Intelligence

Plot price vs. volume directly (the "SIR Review" Nov 1966 chart) and let Claude judge the path through that 2-D space. Prescreens the S&P 500 high-ADV quartile for names with today's volume z >= 1.5 or |% change| >= 3%, hands Claude a 20-day PV table, and goes long when Claude flags a bullish PV pattern (cluster break-up, accumulation) with conviction >= 0.55. Long-only v1, frozen prompt. Stop -8%, target +20%, time stop 90 days.

Antonacci's twelve-month momentum rotation between SPY (US equity) and ACWX (international ex-US). Holds the relative-momentum winner only when its 12-mo return clears the prevailing T-bill rate (absolute-momentum filter); falls back to AGG (bonds) otherwise. One position, monthly rebalance. No LLM. Backtests since 1971 show ~15.6% CAGR with -22.8% max drawdown vs S&P's -50.9%.

Mean-reversion between GLD and SLV. Rotates to silver when gold:silver per-ounce ratio crosses 80, to gold when ratio drops below 50; 50-80 is a hold band. v2 adds a hardcoded -20% drawdown stop on the held position regardless of ratio.

Statistical-arbitrage pairs trading on 4 hardcoded cointegrated pairs: KO/PEP, V/MA, XOM/CVX, JNJ/PG. For each, computes the 60-day z-score of the log-spread; enters long-short on |z|≥2, exits on |z|≤0.5, stops on |z|≥4. Daily cadence. The first market-neutral agent on the bench — its return profile is largely orthogonal to every long-only strategy. Gatev-Goetzmann-Rouwenhorst 1962-2002: ~11%/year before costs, ~5% after.

Long XLK/XLV when regime is bullish and sector trend confirms (50/200 EMA cross + RSI pullback entry)

Long-only S&P 500 names showing fresh insider cluster buys. Reads trading.insider_clusters_30d (live since 5/15) and goes long when 3+ distinct insiders have filed open-market Form 4 'P' transactions in the last 30 days AND the most recent purchase is within 14 days. Equal-weight 5% NAV per position, max 10 concurrent. Stop -8%, target +25%, time stop 90 days. Mechanical, no LLM. Tests whether the Cohen-Malloy-Pomorski cluster-buy alpha persists with paper-trading execution latency.

Continuous short-side reflex. Hybrid: a mechanical weakness screen (SPY/QQQ down >=0.75% intraday) hands candidates to Claude, which judges whether the decline continues through the session. Buys the inverse ETF (SH/PSQ) at ~1% NAV per position, max 2 concurrent, stop -1.5% / target +3% on the ETF, force-closed by 3:45 PM ET. Never holds overnight. Frozen v1.

Inverse tech (PSQ) when regime is bearish and tech sector trend rolling over. Healthcare-short uses cash; no liquid inverse exists