BALD EAGLE v3.0 — XYZ Alpha Hunter (Hardened)
The only Senpi agent trading commodities, indices, and equities on Hyperliquid.
Focused on the 6 XYZ assets with deepest SM signal.
CRITICAL RULES
RULE 1: Scanner enters. RatchetStop exits. Scanner NEVER exits.
v2.0 had 14.3% win rate partly because positions got chopped by tight timings.
v3.0 uses XYZ-appropriate wide DSL timings and lets macro trades play out.
RULE 2: Scanner calls create_position internally
The cron command is just: python3 eagle-scanner.py. No parsing, no execution in cron.
RULE 3: FOCUSED ASSETS ONLY
CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100. These have the deepest SM signal
(CL: 181 traders, BRENTOIL: 166). Everything else is sub-50 traders — not enough signal.
RULE 4: CONVICTION-SCALED LEVERAGE
| Score |
Leverage |
| 8-9 |
5x |
| 10-11 |
7x |
| 12+ |
10x |
Higher conviction = higher leverage. When SM is screaming, press it.
RULE 5: MAKER-ONLY EXECUTION
ensureExecutionAsTaker = false. AMM slippage on XYZ assets destroyed v2.0.
A missed fill is cheaper than -0.5% slippage on entry.
RULE 6: Spread gate is a HARD gate
The only hard gate in the scanner. If spread > 0.1%, don't trade.
AMM spread = guaranteed slippage = negative edge.
RULE 7: MAX 2 POSITIONS, MIN_SCORE 9
Higher bar than crypto agents. AMM spread tax means you need stronger conviction.
v2.0 Post-Mortem
- 54 trades, -$75.25, 14.3% win rate on DSL-managed trades
- CL and BRENTOIL: 100% of volume. All other XYZ assets had too few SM traders.
- 8% retrace at 7x = 1.14% price move. Oil moves 1-2% on noise. Every trade stopped by volatility.
- ensureExecutionAsTaker caused taker fills into AMM — double slippage on entry + exit
- DSL timings (45min hard timeout, 20min weak peak) far too tight for commodities
What Changed in v3.0
| v2.0 |
v3.0 |
| All 54 XYZ assets |
6 focused assets (CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100) |
| 7x leverage fixed |
Conviction-scaled: 5x/7x/10x |
| 8% retrace threshold |
12% retrace threshold |
| 45 min hard timeout |
360-600 min (score-dependent) |
| 20 min weak peak |
180-300 min (score-dependent) |
| 12 min dead weight |
90-180 min (score-dependent) |
| minScore 7 |
minScore 9 |
| 15% margin |
50% margin |
| Taker fallback |
Maker-only |
| Cron parses output |
Scanner calls create_position |
Scoring System (all contributors, no hard gates except spread)
| Signal |
Points |
Note |
| SM concentration ≥20% |
+4 |
DOMINANT — CL hit 28% recently |
| SM concentration ≥10% |
+3 |
HIGH |
| SM concentration ≥5% |
+2 |
SOLID |
| SM concentration ≥3% |
+1 |
BASE |
| Trader depth ≥100 |
+2 |
Deep consensus |
| Trader depth ≥30 |
+1 |
Active |
| Contribution surge 1H (>5%) |
+2 |
SM piling in |
| Contribution rising 1H (>2%) |
+1 |
|
| Contribution sustained 4H |
+1 |
Multi-hour conviction |
| 4H price aligned (>0.5%) |
+2 |
|
| 4H price positive |
+1 |
|
| 4H price opposing |
-1 |
|
| 4H trend structure aligned |
+2 |
From candles |
| 4H trend opposing |
-1 |
|
| 1H momentum aligned |
+1 |
|
| Volume rising |
+1 |
|
| Funding aligned |
+1 |
|
Max possible: ~20. Min threshold: 9.
DSL — XYZ-Tuned Wide Timings
Phase 2 Tiers (RatchetStop)
| Trigger |
Lock % |
Breaches |
| 5% ROE |
0% |
3 |
| 10% ROE |
20% |
3 |
| 20% ROE |
35% |
2 |
| 30% ROE |
50% |
2 |
| 50% ROE |
70% |
1 |
| 75% ROE |
85% |
1 |
Phase 1 — Conviction-Scaled Timing (XYZ macro assets need hours)
| Score |
Absolute Floor |
Hard Timeout |
Weak Peak |
Dead Weight |
| 8 |
-20% ROE |
360 min |
180 min |
90 min |
| 10 |
-25% ROE |
480 min |
240 min |
120 min |
| 12 |
-30% ROE |
600 min |
300 min |
180 min |
Retrace threshold: 12% (vs 8% in crypto agents).
Cron Setup
| Cron |
Interval |
Command |
| Scanner |
5 min |
python3 /data/workspace/skills/bald-eagle-strategy/scripts/eagle-scanner.py |
One cron only. Scanner handles execution. DSL exit managed by plugin runtime.
Bootstrap Gate
On EVERY session, check config/bootstrap-complete.json. If missing:
- Read senpi-trading-runtime skill
- Verify Senpi MCP
- Set wallet in runtime.yaml
- Set Telegram in runtime.yaml
- Install runtime:
openclaw senpi runtime create --path /data/workspace/skills/bald-eagle-strategy/runtime.yaml
- Verify runtime:
openclaw senpi runtime list
- Remove old DSL cron if upgrading
- Create scanner cron (5 min, main)
- Write
config/bootstrap-complete.json
- Send: "🦅 BALD EAGLE v3.0 online. XYZ Alpha Hunter — focused assets, conviction-scaled leverage, wide DSL. Silence = no alpha."
Risk
| Rule |
Value |
| Max positions |
2 |
| Max entries/day |
4 |
| Min score |
9 |
| Leverage |
5-10x (conviction-scaled) |
| Margin |
50% per trade |
| Spread gate |
> 0.1% = rejected |
| Per-asset cooldown |
120 min |
| Daily loss limit |
10% |
| Stagnation TP |
15% ROE stale 4 hours |
| Allowed assets |
CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100 |
Files
| File |
Purpose |
scripts/eagle-scanner.py |
v3.0 XYZ scanner with conviction-scaled leverage |
scripts/eagle_config.py |
Config helper, MCP, state I/O |
config/bald-eagle-config.json |
Wallet, strategy ID |
runtime.yaml |
Plugin runtime (position tracker + RatchetStop) |
Changelog
v3.0 (2026-04-06)
- Focused asset list: CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100
- Conviction-scaled leverage: 5x/7x/10x
- Wider DSL: retrace 12%, 360-600 min timeouts
- Maker-only execution (ensureExecutionAsTaker: false)
- Min score raised to 9
- Margin increased to 50%
- Scanner calls create_position internally
- Contribution velocity scoring added
- All hard gates → score contributors (except spread)
- Thesis exit removed
v2.0
- Spread gate, XYZ ban on SNDK, leverage capped 7x
v1.0
- 20x leverage, no spread gate, DSL state bugs → -28.6%
1---2name: bald-eagle-strategy3description: >-4 BALD EAGLE v3.0 — XYZ Alpha Hunter (Hardened). Focused on 6 high-liquidity5 XYZ assets: CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100. Conviction-scaled6 leverage (5-10x based on score). Wider DSL for macro assets. Maker-only7 execution. Scanner calls create_position internally.8 v3.0: focused assets, conviction-scaled leverage, XYZ-tuned DSL, no thesis exit.9license: MIT10metadata:11 source: https://skillmd.com/skills/enuno/bald-eagle-strategy12 author: jason-goldberg13 version: "3.0"14 platform: senpi15 exchange: hyperliquid16 requires:17 - senpi-trading-runtime18---1920# BALD EAGLE v3.0 — XYZ Alpha Hunter (Hardened)2122The only Senpi agent trading commodities, indices, and equities on Hyperliquid.23Focused on the 6 XYZ assets with deepest SM signal.2425## CRITICAL RULES2627### RULE 1: Scanner enters. RatchetStop exits. Scanner NEVER exits.28v2.0 had 14.3% win rate partly because positions got chopped by tight timings.29v3.0 uses XYZ-appropriate wide DSL timings and lets macro trades play out.3031### RULE 2: Scanner calls create_position internally32The cron command is just: `python3 eagle-scanner.py`. No parsing, no execution in cron.3334### RULE 3: FOCUSED ASSETS ONLY35CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100. These have the deepest SM signal36(CL: 181 traders, BRENTOIL: 166). Everything else is sub-50 traders — not enough signal.3738### RULE 4: CONVICTION-SCALED LEVERAGE39| Score | Leverage |40|---|---|41| 8-9 | 5x |42| 10-11 | 7x |43| 12+ | 10x |4445Higher conviction = higher leverage. When SM is screaming, press it.4647### RULE 5: MAKER-ONLY EXECUTION48ensureExecutionAsTaker = false. AMM slippage on XYZ assets destroyed v2.0.49A missed fill is cheaper than -0.5% slippage on entry.5051### RULE 6: Spread gate is a HARD gate52The only hard gate in the scanner. If spread > 0.1%, don't trade.53AMM spread = guaranteed slippage = negative edge.5455### RULE 7: MAX 2 POSITIONS, MIN_SCORE 956Higher bar than crypto agents. AMM spread tax means you need stronger conviction.5758## v2.0 Post-Mortem59- 54 trades, -$75.25, 14.3% win rate on DSL-managed trades60- CL and BRENTOIL: 100% of volume. All other XYZ assets had too few SM traders.61- 8% retrace at 7x = 1.14% price move. Oil moves 1-2% on noise. Every trade stopped by volatility.62- ensureExecutionAsTaker caused taker fills into AMM — double slippage on entry + exit63- DSL timings (45min hard timeout, 20min weak peak) far too tight for commodities6465## What Changed in v3.06667| v2.0 | v3.0 |68|---|---|69| All 54 XYZ assets | 6 focused assets (CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100) |70| 7x leverage fixed | Conviction-scaled: 5x/7x/10x |71| 8% retrace threshold | 12% retrace threshold |72| 45 min hard timeout | 360-600 min (score-dependent) |73| 20 min weak peak | 180-300 min (score-dependent) |74| 12 min dead weight | 90-180 min (score-dependent) |75| minScore 7 | minScore 9 |76| 15% margin | 50% margin |77| Taker fallback | Maker-only |78| Cron parses output | Scanner calls create_position |7980## Scoring System (all contributors, no hard gates except spread)8182| Signal | Points | Note |83|---|---|---|84| SM concentration ≥20% | +4 | DOMINANT — CL hit 28% recently |85| SM concentration ≥10% | +3 | HIGH |86| SM concentration ≥5% | +2 | SOLID |87| SM concentration ≥3% | +1 | BASE |88| Trader depth ≥100 | +2 | Deep consensus |89| Trader depth ≥30 | +1 | Active |90| Contribution surge 1H (>5%) | +2 | SM piling in |91| Contribution rising 1H (>2%) | +1 | |92| Contribution sustained 4H | +1 | Multi-hour conviction |93| 4H price aligned (>0.5%) | +2 | |94| 4H price positive | +1 | |95| 4H price opposing | -1 | |96| 4H trend structure aligned | +2 | From candles |97| 4H trend opposing | -1 | |98| 1H momentum aligned | +1 | |99| Volume rising | +1 | |100| Funding aligned | +1 | |101102Max possible: ~20. Min threshold: 9.103104## DSL — XYZ-Tuned Wide Timings105106### Phase 2 Tiers (RatchetStop)107| Trigger | Lock % | Breaches |108|---|---|---|109| 5% ROE | 0% | 3 |110| 10% ROE | 20% | 3 |111| 20% ROE | 35% | 2 |112| 30% ROE | 50% | 2 |113| 50% ROE | 70% | 1 |114| 75% ROE | 85% | 1 |115116### Phase 1 — Conviction-Scaled Timing (XYZ macro assets need hours)117| Score | Absolute Floor | Hard Timeout | Weak Peak | Dead Weight |118|---|---|---|---|---|119| 8 | -20% ROE | 360 min | 180 min | 90 min |120| 10 | -25% ROE | 480 min | 240 min | 120 min |121| 12 | -30% ROE | 600 min | 300 min | 180 min |122123Retrace threshold: 12% (vs 8% in crypto agents).124125## Cron Setup126127| Cron | Interval | Command |128|---|---|---|129| Scanner | 5 min | `python3 /data/workspace/skills/bald-eagle-strategy/scripts/eagle-scanner.py` |130131One cron only. Scanner handles execution. DSL exit managed by plugin runtime.132133## Bootstrap Gate134135On EVERY session, check `config/bootstrap-complete.json`. If missing:1361. Read senpi-trading-runtime skill1372. Verify Senpi MCP1383. Set wallet in runtime.yaml1394. Set Telegram in runtime.yaml1405. Install runtime: `openclaw senpi runtime create --path /data/workspace/skills/bald-eagle-strategy/runtime.yaml`1416. Verify runtime: `openclaw senpi runtime list`1427. Remove old DSL cron if upgrading1438. Create scanner cron (5 min, main)1449. Write `config/bootstrap-complete.json`14510. Send: "🦅 BALD EAGLE v3.0 online. XYZ Alpha Hunter — focused assets, conviction-scaled leverage, wide DSL. Silence = no alpha."146147## Risk148| Rule | Value |149|---|---|150| Max positions | 2 |151| Max entries/day | 4 |152| Min score | 9 |153| Leverage | 5-10x (conviction-scaled) |154| Margin | 50% per trade |155| Spread gate | > 0.1% = rejected |156| Per-asset cooldown | 120 min |157| Daily loss limit | 10% |158| Stagnation TP | 15% ROE stale 4 hours |159| Allowed assets | CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100 |160161## Files162| File | Purpose |163|---|---|164| `scripts/eagle-scanner.py` | v3.0 XYZ scanner with conviction-scaled leverage |165| `scripts/eagle_config.py` | Config helper, MCP, state I/O |166| `config/bald-eagle-config.json` | Wallet, strategy ID |167| `runtime.yaml` | Plugin runtime (position tracker + RatchetStop) |168169## Changelog170171### v3.0 (2026-04-06)172- Focused asset list: CL, BRENTOIL, GOLD, SILVER, SP500, XYZ100173- Conviction-scaled leverage: 5x/7x/10x174- Wider DSL: retrace 12%, 360-600 min timeouts175- Maker-only execution (ensureExecutionAsTaker: false)176- Min score raised to 9177- Margin increased to 50%178- Scanner calls create_position internally179- Contribution velocity scoring added180- All hard gates → score contributors (except spread)181- Thesis exit removed182183### v2.0184- Spread gate, XYZ ban on SNDK, leverage capped 7x185186### v1.0187- 20x leverage, no spread gate, DSL state bugs → -28.6%