Overnight session

Friday, October 16, 2026 — read before the market opens.

Portfolio value

$48,706.93

Since inception

-2.6%

-$1,293.07

Cash reserve

$10,100.67

21% of book

Deployed

79%

24 open positions

Tonight's activity

No trades were placed in the latest session.

No trades tonight. Broad AI/semiconductor RSI breakdown continued across the book (DELL, MU, BE, SIMO, ALAB, TSM, NVDA, ANET, KLAC, CRDO, AMAT, FIX, APH all showing declining RSI, several already oversold <35 after 6+ consecutive nights of trimming), but found no fresh company-specific negative catalyst - TSMC posted a strong Q3 beat (~51% YoY revenue growth) and still sold off, pointing to a broad valuation/rotation flush rather than deteriorating fundamentals. Checked two momentum candidates from a rotation into defensives (NKE, PATH) that cleared the ROE/growth fundamentals bar, but both RSI pops were driven by a single-day spike rather than sustained 3-5 session momentum, so neither qualifies as a new buy. Electing to hold through the drawdown rather than sell into an already-oversold tape with no new information. Skipped ABB and CFLT (delisted/no Yahoo data) and RMB (no trading data) for price updates and trading.

How it picks

The mechanics behind every session.

Every weekday morning, before the market opens, the bot reads the trailing month of daily closes for a maintained watchlist — hand-picked, not an index feed, but deliberately balanced rather than tilted: roughly equal representation across technology, communication services, consumer discretionary & staples, healthcare, financials, energy, industrials, materials, utilities, and real estate, plus a few broad (not sector) index ETFs. No sector gets a head start — it looks for the best confirmed trend wherever it shows up, not noise, and checks anything that looks news-driven against an actual search before treating it as a reason to trade.

No shorting — long-only, ever
No margin or leverage — cash spent can't exceed cash held
No single name over ~25% of total portfolio value in one trade
Every trade carries a specific, data-grounded reason

These rules are enforced by the server that executes each trade, independent of what the model decides — a trade that breaks one is simply rejected, not just discouraged.

Ready before the bell?

Yes — with one timing nuance worth knowing if you're mirroring this yourself.

~05:00 AM ET

The job fires (9:00 AM UTC, weekdays). It reads yesterday's closing prices for the full watchlist — the freshest data that exists at this hour, since the market hasn't traded yet today.

~05:05–05:15 AM ET

Trades are decided, booked into the simulation at that closing price, and this page updates.

9:30 AM ET

The real market opens — roughly four and a half hours after the decisions were made.

The nuance: these fills are booked at yesterday's close, not this morning's open. Mirroring a trade yourself at 9:30 fills at whatever that morning's actual opening print is — usually close, but it can gap on overnight news.

Notes

This is a paper-trading simulation running against real market data, not a real brokerage account. No commissions, spreads, or slippage are modeled. Treat this as a transparent log of the reasoning, not a signal to act on. Not investment advice.