A-Share Quant Selector

You can add custom strategies by following these steps:

1. Create a new file in the strategy/ directory.
2. Inherit from the BaseStrategy class.
3. Implement the calculate_indicators(self, df) and select_stocks(self, df, stock_name='') methods.
4. Add any necessary parameters to config/strategy_params.yaml.

The system will automatically detect and execute the new strategy.

from strategy.base_strategy import BaseStrategy

class MyStrategy(BaseStrategy):
    def __init__(self, params=None):
        super().__init__("我的策略", params)
    
    def calculate_indicators(self, df):
        # Calculate indicators
        return df
    
    def select_stocks(self, df, stock_name=''):
        # Selection logic
        return signals

The project includes a web-based dashboard accessible at http://localhost:5000. It provides the following views:

• System Overview (📊 系统概览): View total stock counts and the latest data update date.
• Stock List (📈 股票列表): Browse all stocks with basic information and search capabilities.
• Selection Results (🎯 选股结果): View detailed signals from executed stock selection strategies.
• Strategy Configuration (⚙️ 策略配置): Modify strategy parameters online.

Use the following commands to execute the system's core workflows:

• Full Workflow (Update data $\rightarrow$ Select stocks $\rightarrow$ Send DingTalk notifications):

  python3 main.py run

• Quick Test (Process only the first 500 stocks):

  python3 main.py run --max-stocks 500

• Category Filtering (Only select stocks in the 'bowl center' category):

  python3 main.py run --category bowl_center

• Start Web Management Interface (Default port 5000):

  python3 main.py web

• Incremental Update (Run daily after market close):

  python3 main.py update

Follow these steps to set up the environment and run the quant selector for the first time:

1. Clone the repository:

   git clone https://github.com/Dzy-HW-XD/a-share-quant-selector.git
   cd a-share-quant-selector

2. Install dependencies:

   pip3 install -r requirements.txt

3. Configure DingTalk notifications (Optional): Edit config/config.yaml to provide your webhook and secret.
4. Initialize data (Fetch 6 years of historical data):

   python3 main.py init

The B1 Pattern Match feature uses a 3D similarity engine to rank stocks based on 10 historical successful cases. It evaluates stocks across four dimensions: Trend Structure (30%), Volume Pattern (25%), Price Shape (25%), and KDJ State (20%).

CLI Usage

• Enable B1 matching:

  python3 main.py run --b1-match

• Custom lookback and similarity:

  python3 main.py run --b1-match --lookback-days 30 --min-similarity 70

Configuration

You can adjust the matching engine in config/strategy_params.yaml under the B1PatternMatch section, including min_similarity, lookback_days, and the weights for each dimension.

To automate the stock selection process daily, add a task to your crontab. For example, to run the selection process every weekday at 15:05 and log the output to /var/log/quant-csv/run.log, use the following command:

1. Open crontab editor: crontab -e
2. Add the following line (adjust the path to your project directory):

# 每天15:05执行，仅工作日
5 15 * * 1-5 cd /root/quant-csv && /usr/bin/python3 main.py run >> /var/log/quant-csv/run.log 2>&1

The system is designed to allow expansion from B1 to B2 or B3 pattern matching. To implement a new pattern (e.g., B2), follow these steps:

1. Configuration: Add the new pattern configuration (e.g., B2_PERFECT_CASES) in pattern_config.py.
2. Library Class: Create a new library class (e.g., B2PatternLibrary) modeled after B1PatternLibrary.
3. CLI Argument: Add a new command-line parameter (e.g., --b2-match) in main.py.
4. Notification: Register a new notification method (e.g., send_b2_match_results()) in dingtalk_notifier.py.

To adjust the number of matching results displayed in DingTalk notifications, edit the config/strategy_params.yaml file. Increasing this value results in longer messages.

Recommended value: 10-20.

B1PatternMatch:
  top_n_results: 15  # Default displays the top 15 matching results

To enable automated notifications, edit config/config.yaml and provide your DingTalk webhook and secret.

Notifications include:

• Standard Selection Results: Displays strategy parameters (e.g., N, M, CAP, J_VAL), timestamp, and categorized stock counts (e.g., 🥣 回落碗中, 📊 靠近多空线). Individual stock details include price, J-value, key K-line dates, and an included K-line image showing trends and volume.
• B1 Perfect Pattern Matching: Triggered via the --b1-match flag. It provides a ranked list of stocks based on similarity percentage, matching historical cases, and a breakdown of scores across four dimensions: 趋势 (Trend), KDJ, 量能 (Volume), and 形态 (Pattern).

# Edit config/config.yaml
webhook: 'YOUR_WEBHOOK_URL'
secret: 'YOUR_SECRET'

The system uses an internal _smart_update mechanism to optimize data fetching and avoid unnecessary network calls:

1. Time Check: If the current time is before 15:00 (market close), the system skips network updates and uses existing local data to prevent fetching incomplete intraday data.
2. Data Freshness Check: It performs a sampled check (up to 100 stocks) to see if the latest date in the local CSV matches today's date. If all sampled stocks already have today's data, the update is skipped entirely.
3. Incremental Update: If updates are needed, it triggers fetcher.daily_update() to fetch only the missing data.