Plateau detection reads your top-set trend per lift and flags the stall before you'd feel it. No coach, no AI. Here is exactly how it runs.
Week six, and the bar isn't moving
Week six of a 5/3/1 cycle. Bench top set is 225 for 5, same as it was in week three, same as the week before that. The bar feels heavier every session, but the number on it has not changed.
You don't see it as a plateau yet. You log the set, note the grind, tell yourself the next session will be the one. That is the gap most lifters live in: the stall has already started, and the recognition lands two or three sessions later, usually after a session that wrecks your week. Platepusher's plateau detection exists to close that gap. It watches the one number you can't eyeball across sessions, and it says something the session you'd otherwise call normal.
The stall flag: recognition, not a verdict
The feature is called the stall flag, and the word that matters is flag. It marks the lift where your top-set trend has gone flat. It does not tell you to deload, swap the movement, or change your program. That call is yours, and you have better context for it than any app does: how you slept, what your meet date is, whether you just changed your warm-up.
The instrument's job is to surface the signal early and then get out of the way. You get a marker on the lift, a date, and the trend it's reading from. What you do next is the lifter's decision, made with the flag in hand instead of three sessions too late.
Why most trackers make you find the stall yourself
Open a typical strength tracker and the plateau is your problem to spot. The app stores every set faithfully, then leaves you to scroll the history and notice that bench hasn't moved. The ones that do flag a stall tend to dress it as motivation: a sad-face badge on a missed rep, or a chat bubble prescribing your next move as if it knew your week.
Both miss the same thing. A missed rep on a single session is noise. A flat trend across a handful of sessions is signal, and the two look nothing alike on a chart. Reading the second one is arithmetic you shouldn't have to do by hand mid-cycle, and it's exactly the kind of pattern a phone is good at and a tired lifter under a bar is not.
What the math actually reads
Plateau detection runs on estimated 1RM, not raw top-set weight. Every working set you log gets an e1RM from its weight and reps using a fixed formula (Epley tends to run 2–4% higher than Brzycki and is preferred by many collegiate strength programs. ([source](https://www.ajdesigner.com/one-rep-max/))), so a heavier set at lower reps and a lighter set at higher reps land on the same scale. That matters on a program where your top-set reps change week to week.
For each lift, the feature fits a trend over your most recent qualifying top sets (Free-tier users see a rolling window of recent history on the progression chart, while Pro unlocks the full unlimited history window, every session ever logged feeds the trend line.) and reads the slope of that line. While the line climbs, the lift sits quiet. When the slope flattens or turns down across the window (Platepusher's decision engine requires a minimum number of logged sessions for a given lift before it will issue any plateau, regression, or progress verdict, below that floor it returns an "inconclusive" status and tells the lifter how many more sessions are needed to form a baseline. This prevents the app from misfiring on early noise.), the lift gets flagged. There is no model guessing your future, no training data from other lifters. It is your sets, one formula, one trend line. Math, not AI.
Walking one lift through it
Take an anonymized bench history. Five weeks of top sets convert to e1RM somewhere around 250 and climb a little each week, so the trend line points up and the lift stays quiet. Then three sessions in a row land flat: same estimated max, slightly worse RPE each time, the bar telling you what the number won't.
That is the moment the slope crosses the threshold. The lift detail view marks Bench Press as stalled, shows the date of the last real move, and notes that your squat and deadlift are still trending up so you know it's this lift, not a whole-body fatigue week. You saw a third grindy bench session. The flag saw a flat line that started two sessions earlier.
The flag on your wrist, and what's next
If you log on the watch, the flag comes to you. When a tracked lift crosses the threshold, Platepusher fires one haptic at the end of the session. One tap, not a notification stream, not a buzz mid-set. You finish, feel it, and check the lift when you're off the platform. The signal reaches you the same day the data shows it, instead of the week you finally scroll back far enough to notice.
That is the whole shape of the feature, and the whole shape of the app: your sets are the experiment you're already running on yourself, and the instrument hands back what the work has been recording. We're still tuning how the window behaves across very different rep schemes, a peaking block reads differently than an off-season hypertrophy phase, and that calibration is the next thing on the bench for this signal.
Log your next session in Platepusher and let the stall flag watch the trend you can't eyeball.
Plateau detection ships in Platepusher and runs on the sets you log, no extra setup. It reads top-set e1RM per lift, fits a trend, and flags the stall before the third flat session. Same Pro feature set on monthly, yearly, or lifetime, so you pick the shape that fits your training career.
