13 years of daily weight · 6 years of daily food · one report.
Where you are now
The one-sentence summary. You've had distinct loss episodes and distinct gain episodes since 2013 — losing isn't the problem, holding is. Gains last longer than losses, which is why the long-term drift has been upward.
The big picture
Trend weight (red), trend fat % (orange), and the smoothed trend of your logged daily calories (blue). Calorie data only starts April 2020. Gaps in the blue line are stretches of 14+ days with no food logs.
A theory worth testing: do you gain weight while dating and lose it while single? I pulled phase boundaries from your annotated fat-mass chart and the 2017–18 relationship you mentioned. Fat mass (trend weight × trend fat %) is the cleaner signal here and is what's plotted below.
Fat mass (kg, trend)Dating phaseSingle phaseAlmost dating
9 out of 9. Every single one of the 4 dating phases gained weight; every single one of the 5 single phases either lost weight or stayed flat. No exceptions.
Every phase, in detail
Phase
Type
Start
End
Days
Δ weight
kg/mo
Δ fat %
Δ fat mass
fat-kg/mo
Statistical significance
Metric
Dating mean
Single mean
Welch t
p-value
p = 0.001 for fat mass means there's roughly 1 in 1,000 chance of seeing this pattern by coincidence. Strong signal, though n=9 phases is small — treat as suggestive, not definitive.
Caveats worth saying out loud
Phase boundaries are approximate. Drawn from your annotated image. Dates off by weeks would shift rates slightly but not the overall direction.
Correlation, not causation. Plausible mechanisms: eating out together, shared desserts, alcohol, disrupted routines, worse sleep. Reverse causality possible too — maybe you tighten discipline after breakups.
"Single" isn't magic. Your 1,156-day Single A (2018–2021) was essentially flat (−0.03 kg/mo). Being single removes the dating-gain force; it doesn't replace it with a loss force. Discipline still has to show up.
"Almost Dating" was the worst rate (+0.87 kg/mo, n=1). Possibly early-courtship energy is especially metabolically expensive. Or just a small-sample fluke.
What this means for you: this is one of the most actionable findings in the dataset — cleaner than calories, cleaner than meal timing, cleaner than macros. If you're entering a dating phase, the data says you should pre-commit to extra discipline especially during the first 3–4 months, when the gain rate is highest.
Every loss and gain episode, 2013–2026
Detected from the trend-weight curve — an episode is a sustained move of ≥1 kg over ≥30 days between local peaks and troughs.
Type
Start
End
Days
Start→End (kg)
Δ (kg)
kg/month
Logged days
%
Logging rate is your best leading indicator
For every day in the food-logging range, I computed the rolling 28-day logging rate (fraction of days you logged food) and the trend-weight change over the next 28 days. Below 75% logging, you drift or gain. Above 75%, you reliably lose.
Avg 28-day-forward trend-weight change (kg) by logging bucket
Logging rate (prev 28d)
Days
Next 28d Δ (kg)
Counter-intuitive finding: Partial logging (1–25%) is the worst bucket — worse than not logging at all. It's the signature of "trying, failing, eating what didn't get logged."
Seasonality — May and June are your danger months
Average 30-day trend-weight change, by the month that window starts in. Computed across all 13 years.
The May–June curse. Across 13 years, you average +1.26 kg gained in May+June combined. July–October is your natural loss season. Don't start a "fresh attempt" in late spring — pre-plan disciplined logging instead.
Weekends are your blind spot
Fraction of days you logged food, by day of week, across the entire food-tracking period.
Saturday is nearly half the logging rate of Monday. If you change one behavior — log Saturdays. Combined with the previous finding that partial logging gains weight fastest, weekend drift is the most likely source of the chronic regain pattern.
The success story: July 2024 → March 2025
Your biggest sustained loss in the entire dataset. 249 days, −6.7 kg trend weight (82.9→76.2), −3.7 pp trend fat (27.7%→24.0%). Logged 181/249 days (73%). Almost entirely diet-driven — only 20 activity days in the whole period.
The three thirds
Phase
Days
Logging %
Δ weight (kg)
Avg kcal
P / C / F %
The middle third was the engine (half the total loss). When logging dropped to 37% in Third 3, the "avg kcal" looks low on paper (1,486) but weight loss stalled — the gaps hid real intake.
Meal pattern — clearly lunch-centric
Meal
Eaten on % of days
Avg kcal when eaten
Avg daily share
Minimal or no breakfast → chicken-potato-veg lunch with olive oil (~770 kcal) → protein-bar snacks → eggs-and-something dinner → protein-bar late dinner. Effective eating window ~10 AM–11 PM.
Top foods in the winning period
Food
Days eaten
Avg kcal
Avg protein (g)
Total kcal
Does skipping breakfast matter?
A common intuition: "if I fast till lunch, I'll lose faster." Or the opposite: "the morning cocoa gets my day started." The data has a clear answer once logging quality is controlled for.
Controlled comparison — only "thorough log" days
Days inside 7+ day consecutive logged streaks, with ≥5 items logged that day. Both groups are comparably well-tracked, so the bias mostly cancels.
Metric
Morning food
Fasted morning
Δ
Streak-level Pearson r between morning-eating frequency (%) and weekly weight change: −0.10 (p = 0.69) — statistically indistinguishable from zero.
The interesting finding: near-perfect compensation. Morning food adds ~278 kcal — but lunch, afternoon snack, dinner and late dinner are collectively ~251 kcal smaller on those days. Net difference in total intake: just +27 kcal. Your body self-regulates hunger at the day level. The morning-food-vs-fasted choice isn't a calorie decision — it's an adherence decision. Pick whichever makes the rest of the day easier to log and eat consistently.
What about the Bellarom Chocolatto + coffee ritual specifically
Within clean-log days, split the morning-food group by whether Bellarom was part of the morning.
Subgroup
n
Total kcal
Morning kcal
Lunch kcal
Next 14d Δ
Next 28d Δ
Bellarom-morning days show slightly worse 28-day loss than either non-Bellarom morning food or fasting (n=35, so noisy). Not strong enough to rule against it — but also no evidence it helps. Possible interpretation: Bellarom shows up on "I'll have my cocoa and that's my effort today" days.
The naive-vs-controlled gap is exactly the bias you suspected
The uncontrolled view (all 687 real-food-logged days) shows morning-food days losing noticeably faster. That's because morning-food days are also days where you're tracking thoroughly in general — selection bias, not a real effect.
View
Morning food Δ 28d
Fasted Δ 28d
Apparent effect
All multi-day logged streaks, ranked
Every run of 5+ consecutive logged days (sorted by total weight change). These are the "clean windows" where food data and weight data agree.
Start
End
Days
Start→End (kg)
Δ (kg)
kg/wk
Avg kcal
Protein %
g/kg
Items/day
Streak-level correlations
Your top 20 kcal contributors, all time
Sorted by total calories consumed. "Protein density" = protein-kcal as a share of food kcal — higher is better. Flagged in red: calorie-dense items with near-zero protein.
Food
Times eaten
Total kcal
Avg kcal / time
Protein density
The one true whale to cut:Costa velké ice latte karamel — 36 occurrences, ~9,750 total kcal, 3 g protein average. Pure liquid sugar-fat. If you cut or halve this one item, you've bought yourself ~5,000 kcal of annual budget back.
What to actually do
Target 75%+ logging always. That's the statistically validated threshold where your body reliably loses weight. Treat anything below 75% as a red flag that a gain cycle is starting.
Log Saturdays specifically. Your single biggest blind spot — the one change with the highest leverage.
Pre-commit May–June discipline. Historically your two worst months. Don't start attempts in late spring — maintain existing discipline instead.
Drop or halve the Costa ice latte. The one item the data flags as pure cost, no benefit.
Reproduce the Jul 2024–Mar 2025 template. Skip/light breakfast → 700–800 kcal lunch of chicken-potato-veg with olive oil → protein-bar snacks → light dinner. ~1,700 kcal, ~1.2 g protein/kg, 25% protein share. Your own data says this works.
Use the current moment as the alarm. Jan–Mar 2026 you logged 100% and lost 1.8 kg in Feb. April logging just slipped; trend reversed to +0.09 kg/wk. You know what happens next if you let it.