13f fund analysis

13F Fund Analysis Skill

Overview

This skill enables deep analysis of 13F fund filings using accession numbers, extracting key metrics like AUM (Assets Under Management), number of holdings, and detailed stock positions.

When to Use

Use this skill whenever you need to:

• Extract AUM from a specific fund filing
• Count the number of stocks held by a fund
• Get detailed holdings data (positions, shares, values)
• Analyze a fund's portfolio composition

Prerequisites

You must have:

• The accession_number from a 13F filing (get this using the 13f-fund-search skill)
• The quarter folder path (/root/2025-q2/ or /root/2025-q3/)

Analysis Process

Step 1: Locate the Filing Files

Given an accession_number, find the corresponding filing folder:

# Accession numbers map to folder names, typically in format:
# /root/2025-q2/[cik]/[accession_folder]/
find /root/2025-q2/ -name "*accession*" -o -name "*.csv" -o -name "*.json"

Step 2: Extract Fund Metadata

Look for files containing:

• COVERPAGE - Contains fund name, AUM, manager info
• HOLDINGS or INFOTABLE - Contains individual stock positions

import pandas as pd
import json
import os

def analyze_fund(quarter_path, accession_number):
    """
    Analyze a fund's holdings and AUM given accession number.

    Args:
        quarter_path: Path to quarter (e.g., '/root/2025-q2/')
        accession_number: The accession number from search

    Returns:
        Dictionary with fund metrics
    """
    results = {
        'accession_number': accession_number,
        'aum': None,
        'number_of_holdings': 0,
        'holdings': [],
        'fund_name': None
    }

    # Search for accession folder
    for root, dirs, files in os.walk(quarter_path):
        # Check if this directory contains our accession number
        if accession_number in root:
            # Look for COVERPAGE
            for file in files:
                if 'COVERPAGE' in file.upper():
                    file_path = os.path.join(root, file)
                    try:
                        if file.endswith('.csv'):
                            df = pd.read_csv(file_path)
                        else:
                            with open(file_path, 'r') as f:
                                df = json.load(f)
                                if isinstance(df, list):
                                    df = pd.DataFrame(df)

                        # Extract AUM - look for columns containing 'aum' or 'assets'
                        for col in df.columns:
                            if 'aum' in col.lower() or 'assets' in col.lower():
                                if len(df) > 0:
                                    results['aum'] = df[col].iloc[0]
                                    break

                        # Extract fund name
                        for col in df.columns:
                            if 'name' in col.lower() or 'fund' in col.lower():
                                if len(df) > 0:
                                    results['fund_name'] = df[col].iloc[0]
                                    break
                    except Exception as e:
                        pass

            # Look for holdings/infotable
            for file in files:
                if 'INFOTABLE' in file.upper() or 'HOLDINGS' in file.upper() or 'POSITION' in file.upper():
                    file_path = os.path.join(root, file)
                    try:
                        if file.endswith('.csv'):
                            holdings_df = pd.read_csv(file_path)
                        else:
                            with open(file_path, 'r') as f:
                                holdings_df = json.load(f)
                                if isinstance(holdings_df, list):
                                    holdings_df = pd.DataFrame(holdings_df)

                        results['number_of_holdings'] = len(holdings_df)
                        results['holdings'] = holdings_df.to_dict('records')
                    except Exception as e:
                        pass

    return results

# Usage
fund_data = analyze_fund('/root/2025-q3/', '0001234567-25-000123')
print(f"AUM: ${fund_data['aum']}")
print(f"Number of holdings: {fund_data['number_of_holdings']}")

Step 3: Parse Holdings Data

Holdings typically include:

• CUSIP - Stock identifier
• Shares - Number of shares held
• Value - Market value of position
• Stock Name - Company name

def get_holdings_summary(fund_data):
    """Get top holdings by value."""
    holdings_df = pd.DataFrame(fund_data['holdings'])

    # Sort by value (look for value or market_value column)
    value_col = None
    for col in holdings_df.columns:
        if 'value' in col.lower():
            value_col = col
            break

    if value_col:
        holdings_df = holdings_df.sort_values(by=value_col, ascending=False)

    return holdings_df.head(10)  # Top 10

Output Format

Returns fund analysis object:

{
    "accession_number": "0001234567-25-000123",
    "fund_name": "Renaissance Technologies",
    "aum": 12345000000,
    "number_of_holdings": 245,
    "holdings": [
        {
            "cusip": "000000001",
            "name": "COMPANY NAME",
            "shares": 1000000,
            "value": 50000000
        }
    ]
}

Key Fields to Extract

Field	Description	Source
AUM	Assets Under Management	COVERPAGE
Number of Holdings	Count of stock positions	INFOTABLE/HOLDINGS
CUSIP	Stock identifier	INFOTABLE/HOLDINGS
Shares	Number of shares held	INFOTABLE/HOLDINGS
Value	Market value of position	INFOTABLE/HOLDINGS

Common Column Names

The 13F files may use varying column names:

• AUM: "AUM", "total_aum", "assets_under_management"
• Holdings: "infotable_entry", "position", "holding"
• Value: "value", "market_value", "value_of_shares"

Tips

• Check file structure first—CSV vs JSON
• AUM may be in different units (thousands vs actual)
• Holdings count = number of rows in INFOTABLE
• Some funds may have partial holdings data in Q3