Run2 quarter comparison

Quarter-to-Quarter Holdings Comparison

Overview

Compare a fund's holdings between two quarters (e.g., Q2 to Q3) to identify which stocks received increased or decreased investment.

Core Workflow

Step 1: Get Accession Numbers for Both Quarters

import pandas as pd

# Q2 (June 30)
coverpage_q2 = pd.read_csv('/root/2025-q2/COVERPAGE.tsv', sep='\t')
fund_q2 = coverpage_q2[coverpage_q2['FILINGMANAGER_NAME'].str.contains('Berkshire', case=False)]
accession_q2 = fund_q2['ACCESSION_NUMBER'].values[0]

# Q3 (September 30)
coverpage_q3 = pd.read_csv('/root/2025-q3/COVERPAGE.tsv', sep='\t')
fund_q3 = coverpage_q3[coverpage_q3['FILINGMANAGER_NAME'].str.contains('Berkshire', case=False)]
accession_q3 = fund_q3['ACCESSION_NUMBER'].values[0]

Step 2: Load Holdings for Both Quarters

# Load INFOTABLE for both quarters
infotable_q2 = pd.read_csv('/root/2025-q2/INFOTABLE.tsv', sep='\t', low_memory=False)
infotable_q3 = pd.read_csv('/root/2025-q3/INFOTABLE.tsv', sep='\t', low_memory=False)

# Filter to specific fund
fund_q2_holdings = infotable_q2[infotable_q2['ACCESSION_NUMBER'] == accession_q2]
fund_q3_holdings = infotable_q3[infotable_q3['ACCESSION_NUMBER'] == accession_q3]

Step 3: Aggregate by CUSIP

# Group Q2 holdings by CUSIP
q2_summary = fund_q2_holdings.groupby('CUSIP').agg({
    'VALUE': 'sum',
    'NAMEOFISSUER': 'first',
    'SSHPRNAMT': 'sum'
}).reset_index()
q2_summary.columns = ['CUSIP', 'VALUE_Q2', 'NAMEOFISSUER_Q2', 'SHARES_Q2']

# Group Q3 holdings by CUSIP
q3_summary = fund_q3_holdings.groupby('CUSIP').agg({
    'VALUE': 'sum',
    'NAMEOFISSUER': 'first',
    'SSHPRNAMT': 'sum'
}).reset_index()
q3_summary.columns = ['CUSIP', 'VALUE_Q3', 'NAMEOFISSUER_Q3', 'SHARES_Q3']

Step 4: Merge and Calculate Changes

# Merge on CUSIP (outer join to capture new and exited positions)
comparison = pd.merge(q2_summary, q3_summary, on='CUSIP', how='outer')

# Fill NaN with 0 (positions not held in that quarter)
comparison['VALUE_Q2'] = comparison['VALUE_Q2'].fillna(0)
comparison['VALUE_Q3'] = comparison['VALUE_Q3'].fillna(0)

# Calculate value change
comparison['VALUE_CHANGE'] = comparison['VALUE_Q3'] - comparison['VALUE_Q2']
comparison['SHARE_CHANGE'] = comparison['SHARES_Q3'].fillna(0) - comparison['SHARES_Q2'].fillna(0)

Step 5: Identify Top Changes

# Top 5 stocks with INCREASED investment (positive value change)
top_5_increases = comparison[comparison['VALUE_CHANGE'] > 0].nlargest(5, 'VALUE_CHANGE')

# Top 5 stocks with DECREASED investment (negative value change)
top_5_decreases = comparison[comparison['VALUE_CHANGE'] < 0].nsmallest(5, 'VALUE_CHANGE')

# Get CUSIPs
increased_cusips = top_5_increases['CUSIP'].tolist()
decreased_cusips = top_5_decreases['CUSIP'].tolist()

print("Top 5 Increased Positions:")
print(top_5_increases[['CUSIP', 'NAMEOFISSUER_Q2', 'VALUE_Q2', 'VALUE_Q3', 'VALUE_CHANGE']])

Important Considerations

Handling Missing Companies

# Use first non-null NAMEOFISSUER
comparison['NAMEOFISSUER'] = comparison['NAMEOFISSUER_Q2'].fillna(comparison['NAMEOFISSUER_Q3'])

# New positions (not in Q2): NAMEOFISSUER_Q2 is NaN
new_positions = comparison[comparison['VALUE_Q2'] == 0][comparison['VALUE_CHANGE'] > 0]

# Exited positions (not in Q3): NAMEOFISSUER_Q3 is NaN
exited_positions = comparison[comparison['VALUE_Q3'] == 0][comparison['VALUE_CHANGE'] < 0]

Duplicate CUSIP Handling

Some fund filings may have the same CUSIP listed multiple times (different options, derivative types, etc.). The aggregation via groupby().agg() handles this automatically by summing.

Data Type Warnings

# Avoid DtypeWarning when loading INFOTABLE
infotable = pd.read_csv(
    'INFOTABLE.tsv',
    sep='\t',
    low_memory=False,  # Critical for INFOTABLE
    dtype={'ACCESSION_NUMBER': 'str', 'CUSIP': 'str', 'VALUE': 'int64'}
)

Complete Example: Berkshire Q2 to Q3

import pandas as pd

# Load files
infotable_q2 = pd.read_csv('/root/2025-q2/INFOTABLE.tsv', sep='\t', low_memory=False)
infotable_q3 = pd.read_csv('/root/2025-q3/INFOTABLE.tsv', sep='\t', low_memory=False)
coverpage_q2 = pd.read_csv('/root/2025-q2/COVERPAGE.tsv', sep='\t')
coverpage_q3 = pd.read_csv('/root/2025-q3/COVERPAGE.tsv', sep='\t')

# Get accession numbers
accession_q2 = coverpage_q2[coverpage_q2['FILINGMANAGER_NAME'] == 'Berkshire Hathaway Inc']['ACCESSION_NUMBER'].values[0]
accession_q3 = coverpage_q3[coverpage_q3['FILINGMANAGER_NAME'] == 'Berkshire Hathaway Inc']['ACCESSION_NUMBER'].values[0]

# Get holdings
q2_holdings = infotable_q2[infotable_q2['ACCESSION_NUMBER'] == accession_q2]
q3_holdings = infotable_q3[infotable_q3['ACCESSION_NUMBER'] == accession_q3]

# Aggregate by CUSIP
q2_agg = q2_holdings.groupby('CUSIP')['VALUE'].sum().reset_index()
q2_agg.columns = ['CUSIP', 'VALUE_Q2']

q3_agg = q3_holdings.groupby('CUSIP')['VALUE'].sum().reset_index()
q3_agg.columns = ['CUSIP', 'VALUE_Q3']

# Merge
comparison = pd.merge(q2_agg, q3_agg, on='CUSIP', how='outer')
comparison['VALUE_Q2'] = comparison['VALUE_Q2'].fillna(0)
comparison['VALUE_Q3'] = comparison['VALUE_Q3'].fillna(0)
comparison['VALUE_CHANGE'] = comparison['VALUE_Q3'] - comparison['VALUE_Q2']

# Top 5 increases
top_5 = comparison[comparison['VALUE_CHANGE'] > 0].nlargest(5, 'VALUE_CHANGE')
cusips = top_5['CUSIP'].tolist()

print(cusips)
# Output: ['02079K305', '037833100', '025816109', '166764100', '674599105']

Output Format

Result should be a list of CUSIPs ordered by value change (descending):

{
    "top_5_cusips": ["CUSIP1", "CUSIP2", "CUSIP3", "CUSIP4", "CUSIP5"],
    "changes": [
        {"cusip": "CUSIP1", "value_change": 4338397000000},
        {"cusip": "CUSIP2", "value_change": 3208516000000}
    ]
}

Validation Checklist

• Confirm accession numbers exist for both Q2 and Q3
• Verify merging on CUSIP is case-sensitive and exact
• Check that outer join captures both new and exited positions
• Validate top 5 are sorted by VALUE_CHANGE (descending)
• Confirm VALUE_CHANGE = VALUE_Q3 - VALUE_Q2 (should be positive for increases)
• Handle NaN values correctly before calculation