eats
Everywhere I've eaten and ranked, mapped, plus an honest attempt to model my own taste.
01 · the pattern
Two quick looks
Each dot is a place, plotted by where I ranked it within its category (100 = my favourite, 0 = my least). The line is a smoothed trend with a 95% band. On the left: do the notes I write match where I rank things? On the right: has my taste drifted since I started logging?
Note warmth vs. rank
Rank vs. when I logged it
02 · where
Everywhere I've ranked
Every circle is a city. Size is how many places I've ranked there; colour is how well they did, the average rank percentile of my picks (cities with three or more). Dashed rings are my want-to-try list. Click a circle for the highlights.
City-level only: the Beli export has no street addresses, so 200-odd New York places share one circle. Scroll-zoom turns on after you click the map.
03 · home turf
Where I eat, and where I want to
Solid bars are places I've ranked; dashed bars are bookmarked places I haven't tried yet.
The axis is cut short so the next cities are readable; bars with a slash run longer than shown (the label has the real number).
04 · the log
When I add things
New entries per month, stacked by category. A place is added when I rank it, so this is a log of when I rated things, which is close to, but not exactly, when I ate there.
05 · the list
My rankings
06 · the model
What lines up with where I rank a place?
Beli ranks each category in order, so every category has its own #1. I turned each rank into a percentile score (100 = my favourite, 0 = my least favourite) and asked what lines up with it. It's exploratory: about 430 data points, observational, nothing causal. Dots are effects in percentile points with 95% confidence intervals; fuchsia means the interval clears zero.
Effect sizes, all at once
Words that show up in my notes
Words that show up more in my notes on places I rank in my top third (left) than my bottom third, and the reverse (right). Bars show how lopsided the word is.
How predictable am I?
07 · one more thing