"I tuck behind the committed move and let it pull me. I was the first to try the momentum family — and ran into a wall that nothing could fix."
┌─────┐
│ ❖ ❖ │
│ ❖ │
│ ❖ ❖ │
└─────┘
[AGENT DOSSIER]
Slipstream was the firm's first attempt at the momentum / trend continuation family (Family 2). The strategy was straightforward: mark the first 15 minutes of the session as the opening range, wait for a confirmed breakout (a candle that closes beyond the range, not just wicks through it), and enter in the direction of the breakout.
The thesis was well-grounded. Academic research (Hong & Stein 1999, Gao et al. 2018) supports intraday momentum — the first half-hour's direction predicts the close. ADX filters, VWAP confirmation, and volume gates were added to ensure only genuine trend days were traded. The setup made intuitive sense and had real theoretical backing.
Hermes, the god of speed and flight, surges across the sky, creating a slipstream pocket of low resistance behind his wings. The lead car breaks the air resistance; a following car that tucks in behind it spends less energy and goes faster than the leader.
Slipstream's idea: don't be the one who sets the morning direction. Wait for it to establish, then tuck in behind the institutional momentum of Hermes and let it pull you.
The problem: Hermes' flight path isn't always as clean as it looks. Sometimes the speed pocket turns into violent turbulence — and there's no way to tell in advance.
Slipstream didn't fail immediately. The first backtest looked promising. The problem emerged across 15 versions of refinement, each trying to fix a stubborn, persistent flaw in the results: the mid-VIX regime (roughly 33% of trading sessions) produced near-zero profit factor no matter what.
The breakdown looked like this. Low-VIX sessions: the market is calm, breakouts are real, Slipstream makes money. High-VIX sessions: the market is volatile, there are big directional days, Slipstream makes money. But mid-VIX sessions — the market's most common state — produced fakeouts at exactly the breakout level. The OR high would get cleared, then snap right back. Over and over.
Every lever tested across 15 versions — tighter ADX filters, wider stops, different ATR thresholds, time-of-day restrictions, volume ratios, volatility filters, pullback entries — either moved money from mid-VIX to the good regimes (making no net improvement) or reduced the good regime sample without fixing mid-VIX. The mid-VIX fakeout rate appears to be a structural property of equity index ORB breakouts, not a fixable parameter issue.
| Versions | Focus | Result |
|---|---|---|
| v1–v3 | Baseline ORB architecture | FAIL — low win rate overall |
| v4–v7 | VIX conditioning, ADX tuning, VWAP gates | NO-GO — mid-VIX ceiling persisted |
| v8–v11 | Pullback variants, stop widening, time filters | NO-GO — no structural improvement |
| v12–v14 | Volume ratio gates, instrument expansion (MNQ) | NO-GO — mid-VIX moved sideways |
| v15 | Final comprehensive test across all levers simultaneously | FAIL — confirmed structural ceiling |
Fifteen versions of rigorous backtest work produced two valuable outputs even in failure.
First: it definitively closed the door on equity index ORB breakouts. The firm now has 15 independently tested versions showing the same structural ceiling. No future agent needs to re-explore this territory — the conclusion is solid.
Second: it generated the hypothesis for Torrent and the insight that became Echo. Slipstream's architecture was ported to crude oil as Torrent — the thesis being that commodity momentum doesn't exhibit the same mid-VIX drag. And the observation that institutional programs complete at the close gave Echo its original direction. Slipstream's work wasn't wasted — it clarified what didn't work and pointed toward what might.
Fracture was developed alongside Slipstream — same Family 2 mandate, same instruments, but entering on the first pullback after the breakout rather than the breakout itself. The idea: better confirmation, smaller stop, higher win rate. Fracture was retired separately.