comparison is the best frequency study this lifter will ever get. It's already finished.

Type 'how often should I train each muscle group' into a search bar and the answers converge on a range and a shrug: twice a week is a solid default, three is fine, total volume matters more. As a population answer, that holds up. But a lifter with years of logs is asking something narrower. Which frequency did my squat respond to?

The finished experiment: your log already toggled the variable

Call it the finished experiment. Almost every lifter with three or more years of honest logs has changed frequency on the same lift more than once, usually by accident. A program switch. A new split. A job that ate a weekday. Each change left a before and an after in the record.

The literature asks what frequency works for the average trainee over a study's length. Your log answers what frequency worked for your body, across blocks you ran, at efforts you wrote down. Those are different questions, and for your own squat the second one outranks the first.

The meta-analyses are sound. They're just measuring someone else.

Why the population answer flattens your answer

Group averages hide the spread underneath them. Hubal and colleagues (2005) put a large cohort through the same arm-training protocol and reported a wide range of responses in both size and strength: some participants barely changed, others changed a lot. Same program, very different bodies.

Apply that to frequency. A finding of 'no significant difference between two and three sessions a week' is fully compatible with a sample where some lifters did better on three, some did better on two, and the averages landed in a tie.

There's a second gap. Studies hold weekly volume equal so they can isolate frequency. In a real gym, frequency almost never changes alone. Add a third squat day and you usually add sets, swap a variation, or drop intensity to survive it. Your log captured all of that, which is a problem and an opportunity at once.

The tie describes the group. It says nothing about which side of it you sit on.

What do the frequency meta-analyses actually show?

Three reviews carry most of the weight in this debate. Schoenfeld, Ogborn and Krieger (2016) pooled hypertrophy studies and found that training a muscle at least twice a week outperformed once a week. Grgic and colleagues (2018) looked at strength and found the apparent benefit of higher frequency mostly tracked the extra volume that came with it; with volume equated, frequency didn't show a significant effect. Schoenfeld, Grgic and Krieger (2019) reached a similar conclusion for hypertrophy under equated volume.

So the honest summary is narrower than the headline. When weekly volume is held constant, average differences between frequencies are small. A lifter can accept that finding completely and still ask the individual question, because it's a finding about averages.

How to read frequency out of your own training log

The method runs one lift at a time.

Tag every block in your history by how many times that lift appeared per week. Drop blocks too short to show a trend; a three-week block mostly measures how fresh you started. Then filter to top sets logged in a narrow effort band, say RPE 8 to 9. Most lifters skip this step, and it's the one that matters most.

Convert those sets to an estimated 1RM. Epley's formula, weight × (1 + reps ÷ 30), is the common default, so a 140 kg top set for 5 comes out around 163 kg. An RPE-to-percentage chart, like the one Reactive Training Systems popularized, does the same job with effort built in. Plot e1RM against week and read each block's slope: how many kilos per week the estimate climbed.

Last, pull weekly working sets for the same windows. If the higher-frequency block also carried more volume, you're reading the two together. Say so to yourself before drawing a conclusion.

Per-set RPE is what makes the comparison valid

Without per-set RPE, a frequency comparison is quietly an effort comparison. A three-day block where most squats landed around RPE 7 and a two-day block grinding RPE 9.5 sets will produce different slopes for reasons unrelated to frequency. RPE is the control variable. Matching on it is what lets the slopes mean anything.

Even matched, the read is noisy. Life stress, sleep, a bodyweight cut, a swap from high-bar to low-bar, and plain training age all shift slope. Your fourth year moves slower than your first no matter the split.

So treat one comparison as a lead. Two switches pointing the same way is a pattern. Three is about as good as personal evidence gets, and for your squat it beats any pooled average.

What changes once your log outranks the average

Frequency stops being a debate and becomes a setting with evidence behind it. A once-a-week main-lift template and a high-frequency squat program are both defensible bets on paper. Your log shows which bet paid on your body, and the answer can differ by lift: bench might respond to more exposure while deadlift barely notices.

It changes how the next block gets designed, too. If frequency is the open question, change frequency, hold the rest steady, and keep the RPE column honest. Each block becomes a cleaner data point than the one before.

And it moves the authority. Published guidance sets a sensible starting range. Inside that range, the lifter's own record decides.

Keeping old blocks on the same timeline as this week

The finished experiment only pays out if the old blocks are still readable. For a lot of multi-year lifters they aren't: one year lives in a spreadsheet, two in a previous app, a summer in a paper notebook. A frequency comparison across that is a reconstruction project.

Platepusher imports a CSV from Strong or Hevy as native data, every lift, every set, every date, so a 2023 block and this month's session sit on one timeline. The plateau flag marks the week a top set stops moving, which is where you'd start looking when one frequency block went flat and another kept climbing. The math describes. The lifter decides.

What we're tracking next

Within-subject designs, where each leg of the same person gets a different protocol, have shown up in recent research on individual response. They're the lab's version of the finished experiment, and they point the same direction this piece does. The useful frequency question is personal, and the study that answers it best for you is already sitting in your training history.