
Weeks 11 & 12 of building Rookify — when the data finally started talking.
After hitting 95% in acceptance tests, I ran Rookify’s first full ELO validation:
At first only 1/60 skills showed correlation
Now we're seeing 16 strong & 13 moderate correlations.
But here some interesting insights, I've picked up:
1. Chess improvement isn’t always linear.
2. Some skills scale steadily (like blunder rate).
3. Others evolve in leaps (like positional understanding).
Next up, I plan to test non-linear correlation models (Spearman/segmented fits).
🧠 Full deep dive → open.substack.com/pub/vibecoding…
🎯 Try Explore Mode → rookify.io/app/explore
#buildinpublic #chess #ai #datascience #vibecoding

English




