Public screens · honest historical record

Watchlists with receipts.

Every public watchlist is a named rule replayed across history. You see exactly which tickers it would have fired on, when, and how each performed against SPY - including the ones that got delisted.

…Public watchlists

…Historical fires

…Median vs SPY · 20d

…

filter

Voting and following need your sk-live-* key. Grab the demo key from access, or use your own. Stored locally in your browser only.

Top watchlists · ranked by alpha

Shrunk median alpha so a single home-run doesn't outweigh a steady performer. Each watchlist needs at least 10 evaluable fires to qualify. Click a row to jump to its card.

20d horizon · close→close · vs SPY / BTC

#	Watchlist	N fires	Median α	Hit rate	Cum. excess · 12m
loading…

Share via MCP

Publish a watchlist.

Build the rule with your agent via add_watch, then flip it public with make_watchlist_public. ClawTerminal replays it across history and keeps it scoring forever.

How to connect →

# in your Claude / agent
add_watch(label="My screen",
  watch_type="insider", params={...})
make_watchlist_public(
  watch_id=42,
  name="Insider clusters · pharma",
  description="P codes, ≥$250k, …",
)

Top authors · 20d alpha

Median vs SPY

House rules

One rule, replayed across every prior tick that matches.
Return = close at fire → close at fire + 20 trading days. Split-adjusted.
Excess: equities vs SPY, crypto vs BTC, macro raw.
Unevaluable matches (delisted, no price) are shown, never hidden.
Shrunk median - N=3 worth less than N=30 in rankings.
No future leak - every feature is snapped at fire date.
Self-vote rejected at the API.