Live production endpoint

Does the ML model actually work?

This page calls the real birth-time prediction API for 100 famous people whose exact birth times are publicly recorded. For each one it checks a simple, honest question: did the model's predicted time window actually contain their true birth time? No mock data, no cherry-picking.

Use this API in your app Try the predictor
0Tested
—Window Hits
—Top-1 Accuracy
—Any-Window Acc.
—Mean Hit Rank
—Random Baseline
Live Classifier
idle
waiting
Predicted birth-time windows across the day · needle = real birth time
Window-Hit Accuracy
—hit rate
Where the truth landed
Per-Person Results 0
# Name Field Real Time Top Predicted Window Verdict
How this test works (and why it's honest)
  1. We load 100 famous people whose birth times are publicly recorded (most with high "AA"/"A" Rodden ratings).
  2. For each, we send their birth date + place + 9 body features to the real production endpoint FindBirthTimeByMachineLearningTopK.
  3. The model never sees the true time — only the date scans the whole day and returns ranked candidate windows.
  4. A hit = the person's real, known birth time falls inside one of the predicted windows. "Top-1" means it landed in the #1-ranked window.
Read the numbers fairly. Compare Any-Window Accuracy against the Random Baseline — the baseline is the share of the day the predicted windows happen to cover, i.e. the odds a random guess would "hit". The model is only doing real work when it beats that line. Finer precision = narrower windows = a tougher, more meaningful test.
The model is built from classical Vedic body rules (BPHS / B.V. Raman), not fitted to this celebrity set — so this is an out-of-sample, live benchmark, not a memorised score.
Advanced settings
Window scan resolution. Smaller = narrower windows = harder test.