1) Specimen Intake
Upload CSV, confirm the 7 number columns, then run training. Defaults are safe. UI stays uncluttered; deep knobs are tucked away.
Tip: upload last ~1000 rows for speed.
Expected: Lotto Max draws with 7 ascending numbers per row (1..50). Any extra columns are okay.
Column mapping (auto-detected, editable)
Format: name(index). We’ll auto-pick 7 numeric columns if possible.
Example: 2,3,4,5,6,7,8. Must resolve to exactly 7 columns.
Advanced training knobs (optional)
Bigger = weights swing faster. For noisy data, keep modest.
Higher = “recent” matters more. Controls tendency & hit EMAs.
Grid-search α inside each window per position (cheap).
If P(under)−P(over) exceeds θ → nudge +1, and vice versa.
Uniqueness can distort position meaning; optional.
Keeps your original global-weights Monster composite + Pure Raw.
Left = cinematic updates. Right = near-instant (still safe on mobile).
Academic disclaimer: This is a forecasting/backtesting sandbox. Lottery draws are effectively random; no model meaningfully changes real odds.
3) Lab Journal (last cycles)
Readable, non-spammy log. No walls of text. If you need forensic detail, export CSV.
shows last 8 cycles
No data yet.