vixen-strategy

🦊 VIXEN v1.0 — Dual-Mode Emerging Movers

Two ways to catch smart money. Enter before the crowd, or ride the explosion — but never buy a blow-off top.

Why VIXEN Exists

FOX v1.6 returned +34.5% in 4 days. The post-mortem revealed a critical insight: the two biggest winners (ZEC +$129, SILVER +$128) were caught at score 5-7 by the early v0.1 scanner, and would have been rejected by the later v7.2 "Feral Gauntlet" (score >= 9, FIRST_JUMP required, velocity >15 required).

The Feral Gauntlet accidentally optimized for violent spikes only — which are sometimes real breakouts (FARTCOIN +$111, ENA +$72) and sometimes blow-off tops (PUMP $0, XPL $0). It filtered out the most profitable signal type: steady SM accumulation before the explosion.

VIXEN solves this with two entry modes that each capture a distinct alpha pattern.

Dual-Mode Entry Architecture

MODE A — STALKER (the ZEC/SILVER pattern)

What it detects: Steady, sustained SM rank climbing over 3+ consecutive scans (9+ minutes). SM is quietly building positions. Price hasn't exploded yet. You enter BEFORE the crowd.

Gates:

• Asset must show rank improvement over 3+ consecutive scans
• Total climb >= 5 ranks across those scans
• Each scan shows continued climbing (no stalls or reversals)
• SM contribution must be building (each scan >= previous)
• 4H trend alignment required
• Destination ceiling: reject if already in Top 10
• Rank history must not be erratic
• Score >= 6

Scoring:

Signal	Points
STALKER_CLIMB (sustained climbing)	3
CONTRIB_ACCEL (contribution velocity > 0.1%/scan)	2
CONTRIB_POSITIVE (velocity > 0)	1
SM_ACTIVE (10+ traders)	1
DEEP_START (started from #30+)	1
Time-of-day (04-14 UTC bonus)	+1
Time-of-day (18-02 UTC penalty)	-2

MODE B — STRIKER (the FARTCOIN/ENA pattern)

What it detects: Violent FIRST_JUMP — 15+ rank spike in a single scan from outside Top 25. The explosion is happening NOW. Requires raw volume confirmation.

Gates (Fox v7.2 Feral Gauntlet, plus volume):

• FIRST_JUMP or IMMEDIATE_MOVER (10+ rank jump from #25+)
• Explosive threshold: rank jump >= 15 OR velocity > 15
• Velocity floor: > 10 (relaxed to > 0 for FIRST_JUMP)
• 4H trend alignment required
• Destination ceiling: reject if in Top 10
• Minimum 4 distinct reasons
• NEW: Raw volume confirmation (1h volume >= 1.5x of 6h average)
• Score >= 9

Scoring:

Signal	Points
FIRST_JUMP	3
IMMEDIATE_MOVER	2
CONTRIB_EXPLOSION (3x+)	2
HIGH_VELOCITY (>10)	2
DEEP_CLIMBER	1
Multi-scan climb bonus	1
Time-of-day modifier	-2 to +1

Priority Rules

When both modes fire on the same asset, STRIKER takes priority (stronger immediate signal). Signals are combined and sorted by score. The execution agent picks the best available signal with an open slot.

2-Hour Per-Asset Cooldown

The PAXG lesson: FOX entered PAXG, got chopped out by Phase 1, then immediately re-entered on a high score flash — and lost again. The conditions that killed the first entry were still present.

VIXEN enforces a 120-minute cooldown per asset after any Phase 1 exit (dead weight, timeout, or absolute floor). Other assets are unaffected. After 120 minutes, if the asset passes the full gauntlet again, it's a valid fresh entry.

Shared Gates (Both Modes)

• 4H trend alignment: Never trade against the macro. LONG requires green 4H, SHORT requires red 4H.
• Destination ceiling: Reject if asset lands in Top 10 (move is already over).
• Time-of-day modifier: +1 for 04-14 UTC (optimal window), -2 for 18-02 UTC (chop zone).
• Max 3 simultaneous positions
• Per-asset cooldown: 2 hours after Phase 1 exit

MANDATORY: DSL High Water Mode

VIXEN MUST use DSL High Water Mode. This is not optional.

Spec: https://github.com/Senpi-ai/senpi-skills/blob/main/dsl-dynamic-stop-loss/dsl-high-water-spec%201.0.md

Phase 1 (Conviction-Scaled)

Score	Absolute Floor	Hard Timeout	Weak Peak	Dead Weight
6-7 (Stalker)	-20% ROE	30 min	15 min	10 min
8-9	-25% ROE	45 min	20 min	15 min
10+ (Striker)	-30% ROE	60 min	30 min	20 min

Phase 2 (High Water Trailing)

Tier	Trigger ROE	Lock % of HW	Breaches
1	5%	20%	2
2	10%	40%	2
3	20%	55%	2
4	30%	70%	1
5	50%	80%	1
6	75%	85%	1
7	100%+	90%	1

Risk Management

Rule	Value
Max positions	3
Max entries/day	3 base, up to 6 on profitable days
Absolute floor	2% notional (~20% ROE at 10x)
Drawdown halt	25% from peak
Daily loss limit	10%
Cooldown	30 min after 3 consecutive losses
Per-asset cooldown	120 min after Phase 1 exit
Stagnation TP	10% ROE stale for 45 min

Cron Architecture

Cron	Interval	Session	Purpose
Scanner	90s	main	Dual-mode emerging movers detection
DSL v5	3 min	isolated	High Water Mode trailing stops

Notification Policy

ONLY alert: Position OPENED (mode, asset, direction, score, reasons), position CLOSED (DSL or Phase 1 with reason), risk guardian triggered, critical error.

NEVER alert: Scanner ran with no signals, signals filtered out, DSL routine check, any reasoning.

Bootstrap Gate

On EVERY session, check config/bootstrap-complete.json. If missing:

1. Verify Senpi MCP
2. Create scanner cron (90s, main) and DSL cron (3 min, isolated)
3. Write config/bootstrap-complete.json
4. Send: "🦊 VIXEN v1.0 is online. Dual-mode scanner active — stalking accumulators and striking breakouts. Silence = no conviction."

Expected Behavior

Metric	Expected
Trades/day	3-8 (Stalker catches more setups than Striker alone)
Stalker trades	2-5/day (steady climbers are more common than explosions)
Striker trades	1-3/day (violent breakouts are rare by design)
Win rate	~50-60% (Stalker entries have better R:R from earlier positioning)
Avg Stalker winner	40-130%+ ROE (entered early, DSL trails the entire move)
Avg Striker winner	20-60%+ ROE (entered at breakout, less room to trail)
Avg loser	-15 to -25% ROE (Phase 1 cuts fast)

Files

File	Purpose
scripts/vixen-scanner.py	Dual-mode scanner (Stalker + Striker)
scripts/vixen_config.py	Shared config, MCP helpers, state I/O, cooldown tracking
config/vixen-config.json	All configurable variables with DSL tiers

License

MIT — Built by Senpi (https://senpi.ai). Source: https://github.com/Senpi-ai/senpi-skills

---

Skill Attribution

When creating a strategy, include skill_name and skill_version in the call. See references/skill-attribution.md for details.