# quant

**Repository Path**: chenchengf/quant

## Basic Information

- **Project Name**: quant
- **Description**: No description available
- **Primary Language**: Python
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-08-07
- **Last Updated**: 2026-08-16

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Quant Framework

A-share quant research framework. See `docs/superpowers/specs/2026-08-04-quant-framework-design.md` for design.

## Install
```bash
pip install -r requirements.txt
```

## Usage

### Run tests
```bash
pytest
```

### Run a backtest
```bash
python jobs/run_backtest.py \
    --strategy momentum_top20 \
    --start 2024-01-01 --end 2026-07-31 \
    --output reports/momentum.html
```

### ML strategy (LightGBM)

Three-step flow: train → backtest → signal.

```bash
# 1. Train (once per model version, ~5-30 min depending on data size)
python jobs/train_model.py \
    --train-start 2021-01-01 --train-end 2023-12-31 \
    --label-horizon 5 \
    --output models/lightgbm_v1

# 2. Backtest
python jobs/run_backtest.py \
    --strategy lightgbm \
    --start 2024-01-01 --end 2026-07-31 \
    --model-path models/lightgbm_v1

# 3. Daily signal
python jobs/daily_signal.py \
    --strategy lightgbm \
    --model-path models/lightgbm_v1
```

Training prints val IC (Spearman rank correlation between predicted and
realized 5-day returns). Rule of thumb: IC ≥ 0.02 weak, ≥ 0.05 medium,
≥ 0.08 strong. Below 0.02 → warning printed but model still saved.

Models live under `models/<name>/` (gitignored). Each contains `model.pkl`
and `meta.json` (train dates, feature list, metrics).

### Generate daily signal
```bash
python jobs/daily_signal.py --strategy momentum_top20
```

### Refresh data
```bash
python jobs/daily_update.py
```

### Install crontab
```bash
crontab jobs/crontab.example
```

## Project structure

See `docs/superpowers/specs/2026-08-04-quant-framework-design.md` section 4.