The whole system, measured daily.
Every number below was measured directly from the live production database at 2026-11-21T20:45 IST — no edits, no rounding up, nothing invented. Where grading is still pending, we write "grading pending" instead of a flattering guess. Yesterday's full picture:
of 11,619 graded predictions yesterday (5,521 right). 3,288 still pending grading — no claims made for these.
registered accounts; 2 in the last 7 days; 6 on a live paid plan.
published. Signal-quality filter blocked 355; Market-risk filter blocked 169.
The rest of the measured picture
| Metric | Measured value |
|---|---|
| Weekend signals (total / last 7d) | 60 / 14 |
| School leads (total) · all leads | 4 · 7 |
| Retrain status | Model retrain: weekly cycle, last calibration day 2026-11-20 |
| Dead models (don't beat own baseline) | 193 of 224 5-minute models do not beat their own majority baseline |
| Confidence honesty score | raw 0.3449 → calibrated 0.013 (fit 2026-11-17) |
Galtiyan / verdicts of the week
Week of 2026-11-16: 1 audits — READY 0, MARGINAL 0, NOT_READY 1; avg readiness score 20.0
Lowest-readiness verdict reason, quoted verbatim: "System scored 20.0/100. Performance does not meet the bar for live trading. Win rate 42.4% and/or P&L ₹13,648 need improvement."
Why this page exists
Because trust should not require believing anyone. This report regenerates every evening at 20:30 IST from the same database the platform runs on. The long record — every prediction ever made, graded — lives on the transparency page. Nothing here is chosen for looking good; if a number is missing, it stays missing.
REPORT DATE 2026-11-20 · MEASURED FROM LIVE DB · AUTO-GENERATED (scripts/generate_daily_report.py) · KOI NUMBER INVENT NAHI