What is trend weight?

An exponentially weighted moving average of your weigh-ins, where each day's trend is a blend of that day's reading and yesterday's trend. It smooths out the water, food volume and glycogen swings that make a raw scale reading close to useless day to day, and it is the number every rate calculation in Base uses.

Why does Base use trend weight instead of my actual weight?

Because day-to-day scale noise is routinely two to three pounds, and a genuine change of half a pound a week is invisible underneath it. Reading raw scale numbers means reacting to noise, which produces exactly the wrong decisions — cutting calories after a high-sodium meal, or adding them after a dehydrated morning.

How often should I weigh in?

Daily, at the same time, under the same conditions. The smoothing works better with more samples, and the cost is thirty seconds. Weekly weighing gives the average one noisy sample per week, which is not enough to separate signal from noise on any useful timescale.

Why does my trend keep moving when I have not weighed in for days?

It should not move much, but it does keep carrying forward. The average runs over calendar days with gaps carried forward rather than over the sparse list of weigh-ins, because otherwise a missed week would compress into a single step and the trend would whipsaw when you returned.

How long before a change in my rate is real?

Roughly two weeks. A weekly rate under about 0.15 pounds is not distinguishable from nothing happening given normal scale noise, so a flat-looking fortnight genuinely is flat rather than being a signal you are missing.

Trend Weight: Why the Scale Number Is Not the Signal

Why the raw number is not usable

Day-to-day bodyweight noise from water, food volume, glycogen and gut contents is routinely two to three pounds.

A well-run lean bulk moves half a pound a week. That signal sits entirely underneath the noise, which means any decision made from a single reading is a decision made from noise:

A high-sodium meal reads as fat gain, and you cut calories
A dehydrated morning reads as progress, and you add them
A carb-heavy day reads as a stall

None of those are real, and all of them are common.

The smoothing

trend[i] = α × raw[i] + (1 − α) × trend[i−1]

Each day's trend is a fraction of today's reading plus the remainder of yesterday's trend. Recent readings matter more than old ones, and the influence of any single reading decays.

That gives it the property you want: a genuinely different reading moves it a little, and a run of different readings moves it a lot.

It runs over calendar days

This detail matters more than it looks. The average runs over calendar days with gaps carried forward, not over the sparse array of weigh-ins.

Run it over the samples alone and a missed week compresses into a single step, so the trend jumps hard when you return — reacting to a fortnight of change as if it happened in one day. Running it over calendar days keeps the time axis honest.

The weight trend with raw weigh-ins behind it. The gap between the noisy points and the smooth line is the reason the trend exists rather than being a presentational choice.

Weigh daily

The single highest-return habit in the app, and it is thirty seconds.

More samples means better smoothing, and the trend converges faster on a genuine change. Weekly weighing gives the average one noisy sample per week, which is not enough to separate signal from noise on any timescale you would want to act on.

Consistency matters more than the time itself: same point in the day, same conditions, same clothing. The trend is robust to random noise and much less robust to a systematic shift, so weighing before breakfast some days and after dinner on others introduces exactly the kind of error smoothing cannot remove.

Practical detail on the screen itself in logging your bodyweight.

Reading the rate

The rate of change is the trend's slope, and it is what feeds your goal tracking and maintenance calculation.

Weekly rate	Reading
Under ~0.15 lb	Flat. Not distinguishable from nothing.
0.2–0.5 lb gain	A lean bulk working
Over ~0.5% of bodyweight	Faster than tissue can be built — the excess is fat
Falling on a bulk	Under-eating, or the surplus is being spent elsewhere

That first row is worth taking seriously. Given normal noise, a weekly slope under about 0.15 pounds is not a signal, and hunting for meaning in it produces bad decisions.

The first two weeks are not fat

Coming off a deficit or starting a bulk, the first fortnight of scale movement is substantially glycogen, its bound water, and gut contents.

Every rate-based judgement is unreliable in that window, and Base's phase logic knows it — during the refeed and stabilise phases it shows the numbers but withholds a verdict, because a rate verdict computed on glycogen is a wrong number.

Common mistakes

Reading the raw number. The trend is the signal.

Weighing weekly. One noisy sample.

Weighing at inconsistent times. Introduces a bias smoothing cannot fix.

Reacting to three days. Two weeks is the shortest window worth a decision.

Skipping a weigh-in because the number will be high. Sodium and carbohydrate move water. A high reading is information; a missing one is not.

Treating the first fortnight of a bulk as fat gain. It is mostly not, and Base declines to render a verdict there for exactly that reason.

What to do next

Weigh yourself daily for two weeks at the same time, then look at the trend line rather than the points. The difference between the two is the entire argument for doing it.