PLATEAU

Paste your logged lift numbers (date, exercise, weight, reps) and see which lifts have actually stalled, from a rolling-regression slope, not vibes.

Training log

Estimated 1RM trend

Lift status

Session log

What Plateau does

Plateau is a plateau detector for people who lift and log their training. You paste the numbers you already keep in a spreadsheet or notes app (date, exercise, weight, reps) and it tells you which lifts have genuinely stalled and which are still moving. The point is to replace a guess with a test. Comparing this week's top set to last week's is a single noisy sample: sleep, stress, bar speed, and rep quality push your estimated one-rep max around enough that one comparison produces both false alarms and false comfort.

Instead, Plateau converts each session to an estimated one-rep max (Epley formula), fits a rolling linear regression over each exercise's recent history, and checks whether the slope is statistically distinguishable from zero at 90% confidence. If zero sits inside that confidence interval, the lift is flat within its own noise and gets flagged as plateaued, with a count of how many weeks the flat window spans. If the interval clears zero, the lift is trending up or down. Lifts with fewer than four logged sessions are marked insufficient data rather than guessed at.

How to use it

  1. Copy your log as one session per line: date, exercise, weight, reps. Commas or tabs both work, and a header row is fine.
  2. Paste it into the box and press Analyze, or click "Try it now" to load a sample.
  3. Read the badges: each exercise is labeled stalled, trending up, trending down, or insufficient data. Click a badge to focus the chart on that lift and see the scatter, the fitted trend line, and its confidence band.

FAQ

What log format do I need?
Four fields per line: date, exercise name, weight, and reps. Dates can be YYYY-MM-DD or MM/DD/YYYY. Rows that don't parse are skipped and reported, so a couple of typos won't block the rest of your log.
How many sessions before it can call a plateau?
At least four sessions for one exercise. The regression uses a trailing window of up to six sessions, so more history gives a tighter confidence interval and a more confident call.
Does my training data get uploaded anywhere?
No. The whole analysis runs in your browser through Pyodide. Nothing you paste is sent to a server, and the only thing stored is your last log in this browser's local storage so the page can restore it on reload.
Why estimated one-rep max instead of the raw weight?
It puts sets of different rep counts on one scale, so a triple at 300 and a set of five at 275 are comparable points on the same trend line rather than two numbers you can't line up.