How it works

The board uses fixed rules to turn analyst research notes into stock lists. A language model reads each report and records, per stock, whether the view is bullish or bearish. The rules then choose the names; nobody edits the lists by hand. Each position is tracked against QQQ. The default board shows a retrospective research-date simulation; the recorded issuance ledger remains separate.

The three lists

New signals (long or short)

The stock's consensus reverses: bearish in the previous 14 days and bullish in the latest 14 (long), or the opposite (short). Only an early reversal counts: the new side has exactly 2 report days. Each side shows at most 10 names: the newest edition first, then the name that had already moved most in the call's direction against the market over the 10 sessions before the signal. After a position closes, another call on the same name in the same direction waits 20 sessions; a call in the opposite direction can open at once.

Continuation (long)

A bullish narrative is intact: its stage is building, crowded or revived, its contest level is low, its evidence is not sparse, and it was covered within the last 5 editions. It names the stock with a bull view in the last 21 days, and the stock's own consensus is bullish. At most 10 are open, ranked by how many such narratives back the name. A position renews for another 20 sessions while it still qualifies.

Watch (long, not ranked)

The same narrative test, but the stock lagged QQQ by 8 points or more over 20 sessions. It has no bearish consensus and no bearish reversal in the last 28 days. It is shown most-lagging first, at most 10 names, has its own track record and is never ranked with New signals.

A stock appears in one list only: New signals first, then Continuation, then Watch.

Entry, exit and the vs QQQ figure

• Simulated entry uses the first final close after the research date, independent of upload or processing time. Publication timestamps are unverified; matching source session dates are explicit assumptions. A missing entry close stays unpriced; it never shifts entry to a later session. In the recorded issuance ledger, entry is the first final close after both the report date and the day the rules ran. A report that arrives more than 5 weekdays late opens nothing. A close is final after 17:30 exchange time.
• Until that close exists, the vs QQQ cell shows —.
• After entry, vs QQQ is the stock's return minus QQQ's return, from the entry close to the latest close, in percentage points. It is signed in the direction of the call: for a short, a positive figure means the stock did worse than QQQ. A closed position is measured to its exit close.
• A position ends on an opposite reversal, after 20 sessions, or (Continuation and Watch) when no listed narrative is still bullish and not highly contested. Counter-evidence, a cooling narrative and quiet periods do not end it.
• Only stocks listed on a US exchange (NYSE, Nasdaq, NYSE American, NYSE Arca or Cboe BZX) and priced in US dollars are eligible. OTC and foreign listings are not. If a stock leaves this universe while its position is open, the position is withdrawn. Its recorded entry stays, and it does not count in the track record.
• Simulated exits follow dated research evidence and the signal's frozen hold rule. Review revisions preserve their own inputs and prices. The original issuance ledger remains unchanged. Entry and exit closes are recorded once and never rewritten.

Limits

• This reviews selected recorded signals, not all historical opportunities or a strategy known at the time. Simulated closures do not count toward live validation. The rules come from a replay of past research, not from live results. They stay in validation until New signals and Continuation each have 30 closed positions. Hit rates appear after 10.
• The consensus counts report days, not independent sources. A view repeated in a later report counts again.
• These are research views tracked against QQQ, not investment advice.

Rules reco_v2 · See today's simulation · Recorded issuance ledger