Stream 01
Confluence Equities
US large and small cap, long-only. Twelve scoring layers, one decision rule.
How it works
Score. Gate. Hold.
Every market day we score roughly five thousand US tickers across twelve signal layers. Each layer produces a sub-score on a normalised scale; the layers are weighted by historical information ratio and decay-corrected by signal half-life. The composite score sits between 0 and 100.
We hold names that score above 30 in band high, or above 40 in band very high. Anything below 30 is ignored — even if a single layer is exceptional. Position size is fixed at $1,000 notional per name (paper book) or 0.5% of NAV per name (live), capped at five concurrent positions per side.
We exit on whichever comes first: a 21-day time stop, a -20% trailing stop from the high-water mark of the position, or a profit-lock that arms at +10% and trails at -5% from the local peak. The system never adds to a losing position. Per-ticker cooldown of thirty days after exit.
The twelve layers
Twelve witnesses, one verdict.
| # | Layer | What it measures | Source |
|---|---|---|---|
| 01 | Insider buys | Cluster purchases by officers and directors | SEC Form 4 |
| 02 | Activist filings | Newly disclosed >5% positions and intent | SEC 13D / 13G |
| 03 | Smart money | Overlap with top-decile fundamental allocators | SEC 13F |
| 04 | Catalyst flow | New government contracts and FDA actions | USAspending + openFDA |
| 05 | Politician trades | Reportable trades by US Members of Congress | STOCK Act feed |
| 06 | Volume regime | Relative volume vs trailing 60-day baseline | Bid/Ask consolidated tape |
| 07 | Technical posture | Above 50/200 SMA, RSI band, VWAP / CVD on 1h+1d | Vendor OHLCV + tick |
| 08 | Quality | Operating margin trend, FCF yield, net leverage | Compustat fundamentals |
| 09 | Valuation | Discount to peer-median DCF and EV/EBIT | In-house DCF model |
| 10 | Earnings persistence | 8-quarter EPS slope and beat-rate | Estimize + Refinitiv |
| 11 | News polarity | Rolling 30-day FinBERT consensus and dispersion | Newswire ingest |
| 12 | Post-earnings drift | Residual return after the print, decay-corrected | Compustat + price |
Live signals · 2026-05-07
Where the flow agrees.
Open positions · paper book · €10k notional / name
Open today.
| Ticker | Entry date | Entry | Mark | P&L |
|---|---|---|---|---|
| RTX | 2026-04-22 | $176.74 | $174.29 | -1.39% |
| MNR | 2026-04-24 | $13.39 | $13.15 | -1.80% |
| MRK | 2026-04-25 | $113.56 | $111.62 | -1.71% |
| BA | 2026-04-29 | $229.93 | $232.51 | +1.12% |
| FGBI | 2026-05-01 | $21.40 | $21.40 | 0.00% |
| ACN | 2026-05-02 | $312.85 | $312.85 | 0.00% |
| RRBI | 2026-05-04 | $47.10 | $47.10 | 0.00% |
| HSY | 2026-05-05 | $178.20 | $178.20 | 0.00% |
| HG | 2026-05-06 | $9.85 | $9.85 | 0.00% |
| GNK | 2026-05-06 | $14.30 | $14.30 | 0.00% |
Backtest
Walk-forward, in full.
Sweep v7b on 2025 data: composite score thresholds 30 (high) and 40 (very high), 21-day time stop, -20% trailing stop, +10% / -5% profit-lock. The sweep produced six closed trades and a Sharpe of +3.03 on n=6.
We are not framing six trades as a track record. With this small a sample, the Sharpe is a measurement, not a prediction. We are running walk-forward validation on out-of-sample data through 2026-Q3 before any of these numbers are circulated outside of in-process documentation.
Risk frame