Why does my lifting app show a strength drop after I switch exercises?

Most trackers treat each exercise name as its own series. Log a new variant under the old name and the chart compares two different movements as one, so a lighter variant reads as a regression. Log it as a new exercise and the new line starts with no history at all.

Should I log high-bar and low-bar squats as the same exercise?

No. They load the hips, knees and torso differently, so their numbers aren't comparable set to set. Keep them as separate variants, ideally linked to the same parent lift, so each one is judged against its own history.

Is there a conversion ratio between low-bar and high-bar squat?

There's no universal one. Most lifters handle less weight high-bar, but the gap depends on build, technique and training history. Treat any fixed conversion with suspicion and let your own paired sessions show the relationship.

You Rotated Your Bench Grip. Your Log Now Thinks You Regressed.

Article · 5 min read

Switch a lift variant and your log calls it a regression.

Q: How long before a new variant gives a useful trend?

Treat the first few sessions as calibration. Bar position, groove and depth are still settling, so early jumps reflect technique as much as strength, and a stall flag that early is mostly noise.

Switch low-bar to high-bar, or close-grip to wide, and most logs read a strength drop. A variant swap starts a new baseline that needs a link back to the old lift.

Week seven, the squat "dropped" 30 pounds

Take a lifter six weeks into a squat block, running low-bar because that's how they learned. Top sets have climbed every week. In week seven they move the bar up onto the traps, high-bar, to chase more quad work and a more upright torso. The first high-bar top set lands about 30 pounds under last week's low-bar number.

The app draws a line down. Estimated 1RM: down. Trend: down. If the tracker has a stall or regression flag, it's lit.

The legs are the same legs they were last Tuesday. What changed is the movement, and the log has no idea the two movements are related, or how.

The Baseline Fork

A variant swap is a fork in the series. The new variant starts its own baseline on day one, and that baseline needs a pointer back to the lift it branched from. Call it the Baseline Fork.

Both halves matter. The new branch gets judged only against itself, so the first high-bar session is measured against the next high-bar session and never against a low-bar top set. And the pointer keeps the story intact: six weeks of low-bar work is part of why the high-bar numbers start where they do, and a lifter looking back over the year should see that lineage in one place.

Most logs get one half and drop the other.

Why does my strength app show a drop after switching exercises?

Trackers generally key history to the exercise entry. "Squat (Barbell)" is one series, "High Bar Squat" is another. If the lifter logs the new variant under the old name to keep the chart going, the chart compares two different movements as if they were one. That's where the false drop comes from.

The obvious fix is just as wrong. Lump every squat variant into one "squat" line and front squats, safety-bar squats and paused squats all bleed into the same trend. It flattens into mush, and a real stall in the competition squat hides under a good week of front squats.

Separate strings lose the history. One bucket loses the signal. The fork keeps both.

Four ways a training log can model a variant swap. Category-level behavior, not a vendor comparison.
	Same name, logged through	New exercise entry	One lumped "squat" line	Linked baseline (the fork)
Chart after the swap	A false drop on swap week	Old line stops; new line has no past	A cliff, then a blurred trend	New line starts its own baseline, tagged to the parent lift
Stall or regression flag	Fires on the swap	Silent until the new variant builds history	Fires on the swap, then misses real stalls	Judges the variant only against itself
Year-end review	One lift with a fake dip	Two lifts that look unrelated	One blurred lift	One movement family, branches visible

Low bar vs high bar squat: why the numbers don't carry over

Variants exist because they load the body differently, and that's exactly why their numbers don't transfer one-to-one. The biomechanics literature on the two back-squat positions describes a more upright torso and greater knee demand high-bar, and more forward lean with greater hip demand low-bar Glassbrook et al. (2017, Journal of Strength and Conditioning Research) reviewed high-bar vs low-bar back squats: high-bar showed greater knee flexion, lesser hip flexion, a more upright torso, and a deeper squat. ([source](https://www.ovid.com/jnls/nsca-jscr/fulltext/10.1519/jsc.0000000000002007~a-review-of-the-biomechanical-differences-between-the)). Different joints carrying the load means a different number on the bar.

The same logic runs through the bench and the deadlift. A narrower grip lengthens the bar path. A pause strips out the rebound. Sumo and conventional shift work between hips, knees and back in ways that depend so heavily on the lifter's build that no rule says which one will be heavier.

And estimated-1RM math assumes a stable movement. Epley and Brzycki turn reps at a weight into a max for that lift. Feed them two different lifts and the output is precise and meaningless.

Direction is the common pattern, not a fixed ratio. Individual lifters vary, and no conversion factor is implied.
From	To	Typical load direction	What changes
Low-bar back squat	High-bar back squat	Usually lighter	Bar higher on the back, more upright torso, more knee demand
Back squat	Front squat	Lighter	Front-rack position, upright torso, upper back limits load
Straight-bar squat	Safety squat bar	Usually lighter	Bar pitches forward, upper back works harder
Competition-grip bench	Close-grip bench	Usually lighter	Longer bar path, more triceps
Touch-and-go bench	Paused bench	Lighter	Rebound off the chest removed
Conventional deadlift	Sumo deadlift	Depends on the lifter	Work shifts between hips, knees and back

One block, three logs

Run the week-seven lifter through each model. Say the last low-bar top set was 315 for 3 and the first high-bar top set is 285 for 3.

Logged under the same name, Epley reads that as roughly 347 dropping to roughly 314. A regression flag fires, and the lifter either ignores it (and learns to ignore the flag) or second-guesses a perfectly good training decision.

Logged as a new exercise, nothing fires. Nothing can. High-bar has one data point, the low-bar line stops dead, and the year-end view shows two lifts that look like strangers.

Lumped into "squat," that 33-point e1RM cliff averages into the trend line and drags it down for weeks.

Under the fork, 285 for 3 becomes the high-bar baseline. Week eight's 295 for 3 reads as progress on high-bar, which is what it is. The low-bar line stays parked at its last value, and the link records why it stopped.

The first few sessions are calibration

On a new variant, bar position, groove and depth are still settling, so early jumps reflect technique as much as strength. A plateau read on a variant with only a handful of sessions behind it is noise.

What changes when the log knows about the fork

Mostly, the lifter stops paying a tax for training well. Rotating variants is standard in conjugate-style programming and in plenty of off-season blocks. A log that punishes every rotation nudges lifters toward the worst outcome: hanging on to a variant they've outgrown because its line looks nice.

The history also starts answering better questions. Did the high-bar phase carry over when they rotated back to low-bar? That question only makes sense when both branches sit in the same family. Did the bench stall, or did the close-grip block just start lower? Same deal.

The pointer carries context and nothing else. A tracker that converts a high-bar 285 into an "equivalent" low-bar number is inventing a ratio the lifter never tested.

A log that punishes every rotation teaches lifters to keep the variant they've outgrown.

How Platepusher approaches variant history

Platepusher is built on one bet: your training history should improve today's training. A false regression on swap week works against that directly, because it teaches the lifter to distrust the chart at the moment the chart has something useful to say.

Two pieces are already in place. Plateau detection is deterministic math on each lift's own series. It flags the week a top set stops moving and shows the numbers behind the flag. And importing from another tracker brings every lift, set and date across as native data under the names they were logged with, so years of low-bar and high-bar work arrive as intact series instead of a merged blur. Platepusher does not let you link a new variant to a parent lift yourself. Built-in catalog exercises are grouped into families behind the scenes, but a custom exercise or an imported name stays a separate lift with its own history. A passive "same lift?" prompt at add time, offering to rename old logs only if you confirm, is on the roadmap but hasn't shipped, and nothing is ever merged automatically.

The target for variant swaps is the fork: each variant judged against itself, the family readable as one story, and no invented conversion between them.

What we're tracking next

Whether there's a defensible way to describe carryover between linked variants without pretending to a fixed ratio. Lifters who rotate low-bar and high-bar across a year generate the paired data themselves, and the honest version may be a per-lifter observed range rather than any published conversion. Until that's clear, the log should show both branches and let the lifter read the relationship.

Bring your training history over and see what the math says about each lift.

Platepusher's plateau detection is deterministic math on each lift's own series, and it shows the numbers behind every flag. History imported from another tracker arrives as native data, every lift, set and date, so years of variant work come across intact and keep informing today's sessions.

Get Platepusher