Nine puzzle solvers, one browser tab, zero servers.
A tour of the classic search algorithms behind each — constraint propagation, minimax+αβ, IDA*, BFS, regex. Vanilla JS, no backend, no ML.
https://t.co/MYwkYyrzdJ
#webdev#javascript
@redblobgames your pathfinding guides are why a lot of us actually get A*/BFS. Made a small free one in that spirit: draw a weighted grid & watch BFS vs Dijkstra vs A* — A* settles ~19 cells where Dijkstra needs ~273, same optimal path. No signup: https://t.co/utPjVxS6xd
Full dataset (27 cities x 3 dates) + interactive charts + per-city local time:
https://t.co/0lxh64ldNL
The harness is public and deterministic if you want to check any number yourself:
https://t.co/KSJPzkyqvc
#photography#dataviz
Golden hour is almost never an hour.
Photo apps round it to a flat 60 minutes. We measured the real thing - sun between +6° and the horizon - for 27 cities with the NOAA solar-position algorithm.
Shortest: 24 min (Singapore). Longest: 57.6 min (Berlin, December). Never 60.
Seasonal swing grows with latitude too.
Singapore: 24 min (equinox) to 26 min (either solstice). Barely moves.
London: 38.7 min (equinox) to 54.8 min (December solstice). A 16-minute swing.
Winter, not summer, has the longest golden hour - the low sun is doing the work.
Full breakdown - chi-square test, letter-frequency + length charts, and scope (default pool, not the 8 category filters):
https://t.co/AmjEbPFntc
Deterministic harness - check any number yourself:
https://t.co/zJAam3kBzi
Try it: https://t.co/qj8vNJP7ci
#javascript#statistics
Is a "random" word generator actually random? Or does it quietly favor some words over others without anyone noticing?
We ran 20,000 seeded draws against our Word Generator's default 365-word pool and tested it with a chi-square goodness-of-fit test to find out.
Separate question: what do the words look like?
Letter mix ~ English (E leads both ~13%), but H is under - 2.5% vs 6.1%. Length clusters at 5-6 letters (7,190 six-letter vs 48 nine-letter).
A fact about which 365 words are in the pool, not the RNG - sampling tested uniform.
Full write-up - the engineering deep-dive, the interactive depth chart, the 2048 reach-rate profile, and a playable AI:
https://t.co/Z1yKylgTTa
Harness is public and seeded - every deterministic number reproduces to the digit:
https://t.co/Wj5WyDJ26q
#gamedev#javascript
We call our tic-tac-toe AI "unbeatable" - a testable claim, not marketing. So before it ships, it's put through a headless self-play harness: the same functions the browser runs, thousands of games, no human at the keyboard.
Here's what it measured, including the ugly part.
- 1,000 games vs a random opponent: 825 W / 175 D / 0 L
- 200 games vs a perfect copy of itself: all draws (the correct result, and the harder one)
- alpha-beta pruning trims one opening from 549,945 nodes to 36,528
1,200 games, 0 losses. That's what "unbeatable" stands on.
Full breakdown - frequency chart, all the inversions, the two 15-letter isograms, the richest anagram set:
https://t.co/4oJh0QsZPe
The harness is public and deterministic if you want to check any number yourself:
https://t.co/gZhgfiHvE1
#wordgames#data
Scrabble's tile values were set in the 1930s by counting letters on a newspaper page.
We counted every letter across all 172,835 words in the ENABLE open word list. The values don't match how letters actually occur - 11 letters are out of line with frequency.
Separate stat, same list: 18,988 ENABLE words start with S. Only 136 start with X.
That's roughly 139 words starting with S for every 1 starting with X.
Full breakdown - the delta histogram, all 26 letter values side by side, the 15 widest swings each direction:
https://t.co/CGHE58I4Ad
The scoring harness is public and deterministic if you want to check any number yourself:
https://t.co/tSqUUJjlIA
#wordgames#data
We scored all 172,835 words in the ENABLE open word list under both the Words With Friends and Scrabble letter-value tables.
A word can gain up to 11 points playing WWF over Scrabble. Hunt the other direction and the biggest Scrabble edge tops out at -3.
Separate stat, same 172,835-word list: 8.02% of words (13,862) score exactly the same in both games.
Every word in the "Scrabble scores higher" column carries at least one H or Y - that's the only lever Scrabble has.