EN
Training
“When will I get there?”: a date drawn from your own logs
Target weight, next strength level, race time: how Azetta draws the trend of your last 12 weeks with a robust method (Theil–Sen), sets aside an isolated log, gives a date in words with its range, and would rather say "not yet" than make one up.
7 min read Updated October 9, 2026
The principle
"When will I get there?" is a question often answered with a promise. Azetta answers it with your own logs: it looks at how they have changed over the last 12 weeks, extends that trend to your goal and gives you an approximate date, with a range. When there aren't enough logs, or when they are going the other way, the app says so instead of making up a date.
It's a projection of your current pace, not a prediction of the future: if your pace changes, the date changes.
The Azetta method
1. Three goals
| Goal | What is measured | Minimum for a date |
|---|---|---|
| Target weight (your profile) | the average of each day's weigh-ins | 5 weigh-ins over at least 3 weeks |
| Next strength level | your strength score at the end of each week | 5 weeks spanning at least 4 weeks |
| Race time (your race plan, before race day) | your best run of each week, converted to a time over your distance | 4 weeks spanning at least 3 weeks |
Only logs from the last 84 days count. The strength level aimed for is the one after yours (Novice at 200 points, Proficient at 350, Advanced at 500, Expert at 650, Champion at 800, Titan at 900: see the article on the strength score).
For running, each run of at least 3 km (hikes excluded) becomes a 10 km time using Riegel's formula (exponent 1.06), reduced by 4% because a training run is slower than a race effort, then converted to your race distance. A 10 km time outside 27 minutes to 2 h 30 min is ignored (a walk, an input error).
2. The trend: Theil–Sen
Instead of the least-squares line, which the slightest odd log can tip over, the app draws a Theil–Sen line:
A median doesn't follow an isolated value — a weigh-in after a very salty meal, a run cut short —: the line keeps what most of your logs say.
3. Set-aside logs
The app then computes how far each log lies from this line. A log whose gap exceeds three times the median absolute deviation (multiplied by 1.4826) and a floor is set aside, and the line is redrawn without it:
| Goal | Floor |
|---|---|
| Weight | 1 kg (a water day) |
| Strength score | 15 points |
| Race time | 60 seconds |
Set-aside logs stay on the chart, crossed out. A log is never set aside if that would drop below the minimum number of logs.
4. The range
The slope comes with an interval, Sen's, at about 80%: among all the slopes between pairs, sorted, the app takes those of rank (N − C) ÷ 2 and (N + C) ÷ 2, where N is the number of slopes and n the number of logs:
The fastest slope of the interval gives the earliest date, the slowest the latest date. The steadier your logs, the narrower the range. If the slowest slope doesn't reach your goal within two years, the latest date is left open rather than made up.
5. What the app tells you
| Situation | Message |
|---|---|
| Not enough logs | how many are missing, or the date of the first estimate |
| Trend already at the goal | goal reached |
| Flat trend: less than 50 g, 1 point or 2 seconds a week | trend flat, no date |
| Trend moving away at least as fast | trend going the other way, with its pace, no date |
| Goal more than two years away | at this pace, over two years: no date |
| Otherwise | a date and its range |
The date is given in words: "early" (1st to 10th), "mid-" (11th to 20th) or "late" (21st to the end of) the month. For running, the app also says whether that date falls before or after your race.
6. For weight, two more markers
- A loss of more than 1% of your weight a week is flagged as fast.
- With at least 10 complete days of logged meals over the last 4 weeks (a day logged at less than half your expenditure is left out), the app adds what your calorie balance says:
It's a rough estimate, presented as such: the 7700 kcal per kilo rule is an approximation, and your expenditure is itself estimated.
Why this choice
- Theil–Sen is a classic robust trend method (Theil, 1950; Sen, 1968): because it relies on medians, a few aberrant logs don't move it, whereas a single wrong weigh-in can tip over a least-squares line.
- The median absolute deviation is recommended for spotting isolated logs rather than the standard deviation, which outliers themselves inflate (Leys et al., 2013).
- Sen's interval gives a range without assuming that your logs follow any particular distribution.
- Not giving a date when the trend is flat, reversed or too far off avoids turning noise into a promise.
- The 7700 kcal rule is kept for a simple "what if", with a warning: real weight loss slows over time and drifts away from this rule (Hall et al., 2011).
Limits
- It's a straight line: weight, strength and speed rarely progress at the same pace all the way; progress often slows as you approach a goal.
- Sparse or very irregular logs give a wide range, or no date at all.
- For running, converting a run into a race time is an estimate; a very hilly or very windy run counts like the others (unless it is set aside).
- The app only knows what you log: forgotten weigh-ins or unrecorded sessions change the trend.
References
- Theil H. A rank-invariant method of linear and polynomial regression analysis, I, II, III. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen, 1950;53.
- Sen PK. Estimates of the regression coefficient based on Kendall's tau. Journal of the American Statistical Association, 1968;63(324):1379-1389.
- Leys C, Ley C, Klein O, Bernard P, Licata L. Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 2013;49(4):764-766.
- Hall KD, Sacks G, Chandramohan D, Chow CC, Wang YC, Gortmaker SL, Swinburn BA. Quantification of the effect of energy imbalance on bodyweight. The Lancet, 2011;378(9793):826-837.
- Riegel PS. Athletic records and human endurance. American Scientist, 1981;69(3):285-290.