Good progress! I understand blurring the y-axis, though I'm curious about how large those up/down swings are. Have you considered using an exponentially-weighted average to smooth out the noise in the measurements?
It's just $ f(x_n) = f(x_{n-1}) * (1 - alpha) + x_n * alpha , f(x_0) = x_0 $ where 0 < alpha < 1 (lower values give a smoother graph but lags more)
Here's a quick sketch of how to do it in gnuplot without having to preprocess the data:
# Getting previous values via https://stackoverflow.com/a/25753509/5921362
alpha = 0.1
prev = 0
shift(x) = (prev = x)
ewma(prev, cur) = shift(prev * (1 - alpha) + cur * alpha)
plot data using 1:2 with p,\
data using 1:(prev < 1 ? shift($2) : ewma(prev, $2)) with l
Since I'm really not that careful during non-diet days (some foods such as home-made chicken pot pies or pizzas etc.) can result in "significant" gains (2-3%) and other items my gf sometimes brings over (potato chips; ice cream; ...). Short answer: that week-to-week swing ranges ~4.5% of overall body weight. I try to limit (within reason) processed synthetic fake foods like margarine, Canola (lifeless, NaOH/heat-treated, hexane-extracted, bleached, deoderized "food") in favor of known-sourced, i.e. non-fake EVOO (look for the oleocanthal-induced burn at the back of the throat: http://www.sciencedaily.com/releases/2011/01/110118180514.ht...), butter, etc. I make my own bread, yogurt -- again for control of what goes into them. But overall, not obsessed with any of that. I like to keep it simple; those preparations take minutes not hours ... :-p
Edit: thank you for the plotting comments (I heart these: I love all things tech, programmatic, bio*); I just wanted a quick way to see my overall progress over time. I enter my morning weight (digital scale) daily, and occasionally plot it (Ctrl-Alt-T shortcut to terminal; "gp" bashrc alias to run that plot command, "qq" to quit the terminal). ;-)
It's just $ f(x_n) = f(x_{n-1}) * (1 - alpha) + x_n * alpha , f(x_0) = x_0 $ where 0 < alpha < 1 (lower values give a smoother graph but lags more)
Here's a quick sketch of how to do it in gnuplot without having to preprocess the data: