Eleven columns in my log and only three track my losses

Road warrior here — often 100+ hotel nights a year. My 12th column would be a boring binary: travel vs home. That one predicted my 7-day rolling average better than protein or water, mostly because airport/hotel sodium and weird meal timing mask progress for 2–3 days. I’d also split “ate past 9pm” into home vs on the road — for me only the travel version mattered. Which three columns are winning for you? I’m betting sleep and lifting, but I’m weirdly curious if stairs/elevator snuck in.

Curious which three? My guess is steps, RHR, and the Sunday tape — the slow movers. One thing that changed my whole log: I stopped correlating columns against the 7-day average level and started correlating against the week-over-week change in that average. Level comparisons mostly surface anything that drifts alongside weight and hide what actually precedes a move. After the flip, my only real predictor was the 9pm flag — and it shows up two days later, not the next morning. Water and mood have been pure noise for me, eleven months running. Which three came out on top for you?

I’ve got a similarly crowded log (though only 8 columns), and the thing that surprised me was lag. My 7-day weight average didn’t move with same-day anything; it followed a 3-day-lagged sleep score and a 10-day-lagged waist tape. Also, eyeballed protein was too noisy to rank, so I switched to a weekly “protein hit days” count and that finally lined up with my losses. Gentle question: when you compared columns, did you test lagged versions, or only same-day/week? I’d be curious whether your three winners still win at a 2–3 day delay.

Three out of eleven sounds about right. I kept a similar log after my kids moved out, and the two things that actually nudged my 7-day average were boring: a 20-minute walk after supper and pre-chopping vegetables on Sunday. Steps alone didn’t do it because I’d bank 8k running errands and call it a day. Which three columns won for you? I’m guessing sleep and lifting, but I’m curious if the 9pm one beat them.

this is the kind of nerd rabbit hole i live for. which three popped? i ran a similar thing with soda-swaps and the only columns that moved my 7-day avg were sleep, steps, and whether i ate after 9. the weird one: late-night gaming snacks didn’t show up same day, they showed up 2–3 days later as a water bump, so a straight correlation made them look useless. try shifting each column by 1–3 days against the rolling avg and see if new stuff lights up. also how many days before you trust a column? 209 rows is solid but i still feel like under a month is noise.

Which three surprised you? In my log, the only things that ever moved my rolling average were sleep consistency, a 20-minute walk after dinner, and waist tape lagged by about five days. Steps alone looked useless until I split weekday vs. weekend. Also, eyeballed protein was my noisiest column—switching to a weekly average made it line up. Have you tried lagging each column a few days? For me, a salty dinner showed up in weight 2–4 days later, which can make a real signal look like noise.

Empty nester here too, and I had the same shock — my 7-day average barely flinched for anything I logged same-week. What finally lined up was a lag: steps and home-cooked dinners only showed up on the scale two to three weeks later. Same numbers, shifted window, completely different picture.

Curious which three columns actually earned their keep, and whether you tried lagging them at all.

Also, stairs vs elevator is genius. My version was "ate standing at the counter," which tracked my mood column almost exactly — so half my log was really just measuring the same Tuesday twice.

Nice to see someone else with a spreadsheet habit. One thing I’d check before trusting the three: lag. When I did this in

Curious which three columns came out on top. With 209 rows you've got decent power, but a 7-day rolling average on daily weight will autocorrelate like crazy—any r below ~0.3 is probably noise. Also, eyeballed protein is basically a random number generator. Have you tried lagging variables? Sleep might hit weight two days later. And waist tape only Sundays gives you ~30 points, so that correlation is way weaker than it looks. What r values are you seeing?

Which three? I ran the same nerd experiment on ~180 rows and the only columns that actually moved my 7-day average were protein and Sunday waist—but waist lagged weight by about 10 days. Steps and sleep looked useless same-week, then showed up when I shifted them 2 weeks back. If you haven’t, try lagging each column 7/14/21 days against the rolling average. My “ate past 9pm” flag was noise until I split weeknights vs weekends, then it tracked hard. What are your three? Curious if lifting is one, because for me it masked fat loss for a solid month.

One thing I started noting in my dorm log: “dining hall actually had chicken breast? y/n.” When it’s yes, my protein number is an estimate; when it’s no, I’m eating cereal for dinner and calling it intuitive. I also scribbled “sodium-ish”—ramen broth, soy sauce, gravy—because the scale jumped 2 lbs overnight and then vanished, which past-me treated like a personal betrayal. Which three columns actually correlated for you? I’m nosy and need to steal them.

cafe owner here—my 12th column would be “sat down for my own meal.” Not what I ate, just whether I ate it standing over the sink between rushes. That one tracks my evening picking way better than steps do. Also, did you test lag? My sleep and mood shifts show up two or three days later in the 7-day average, so lining them up same-day made them look useless. Which three actually tracked your losses?

I’d bet two of your three are lagging. When I was doing 3–4 hotel nights a week, same-day lining-up made everything look flat: a late airport dinner or crummy sleep didn’t hit my 7-day average until 48ish hours later. Try shifting sleep, late eating, and water by 1–3 days and see if the correlations firm up. I also added a dumb column: “real meal vs scavenged terminal food.” That one tracks way harder than steps for me—hotel treadmill steps and concourse steps are not the same animal. Did you test any lags, or only same-day?

Which three? I ran the same correlation on ~140 rows and got humbled. My steps column looked like a hero until I realized it was just a proxy for my work schedule. What actually tracked my 7-day average was the timestamp of my last bite of the day — later than 8pm and the average stubbornly flatlined for the next four or five days. Sleep hours moved it maybe a third of a pound. Were your three the ones you'd have guessed, or the ones you almost didn't bother logging?