Multi Factor Strategy

---
name: multi-factor-strategy
description: Guide users to create multi-factor stock selection strategies and generate independent YAML configuration files
---
{"homepage":"https://gitcode.com/datavoid/quantcli","user-invocable":true}

# Multi-Factor Strategy Assistant

Guide you to create multi-factor stock selection strategies and generate independent YAML configuration files.

## Install quantcli

```bash
# Install from PyPI (recommended)
pip install quantcli

# Or install from source
git clone https://gitcode.com/datavoid/quantcli.git
cd quantcli
pip install -e .
```

Verify installation:
```bash
quantcli --help
```

## Quick Start

A complete multi-factor stock selection strategy YAML example:

```yaml
name: Value-Growth Hybrid Strategy
version: 1.0.0
description: ROE + Momentum factor stock selection

screening:
  fundamental_conditions:    # Stage 1: Financial condition screening
    - "roe > 0.10"           # ROE > 10%
    - "pe_ttm < 30"          # P/E < 30
    - "pe_ttm > 0"           # Exclude losses
  daily_conditions:          # Stage 2: Price condition screening
    - "close > ma10"         # Above 10-day MA
  limit: 100                 # Keep at most 100 stocks

# Factor configuration (supports two methods, factors at top level)
factors:
  # Method 1: Inline factor definition
  - name: ma10_deviation
    expr: "(close - ma(close, 10)) / ma(close, 10)"
    direction: negative
    description: "10-day MA deviation"

  # Method 2: External reference (reference factor files in factors/ directory, include .yaml suffix)
  - factors/alpha_001.yaml
  - factors/alpha_008.yaml

ranking:
  weights:                   # Weight fusion
    ma10_deviation: 0.20     # Inline factor
    factors/alpha_001.yaml: 0.40  # External reference factor
    factors/alpha_008.yaml: 0.40
  normalize: zscore          # Normalization method

output:
  limit: 30                  # Output top 30 stocks
  columns: [symbol, name, score, roe, pe_ttm, close, ma10_deviation]
```

### Factor Configuration Methods

**Factor configuration supports two methods (can be mixed):**

| Method | Type | Example | Description |
|--------|------|---------|-------------|
| **Inline** | `dict` | `{name: xxx, expr: "..."}` | Define expression directly in YAML |
| **External** | `str` | `factors/alpha_001.yaml` | Load factor file from `factors/` directory |

**Example: Mixed usage**

```yaml
factors:
  # Inline: Custom factor
  - name: custom_momentum
    expr: "close / delay(close, 20) - 1"
    direction: positive

  # External: Alpha101 factor library (include .yaml suffix)
  - factors/alpha_001.yaml
  - factors/alpha_005.yaml
  - factors/alpha_009.yaml

ranking:
  weights:
    custom_momentum: 0.3
    factors/alpha_001.yaml: 0.3
    factors/alpha_005.yaml: 0.2
    factors/alpha_009.yaml: 0.2
```

Run strategy:
```bash
quantcli filter run -f your_strategy.yaml
```

## Invocation

```
/multi-factor-strategy
```

## Available Expression Functions

### Data Processing Functions
| Function | Usage | Description |
|----------|-------|-------------|
| delay | `delay(x, n)` | Lag n periods |
| ma | `ma(x, n)` | Simple moving average |
| ema | `ema(x, n)` | Exponential moving average |
| rolling_sum | `rolling_sum(x, n)` | Rolling sum |
| rolling_std | `rolling_std(x, n)` | Rolling standard deviation |

### Technical Indicator Functions
| Function | Usage | Description |
|----------|-------|-------------|
| rsi | `rsi(x, n=14)` | Relative strength index |
| correlation | `correlation(x, y, n)` | Correlation coefficient |
| cross_up | `cross_up(a, b)` | Golden cross (a crosses above b) |
| cross_down | `cross_down(a, b)` | Death cross (a crosses below b) |

### Ranking & Normalization Functions
| Function | Usage | Description |
|----------|-------|-------------|
| rank | `rank(x)` | Cross-sectional ranking (0-1) |
| zscore | `zscore(x)` | Standardization |
| sign | `sign(x)` | Sign function |
| clamp | `clamp(x, min, max)` | Clipping function |

### Conditional Functions
| Function | Usage | Description |
|----------|-------|-------------|
| where | `where(cond, t, f)` | Conditional selection |
| if | `if(cond, t, f)` | Conditional selection (alias) |

### Base Fields
| Field | Description |
|-------|-------------|
| open, high, low, close | OHLC prices |
| volume | Trading volume |
| pe, pb | P/E ratio, P/B ratio |
| roe | Return on equity |
| netprofitmargin | Net profit margin |

## Guided Workflow

### Step 1: Strategy Goal定位

I will first understand your strategy needs:
- **Strategy Type**: Value, Growth, Momentum, Volatility, Hybrid
- **Selection Count**: Concentrated(10-30), Medium(50-100), Diversified(200+)
- **Holding Period**: Intraday, Short-term(week), Medium-term(month), Long-term(quarter)

### Step 2: Factor Selection

Based on your strategy goals, recommend suitable factor combinations:

**Common Fundamental Factors**:
| Factor | Expression | Direction | Description |
|--------|------------|-----------|-------------|
| roe | `roe` | positive | Return on equity |
| pe | `pe` | negative | Lower P/E is better |
| pb | `pb` | negative | Price-to-book ratio |
| netprofitmargin | `netprofitmargin` | positive | Net profit margin |
| revenue_growth | `revenue_yoy` | positive | Revenue growth rate |

**Common Technical Factors**:
| Factor | Expression | Direction | Description |
|--------|------------|-----------|-------------|
| momentum | `(close/delay(close,20))-1` | positive | N-day momentum |
| ma_deviation | `(close-ma(close,10))/ma(close,10)` | negative | MA deviation |
| ma_slope | `(ma(close,10)-delay(ma(close,10),5))/delay(ma(close,10),5)` | positive | MA slope |
| volume_ratio | `volume/ma(volume,5)` | negative | Volume ratio |

**Alpha101 Built-in Factors** (can reference `{baseDir}/alpha101/alpha_XXX`):

QuantCLI includes 40 WorldQuant Alpha101 factors that can be directly referenced:

| Factor | Category | Description |
|--------|----------|-------------|
| `alpha101/alpha_001` | Reversal | 20-day new high then decline |
| `alpha101/alpha_002` | Reversal | Down volume bottom |
| `alpha101/alpha_003` | Volatility | Low volatility stability |
| `alpha101/alpha_004` | Capital Flow | Net capital inflow |
| `alpha101/alpha_005` | Trend | Uptrend |
| `alpha101/alpha_008` | Capital Flow | Capital inflow |
| `alpha101/alpha_009` | Momentum | Long-term momentum |
| `alpha101/alpha_010` | Reversal | MA deviation reversal |
| `alpha101/alpha_011` ~ `alpha_020` | Extended | Volatility, momentum, price-volume factors |
| `alpha101/alpha_021` ~ `alpha_030` | Extended | Price-volume, trend, strength factors |
| `alpha101/alpha_031` ~ `alpha_040` | Extended | Position, volatility, capital factors |

**View all built-in factors:**
```bash
quantcli factors list
```

**Usage Example:**
```yaml
factors:
  - alpha101/alpha_001   # Reversal factor
  - alpha101/alpha_008   # Capital inflow
  - alpha101/alpha_029   # 5-day momentum
ranking:
  weights:
    alpha101/alpha_001: 0.4
    alpha101/alpha_008: 0.3
    alpha101/alpha_029: 0.3
```

**Screening Conditions Example**:
```yaml
screening:
  conditions:
    - "roe > 0.10"              # ROE > 10%
    - "netprofitmargin > 0.05"  # Net profit margin > 5%
```

### Step 3: Weight Configuration

Allocate weights based on factor importance, 0 means only for screening, not scoring:

```yaml
ranking:
  weights:
    # Fundamental factors
    roe: 0.30
    pe: 0.20
    # Technical factors
    ma_deviation: 0.30
    momentum: 0.20
  normalize: zscore
```

### Step 4: Generate Strategy File

I will generate a complete strategy YAML file for you:

```yaml
name: Your Strategy Name
version: 1.0.0
description: Strategy description

# Stage 1: Fundamental screening
screening:
  conditions:
    - "roe > 0.10"
    - "pe < 30"
  limit: 200

# Stage 2: Technical ranking
ranking:
  weights:
    roe: 0.30
    pe: 0.20
    ma_deviation: 0.30
    momentum: 0.20
  normalize: zscore

output:
  columns: [symbol, score, rank, roe, pe, momentum]
  limit: 30
```

### Step 5: Run & Evaluate

**Run strategy**:
```bash
quantcli filter run -f your_strategy.yaml --top 30
```

**Evaluation points**:
1. **Selected stock count**: Check if screening conditions are reasonable
2. **Factor distribution**: Distribution of factor scores
3. **Industry diversification**: Avoid over-concentration

## FAQ

**Q: How to allocate factor weights?**
A: Core factors 0.3-0.4, auxiliary factors 0.1-0.2, ensure weights sum close to 1

**Q: Screening conditions too strict resulting in empty results?**
A: Gradually relax conditions, first see how many stocks meet each condition

**Q: What expression syntax is supported?**
A: Supports 40+ built-in functions: `ma()`, `ema()`, `delay()`, `rolling_sum()`, `rsi()`, `rank()`, `zscore()`, etc.