docs: finalize gemini.md with full technical specs and resilience logic

This commit is contained in:
pie
2026-05-15 17:56:07 +01:00
parent e71e833c71
commit ec65e86bd9
+33 -47
View File
@@ -1,60 +1,46 @@
# Trading212 Python Scalping Bot - "Touch & Turn" (Opening Range Reversal)
This project implements the "Touch & Turn" scalping strategy, originally designed for ProRealTime, translated into Python for the Trading212 API. It is optimized for the UK ISA environment using Inverse ETPs for shorting.
## Project Overview
* **Strategy:** Opening Range Liquidity Reversal (Touch & Turn).
* **Asset Class:** US Stocks (e.g., NVDA, ARM, TSLA).
* **Timeframe:** 15-minute chart.
* **Operating Window:** 09:30 - 11:00 EST (Opening of the US Regular Trading Session).
This project implements the "Touch & Turn" scalping strategy for the Trading212 API, optimized for the UK ISA environment.
## Strategy Logic (The Workflow)
1. **Identify the Opening Candle:** Capture the `High`, `Low`, `Open`, and `Close` of the first 15-minute candle of the session (09:30 to 09:45 EST).
2. **Filter for Liquidity:**
- The opening range (`High - Low`) must be at least **25% of the 14-day ATR**. If smaller, the bot stays flat.
1. **Identify Opening Candle:** Capture the 15m candle (09:30 to 09:45 EST).
2. **Filter for Liquidity:** Opening range must be >= 25% of 14-day ATR.
3. **Determine Direction:**
- If the candle is **Bearish** (Close < Open): Prepare for a **LONG** entry.
- If the candle is **Bullish** (Close > Open): Prepare for a **SHORT** entry.
4. **ISA-Compliant Shorting:**
- Since standard shorting is forbidden in a UK ISA, the bot automatically substitutes SHORT signals with **BUY orders for 3x Inverse ETPs** (e.g., 3SLA for TSLA).
5. **Execution:**
- Uses **Market Orders** at 09:45 EST for immediate entry.
- Recalculates SL/TP brackets based on the **Actual Fill Price** fetched from the portfolio.
6. **Risk Management:**
- **Position Sizing:** Risks **1% of account balance** per trade.
- **Virtual Balance:** In demo mode, subtracts £4,750 from total value to simulate a realistic £250 starting point.
- **Targets:** 38.2% Fibonacci retracement level. Risk:Reward ratio of 1:2.
7. **Automatic Exit:** Force close any open positions at 11:00 EST via Market Order.
- Bearish (Close < Open): Prepare **LONG** (Buy at Low).
- Bullish (Close > Open): Prepare **SHORT** (Substitute with **3x Inverse ETP BUY** in ISA).
4. **Execution (09:45 EST):**
- Entry via **Market Order** for immediate fill.
- **Actual Fill Price** fetched from portfolio is used for all bracket calculations.
5. **Hybrid Exit Strategy:**
- **Broker-Side:** Physical **Stop Loss** order placed immediately for protection.
- **Bot-Side:** **Take Profit** monitored manually by polling current market price.
- This bypasses ISA restrictions against multiple pending sell orders for the same shares.
6. **Automatic Exit (11:00 EST):** Force close via Market Order and cleanup pending SL.
## Risk & Capital Management
* **Virtual Balance Simulation:** In demo mode, subtracts £4,750 from total equity to simulate a realistic £250 starting point.
* **1% Risk Rule:** Risks exactly 1% of the Virtual Balance per trade.
* **Capital Partitioning:** Divides total available capital (£250) and risk budget equally among all active ticker threads for the day (max 3).
* **Precision & Minimums:** Automatically detects "precision-mismatch" or "min-quantity-exceeded" errors from T212 and retries with corrected values.
## Technical Architecture
* **API Client (`src/api/client.py`):** Handles REST Basic Auth calls to Trading212.
* **Strategy Engine (`src/strategy/touch_turn.py`):** Calculates setup, Fibonacci levels, and percentage-based targets.
* **Execution Manager (`src/execution/manager.py`):** Handles ticker swapping for ISA mode, market order placement, protection bracketing, and status monitoring.
* **Orchestrator (`main.py`):** Automates the morning routine: scanning, backtesting, and spawning parallel execution threads.
* **Data resilient:** Implements retry loops and random jitter to handle API delays and rate limits (429/403/404 errors).
* **`main.py`:** Daily orchestrator. Scan -> Backtest -> Select Top 3 -> Spawn Parallel Threads. Handles early API verification and unbuffered logging.
* **`src/api/client.py`:** REST wrapper with Basic Auth.
* **`src/strategy/touch_turn.py`:** Setup logic, Fibonacci calculation, and timezone conversion (UTC -> Eastern).
* **`src/execution/manager.py`:** Handles ticker swapping (Inverse ETPs), market entries, hybrid brackets, and retry loops with jitter.
* **`src/strategy/inverse_mapping.py`:** Map of US stocks to 3x Short Inverse ETPs (GraniteShares/Leverage Shares).
## Getting Started
## Resilience Features
1. **Setup Environment:**
```bash
pip install -r requirements.txt
```
2. **Configuration:**
- Set credentials in your `.env` file.
- `ISA_MODE=True` enables Inverse ETP substitution.
3. **Operation:**
- Best deployed via a systemd timer at 09:30 America/New_York time.
* **API Backoff:** Random jitter (1-10s) and exponential retry on 429 errors.
* **Order Tracking:** Uses portfolio checks to infer status if order IDs disappear (404).
* **Unbuffered Logging:** Force-flushes logs to `logs/bot_*.log` immediately for real-time monitoring.
## TODOs
## Operation
- [x] Document the strategy logic.
- [x] Implement ATR calculation and Fibonacci targets.
- [x] Implement 15m candle capture with timezone resilience.
- [x] Implement ISA-compliant shorting via Inverse ETP mapping.
- [x] Implement Risk-based position sizing (Virtual Balance logic).
- [x] Create a leaderboard-based backtesting engine.
- [x] Implement resilient execution (Retry loops, Jitter, Portfolio-checks).
- [x] Add daily logging and P&L tracking (CSV).
1. **Timer:** Service managed by `systemd` timer firing at 09:30 America/New_York.
2. **Tracking:** P&L recorded in `pnl_tracking.csv` (R-multiple based).
3. **Verification:** Always run `./venv/bin/python3 test_api_connection.py` before live days.