rhino-strategy（犀牛策略）

RHINO — Momentum Pyramider

Enter small. Add to winners. Build into conviction.

The only skill in the Senpi zoo that adds to winning positions. Every other skill enters full size once and lets DSL manage the exit. RHINO enters at 30% of max position, then pyramids in as the trade proves correct — adding at +10% ROE and +20% ROE if the thesis still holds.

Why Pyramiding

Standard approach: risk 100% of position size at entry, when the thesis is unproven.

RHINO approach: risk 30% at entry, 70% only after the trade confirms.

If Stage 1 fails → you lose on 30% of max size (not 100%). If it reaches Stage 3 → your full position is built across the move with a better average entry than all-in would give.

The math: on a $300 max position, RHINO risks $90 at entry. If it fails, you lose $90 × stop = ~$18. An all-in entry would lose $300 × stop = ~$60. Same signal, 3x less risk on the initial bet.

MANDATORY: DSL High Water Mode

RHINO 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

When creating DSL state files, you MUST include:

{
  "lockMode": "pct_of_high_water",
  "phase2TriggerRoe": 8,
  "tiers": [
    {"triggerPct": 8,  "lockHwPct": 25, "consecutiveBreachesRequired": 3},
    {"triggerPct": 15, "lockHwPct": 45, "consecutiveBreachesRequired": 2},
    {"triggerPct": 25, "lockHwPct": 60, "consecutiveBreachesRequired": 2},
    {"triggerPct": 40, "lockHwPct": 75, "consecutiveBreachesRequired": 1},
    {"triggerPct": 60, "lockHwPct": 85, "consecutiveBreachesRequired": 1}
  ]
}

FALLBACK: Use tiersLegacyFallback from config until engine supports pct_of_high_water.

DSL applies to the full position, not per-add. Each pyramid add increases position size but the DSL state tracks one high-water mark for the entire position. After adding, the DSL floor recalculates based on the new total margin.

The Three Stages

Stage 1 — SCOUT (30% of max position)

When: Score 10+ convergence on a top-10 asset.

Size: 30% of max margin. At $1,000 account with 30% max margin = $300 max position, scout enters with $90.

Purpose: Skin in the game. Prove the thesis with minimal risk.

Stage 2 — CONFIRM (add 40% at +10% ROE)

When: Position is at +10% ROE AND thesis re-validated:

• 4h trend still intact (same direction)
• SM still aligned (not flipped)
• Volume still present (ratio ≥ 0.5x average)

Size: 40% of max margin. Adds $120 to the $90 scout. Position is now $210 (70% of max).

Purpose: The trade is working. Double down while the signal is confirmed.

Stage 3 — CONVICTION (add final 30% at +20% ROE)

When: Position is at +20% ROE AND thesis re-validated (same checks as Stage 2).

Size: Final 30% of max margin. Adds $90. Position is now $300 (100% of max).

Purpose: The trade is running. Full conviction deployed. DSL High Water trails the full $300 from here.

What If a Stage Doesn't Trigger?

Stage 2 and 3 are optional. If the trade reaches +10% ROE but the 4h trend has broken, the add doesn't happen — the scout position rides with DSL protection. If it never reaches +10%, the scout either hits its stop or trails out on DSL. No forced adds.

Scan Priority

Every 3-minute scan, RHINO checks in this order:

1. Add to existing winners — do any positions qualify for Stage 2 or 3? If yes, add and return.
2. Entry cap check — any slots available?
3. Scout new positions — scan top 10 for new Stage 1 entries.

Adds take priority over new entries. A winning position that needs Stage 2 is more valuable than a new scout.

Risk Management

Cron Architecture

Notification Policy

ONLY alert:

• Stage 1 SCOUT opened (asset, direction, margin, score)
• Stage 2 CONFIRM added (asset, +ROE at add, new total margin)
• Stage 3 CONVICTION added (asset, +ROE at add, full position)
• Position CLOSED (DSL or structural, total P&L across all stages)
• Risk triggered, critical error

NEVER alert: Scanner found nothing, DSL routine, pyramid evaluation found no adds, any reasoning. All crons isolated. NO_REPLY for idle cycles.

Bootstrap Gate

Check config/bootstrap-complete.json every session. If missing:

1. Verify Senpi MCP
2. Create scanner cron (3 min, isolated) and DSL cron (3 min, isolated)
3. Write config/bootstrap-complete.json
4. Send: "🦏 RHINO is online. Scouting top 10 by OI + volume. Enter small, add to winners. Silence = no conviction."

Expected Behavior

The edge: losers are small (30% position × stop). Winners are big (100% position × High Water trail). Asymmetric by design.

Files

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.