You weigh yourself every morning for four weeks and sit at 80.0 kg (176.4 lb). Then the scale shows 79.0 kg (174.2 lb) two days in a row, and you wonder whether a loss has started. Whether your trend weight answers that after three days or after two weeks depends on the calculation behind it.
For most people, the exponential smoothing that Trendcurve uses by default fits. It is calm and stable. If you want faster feedback, pick the 7 day average or double exponential smoothing. If you want a number you can check by hand, pick one of the two averages. The choice trades calm against speed, and you can change it at any time.
The principle: noise against lag
Every method has two jobs. It should filter out daily swings from water, salt and meals (more in Why your weight fluctuates daily). And it should show real change soon enough to be useful. You cannot fully have both. Stronger smoothing gives a calmer line that trails behind a real change. Weaker smoothing shows change earlier, along with more wobble.
A study by Turicchi et al. 2020 found that in smart scale data an exponentially weighted moving average agreed well with real values. That suggests the default is a sensible choice, but it does not prove it is the best one for you.
The four methods in plain words
Exponential smoothing (default). Each new daily value counts for 10 percent, the previous trend for 90 percent. Older values fade evenly, so no edge exists where a value suddenly drops out. For the first few days Trendcurve uses a running average so the trend does not cling to your first entry. A change arrives about halfway after one week.
Moving average over 7 days. The trend is the mean of the last seven days. It reacts quickly and cancels a full weekly pattern. Its drawback: an outlier counts for seven days and then drops out of the window at once, which makes the line jump.
Moving average over 14 days. The same calculation with a window twice as long. The trend is calmer and easiest to understand, but it trails a change the longest. It also jumps when an old value leaves the window.
Double exponential smoothing (Holt). This method estimates not only the level but also the direction of the trend and carries it forward. With a steady loss it therefore sits closest to reality. Trendcurve weights the level at 15 percent and the direction at 10 percent, and damps the direction by 0.95 so it does not run on forever. Across gaps without a measurement the method continues the trend. At a change of direction it can briefly overshoot.
| Method | Reaction to change | Calm with outliers |
|---|---|---|
| Exponential (default) | medium | high |
| 7 day average | fast | medium, jump after 7 days |
| 14 day average | slow | high, jump after 14 days |
| Double exponential (Holt) | fastest with a steady course | lower |
A worked example with fixed numbers
Say your weight sits at exactly 80.0 kg for four weeks. Then your readings drop to 79.0 kg and stay there. This is how much of that drop each trend shows after seven and after fourteen days:
| Method | Trend after 7 days | Trend after 14 days |
|---|---|---|
| Exponential | 79.5 kg | 79.2 kg |
| 7 day average | 79.0 kg | 79.0 kg |
| 14 day average | 79.5 kg | 79.0 kg |
| Holt | 79.2 kg | 78.9 kg |
The 7 day average has taken on the new value completely after one week. Exponential smoothing needs clearly longer and still shows a gap of 0.2 kg (0.4 lb) after 14 days. Holt even ends slightly below, an example of the overshoot.
Now the opposite case: one morning you measure 81.5 kg (179.7 lb), 1.5 kg (3.3 lb) above your level. On that day the exponential trend rises to 80.15 kg, the 14 day average to 80.11 kg, the 7 day average to 80.21 kg and Holt to 80.22 kg. No method takes the outlier seriously, and the differences are tenths of a kilo.
A steady loss of 0.1 kg per day, which is 0.7 kg (1.5 lb) per week, shows the price of calm. After four weeks the real value is 77.2 kg (170.2 lb). Exponential smoothing then stands at 78.05 kg, the 14 day average at 77.85 kg, the 7 day average at 77.5 kg and Holt at 77.4 kg. The stronger the smoothing, the larger the gap to the real value.
How to switch the method in Trendcurve
Open "Settings & more" and then "Trend algorithm". All four methods are listed there with a description, pros and cons, and today's trend value that each method produces from your own measurements. Use that comparison instead of choosing by feel: if all four sit within a few hundred grams, the choice hardly matters for you.
After a switch the app recalculates everything. Your measurements stay untouched. The labels on the trend card, in the history and on the share image name the chosen method, so a shared image is not mistaken for a different calculation. A trend restart (Pro) applies to all four methods. The setting is part of the backup.
The feature is part of the free version. You need Pro neither for switching nor for comparing the four values.
Which method suits which situation
- You want a calm line and weigh daily: exponential smoothing, the default.
- You weigh irregularly or have gaps: exponential smoothing, because it copes well with gaps.
- You want to follow the calculation in your head: the 7 or 14 day average. How that works is covered in Trend weight and the moving average.
- You track a steady loss or gain and want early feedback: Holt.
Limits of all four methods
No method recognises the cause of a change. A trend that has risen by 0.5 kg (1.1 lb) can be fat, muscle or water, and the maths does not decide that. Every method also needs weigh-in days: with only two or three entries a week, any smoothing gets less precise and the start-up phase lasts longer.
Switching the method changes the displayed trend value, not your body weight. A trend that shows 79.5 kg with one method and 79.2 kg with another is not an improvement. Always compare weeks using the same method.
So pick one method and leave it unchanged for at least two to three weeks before you judge whether it suits you.