feat: add florida module for unified hedging and monitoring
This commit is contained in:
117
florida/summaries/6153292/6153292_analisis.md
Normal file
117
florida/summaries/6153292/6153292_analisis.md
Normal file
@ -0,0 +1,117 @@
|
||||
# CLP Strategy & Configuration Log
|
||||
|
||||
This document detailed information about clp position (6153292) + auto hedge on Hyperliquide.
|
||||
|
||||
---
|
||||
|
||||
## 1. Low Volatility / Weekend Optimization (Narrow Range)
|
||||
**Date:** 2025-12-29
|
||||
**Status:** Active
|
||||
**Objective:** Further optimalization of hedging on Hyperliquide.
|
||||
|
||||
### 🔍 Context
|
||||
* **Market Condition:** Monday, th 29th of Dec
|
||||
* **Capital:** $1,000 USDC
|
||||
* **Range:** +/- 1.5% (Narrow)
|
||||
|
||||
### ⚙️ Configuration of scripts
|
||||
(See original file for full config dump)
|
||||
|
||||
### test results
|
||||
1. clp position:
|
||||
{
|
||||
"type": "AUTOMATIC",
|
||||
"token_id": 6153292,
|
||||
"status": "CLOSED",
|
||||
"target_value": 993.31,
|
||||
"entry_price": 869.418,
|
||||
"amount0_initial": 500.0094,
|
||||
"amount1_initial": 0.5674,
|
||||
"liquidity": "2284728345715808667084",
|
||||
"range_upper": 882.4136,
|
||||
"range_lower": 856.6782,
|
||||
"token0_decimals": 18,
|
||||
"token1_decimals": 18,
|
||||
"timestamp_open": 1766982584,
|
||||
"target_value_end": 982.48,
|
||||
"timestamp_close": 1767000734
|
||||
}
|
||||
2. hedge transactions (from Hyperliquide)
|
||||
(See original file for table)
|
||||
|
||||
## results of tests are not satisfactional:
|
||||
1. the main problem is that the hedge doesn't cover lost value of clp pool (without earned fees) -> **-$2.23**
|
||||
|
||||
---
|
||||
|
||||
### 🚀 Analysis & Diagnosis (2025-12-29)
|
||||
|
||||
After reviewing the logs and transaction history, here is the breakdown of the PnL and the root cause of the slippage.
|
||||
|
||||
#### **1. PnL Breakdown (The "Missing" $2.23)**
|
||||
|
||||
* **LP Position Value Change:**
|
||||
* Initial Value: ~$993.31
|
||||
* Final Value: ~$982.50
|
||||
* **LP Loss:** **-$10.81** (This is the Impermanent Loss + Delta Loss from holding BNB as it dropped).
|
||||
* **Hedge Compensation:**
|
||||
* Hedge Gross Profit: **+$10.61** (from your table).
|
||||
* **Net Delta Efficiency:** The hedge covered the LP loss almost perfectly (Diff: -$0.20).
|
||||
* *Conclusion:* The **Delta Calculation is CORRECT**. The math effectively neutralized the market move.
|
||||
* **The Costs (The Real Leak):**
|
||||
* **Trading Fees:** You paid ~$0.79 in fees on Hyperliquid (mostly from the initial Taker entry and the final Taker exit).
|
||||
* **Funding Fees:** Since you were Shorting BNB for ~5 hours, and rates might have been negative (Shorts pay Longs) or just the cost of carry, the remaining discrepancy (~$1.20) is likely **Funding Costs** or slight slippage between LP exit price and Hedge exit price.
|
||||
|
||||
**Total Net:** -$0.20 (Delta Slippage) - $0.79 (Fees) - $1.24 (Funding/Execution Slippage) = **-$2.23**
|
||||
|
||||
#### **2. Execution Inefficiency (The "Panic Loop")**
|
||||
|
||||
The logs reveal a flaw in the execution strategy during trending moves:
|
||||
1. **Drift:** Price drops, Delta drift exceeds threshold (`0.05`).
|
||||
2. **Maker Attempt:** Bot places an `ALO` (Maker) order at the Bid.
|
||||
3. **Timeout:** Market moves down faster than the order fills. The order sits pending (`[WAIT]`).
|
||||
4. **Cancel:** Bot cancels the stale order.
|
||||
5. **Drift Worsens:** The drift continues to grow as price drops further.
|
||||
6. **Panic:** Eventually, `Drift > 5x Threshold` (Large Hedge).
|
||||
7. **Taker Smash:** Bot forces an `IOC` (Taker) trade, paying high fees (0.035%) and eating slippage.
|
||||
|
||||
**Evidence:**
|
||||
* Initial Entry: `[WARN] LARGE HEDGE` -> Taker (Fee: $0.21).
|
||||
* Final Exit: `[URGENT] ... Force Taker Exit` -> Taker (Fee: $0.45).
|
||||
* These two trades alone account for ~85% of your fee costs.
|
||||
|
||||
---
|
||||
|
||||
### 🛠️ Answers to Questions
|
||||
|
||||
**1. Is the calculation of hedge wrong?**
|
||||
**No.** The Delta calculation is accurate. The gross profit of the hedge (+$10.61) almost exactly matched the raw value loss of the LP (-$10.81). The math works.
|
||||
|
||||
**2. How we can proactively fix it?**
|
||||
We need to fix the **Execution Strategy** to avoid the "Wait -> Cancel -> Panic Taker" loop.
|
||||
* **Soft Taker Fallback:** If a Maker order times out (e.g., after 30s), retry as Taker immediately *before* the drift becomes huge.
|
||||
* **Asymmetric Compensation:** Increase the "Over-Hedge" factor. Since V3 LP accumulates "Long" exposure as price drops (Negative Gamma), we should short *more* aggressively early on.
|
||||
|
||||
**3. What we can do with configuration?**
|
||||
We need to tune for **Narrow Ranges** (High Gamma).
|
||||
|
||||
#### **Recommended Config Changes**
|
||||
|
||||
| Parameter | Current | Recommended | Reason |
|
||||
| :--- | :--- | :--- | :--- |
|
||||
| `MIN_HEDGE_THRESHOLD` | `0.05` | **0.02** | Hedge smaller deviations sooner to prevent runaway gamma. |
|
||||
| `LARGE_HEDGE_MULTIPLIER` | `5.0` | **2.5** | Trigger Taker/Urgent correction sooner, before the hole gets too deep. |
|
||||
| `MAKER_ORDER_TIMEOUT` | `600` | **60** | Don't let orders rot for 10 minutes. Cancel and retry faster. |
|
||||
| `SHADOW_ORDER_TIMEOUT` | `600` | **30** | Same as above. |
|
||||
| `BASE_REBALANCE_THRESHOLD_PCT` | `0.25` | **0.15** | Tighten the percentage-based trigger. |
|
||||
|
||||
#### **Code Change Recommendation (Unified Hedger)**
|
||||
We should update `unified_hedger.py` to:
|
||||
1. **Persist Realized PnL:** Fix the bug where `_status.json` isn't updated with hedge PnL, so your logs reflect reality.
|
||||
2. **Execution Fallback:** If an `ALO` order fails/cancels, decrement a counter. If it fails twice, force `IOC`.
|
||||
|
||||
---
|
||||
|
||||
### 📝 Action Plan
|
||||
1. Apply the **Config Changes** above to `clp_config.py`.
|
||||
2. (Optional) I can patch `unified_hedger.py` to fix the PnL logging and execution logic if you approve.
|
||||
Reference in New Issue
Block a user