The night market notebook that guessed missing trades
Under the lanterns, I ran the bracelet swap table with a grubby notebook. A gust flipped pages and a few lines smeared into nothing. I still had two trails to follow: who swapped with whom, and the look and feel of each bracelet on each wrist.
The trouble was, most people never swapped with most others, so the notebook was mainly blank space. Guessing from blanks felt like guessing from silence. Old shortcuts usually stared at the swap lines or the bracelet details, but not both, and they acted sure even when the notes were thin.
So I tried a newer trick. I gave each person a hidden profile like a pencil sketch: a sharp centre and a soft blur to show doubt. I built it by passing hints twice, from you to your swap partners, then through their partners too, so your sketch carried your bracelet and your nearby crowd.
To rebuild a lost line, I compared two sketches. If they leaned the same way, a swap looked more likely; if not, less likely, and I squeezed it into a clean chance from no to yes. I also let each person “swap with themselves” once, so their own bracelet didn’t get drowned out. And I treated real swaps as extra important, because “no swap” is the easy answer for almost every pair.
There was a simpler version as well: give everyone one fixed sketch with no blur, then fill the notebook from that. It can work nicely. But when the notebook is sparse and smudged, the blurry version helps because it can admit, in effect, “I’m not sure about this person yet.”
Using both the swap links and the bracelet details, the rebuilt notebook got better at putting true missing swaps above random non-swaps in different crowds. Even without bracelet details, it held up well against older swap-only guessing. I looked at the clean pages we’d saved and thought: the big change was letting the notebook say how sure it was, not just what it guessed.