YNBrainStock LIVE
YN Finance
The forecaster

A neural net calls the market.
And grades itself in public.

BrainStock forecasts hundreds of stocks every market morning, then scores every call against real closing prices, a real backpropagation network building an un-cherry-picked track record.

11→16→12→1
Architecture
--
Examples trained
warming up
Directional acc.
BOOTSTRAPPING
Status
Step 1 · Today's call

Point it at any ticker. Watch it forecast.

This is the same engine that runs the morning board, live, on demand, for any name you type.

$
HORIZON5d
TRY
ENGINEbuild your own robot & use it here →
Step 2 · The brain's book

We handed the brain $10,000.

Every graded call becomes an equal-weight position. The wins and the misses compound below, marked to real closes. Nothing hidden, nothing reset.

🥊 The Arena of Ghosts · watch the Brain fight Wall Street →

The first calls are in flight, the scoreboard fills itself as they resolve.

Open the full track record
Step 3 · How it works

The net, and the grading loop.

No black box. Eleven live market features feed a real multi-layer network; every prediction is logged, then re-graded against what actually happened.

01
Eleven features
Trend, momentum, volatility and more, extracted live from each ticker’s price action.
02
A real forward pass
Backpropagation through an 11→16→12→1 network, not a chart trick. Watch it fire below.
03
Graded in public
Each call is scored against real closes, the track record builds itself, un-cherry-picked.
Step 4 · Losses, absorbed

Every miss gets an autopsy.
Every night, a lesson.

Most models hide their losers. This one opens them up: the stored evidence behind each missed call, and the nightly training run where the net absorbs every graded outcome. The flywheel, on the floor, in public.

The training floorNightly · after grading

The first training run prints here after tonight's grading pass. The net only learns from calls that have been graded against real closes, so the floor opens when the receipts do.

Second desk · the book net

A second net is in school: it learns P(drawdown) from every Risk Desk scan once the book's 63 sessions run out. 0 of 40 books matured. Feed it your book →

Loss autopsiesMisses only · stored evidence

pulling the case files…

Keep going

The receipts, and the games.

Educational research tool. Forecasts are model estimates, not financial advice. Past performance does not guarantee future results.