Stock Splits: Detect Split Events and Normalize Per-Share Metrics with Python

Stock splits change share counts and per-share values, making historical comparisons difficult. A company reporting $10 EPS before a 2-for-1 split should show $5 adjusted EPS for that period -- but SEC filings contain both adjusted and unadjusted values depending on when they were filed. EdgarTools detects splits from XBRL data and automatically normalizes per-share metrics so you can build consistent time series.

from edgar import Company
from edgar.ttm import detect_splits

company = Company("NVDA")
facts = company.get_facts()
splits = detect_splits(facts.get_all_facts())

for split in splits:
    print(f"{split['date']}: {split['ratio']:.0f}-for-1 split")

A few lines to find every stock split in a company's SEC filing history.

---

Detect Stock Splits from XBRL Data

EdgarTools finds splits by looking for StockSplitConversionRatio facts in XBRL filings. These facts appear in 10-Q, 10-K, and 8-K reports whenever a company reports a split event.

from edgar import Company
from edgar.ttm import detect_splits

nvidia = Company("NVDA")
entity_facts = nvidia.get_facts()
facts = entity_facts.get_all_facts()

splits = detect_splits(facts)

for split in splits:
    print(f"Date: {split['date']}, Ratio: {split['ratio']}")

Output:

Date: 2021-07-20, Ratio: 4.0
Date: 2024-06-10, Ratio: 10.0

How Split Detection Works

The detect_splits() function:

1. Finds split facts - Searches for StockSplitConversionRatio in XBRL concepts
2. Filters stale data - Rejects facts filed >280 days after the split date (historical echoes)
3. Validates duration - Accepts instant facts or short-duration facts (≤31 days), rejects quarterly/annual aggregations
4. Deduplicates - One split per year/ratio combination (same split reported in multiple filings)

The ratio represents the multiplier applied to share counts. A 10-for-1 split has ratio=10.0.

---

Find Stock Split Announcements in 8-K Filings

Companies typically announce splits via 8-K current reports, usually under Item 8.01 ("Other Events") or occasionally Item 5.03 ("Amendments to Articles of Incorporation").

from edgar import Company

apple = Company("AAPL")
filings_8k = apple.get_filings(form="8-K")

# Filter to filings that might contain split announcements
for filing in filings_8k[:50]:  # Check recent 50 filings
    eight_k = filing.obj()

    # Look for Item 8.01 (where most splits are announced)
    if 'Item 8.01' in eight_k.items or 'Item 5.03' in eight_k.items:
        content = eight_k.get('8.01') or eight_k.get('5.03') or ''
        if 'split' in content.lower():
            print(f"{filing.filing_date}: {filing.accession_no}")
            print(f"Items: {', '.join(eight_k.items)}")

Press releases attached as EX-99 exhibits often contain the split announcement details:

if eight_k.has_press_release:
    for release in eight_k.press_releases:
        text = release.text()
        if 'stock split' in text.lower():
            print(f"Split announced: {filing.filing_date}")
            print(text[:500])  # First 500 chars

---

Normalize Per-Share Metrics

Once you've detected splits, use apply_split_adjustments() to retroactively adjust historical data. This makes pre-split and post-split values directly comparable.

from edgar import Company
from edgar.ttm import detect_splits, apply_split_adjustments

tesla = Company("TSLA")
facts = tesla.get_facts().get_all_facts()

# Detect splits
splits = detect_splits(facts)

# Apply adjustments
adjusted_facts = apply_split_adjustments(facts, splits)

# Compare original vs adjusted EPS
eps_facts = [f for f in facts if 'EarningsPerShare' in f.concept]
eps_adjusted = [f for f in adjusted_facts if 'EarningsPerShare' in f.concept]

for orig, adj in zip(eps_facts[:3], eps_adjusted[:3]):
    print(f"{orig.period_end}: ${orig.numeric_value:.2f} → ${adj.numeric_value:.2f}")

Output shows EPS before and after split adjustment:

2021-12-31: $4.90 → $1.63
2022-03-31: $3.22 → $1.07
2022-06-30: $2.27 → $0.76

What Gets Adjusted

The adjustment logic depends on the fact's unit and concept:

Type	Unit Pattern	Adjustment	Examples
Per-share metrics	/share in unit or EarningsPerShare in concept	Divide by ratio	EPS, Dividends per share, Book value per share
Share counts	shares in unit (but not per-share)	Multiply by ratio	Shares outstanding, Weighted average shares
Other metrics	All others	No adjustment	Revenue, Net Income, Assets

# Per-share: Divide by split ratio
# If 10-for-1 split, $10 EPS becomes $1 EPS
adjusted_eps = original_eps / 10.0

# Share counts: Multiply by split ratio
# If 10-for-1 split, 100M shares becomes 1B shares
adjusted_shares = original_shares * 10.0

Retroactive Adjustment Rules

Splits only adjust facts from periods before the split date. The function applies cumulative ratios for multiple splits:

# Example: Company had two splits
# 2021-07-20: 4-for-1 split
# 2024-06-10: 10-for-1 split

# For a fact from 2020:
# - Both splits occurred after 2020
# - Cumulative ratio = 4.0 * 10.0 = 40.0
# - Adjust by dividing by 40

# For a fact from 2022:
# - Only 2024 split occurred after 2022
# - Cumulative ratio = 10.0
# - Adjust by dividing by 10

# For a fact from 2024-07:
# - No splits after this date
# - No adjustment needed (already post-split values)

The function also checks filing dates. If a fact was filed after the split, it's already adjusted by the company and doesn't need further modification.

---

Automatic Split Handling in TTM Calculations

TTM (Trailing Twelve Months) methods automatically detect and apply split adjustments. You don't need to call detect_splits() or apply_split_adjustments() manually when using these methods.

from edgar import Company

nvidia = Company("NVDA")

# TTM calculations handle splits automatically
ttm_revenue = nvidia.get_ttm_revenue()
ttm_net_income = nvidia.get_ttm_net_income()

print(f"TTM Revenue: ${ttm_revenue.value / 1e9:.1f}B")
print(f"TTM Net Income: ${ttm_net_income.value / 1e9:.1f}B")
print(f"As of: {ttm_revenue.as_of_date}")
print(f"Periods: {ttm_revenue.periods}")

For any XBRL concept:

# Per-share metrics are automatically split-adjusted
ttm_eps = nvidia.get_ttm("EarningsPerShareBasic")
print(f"TTM EPS: ${ttm_eps.value:.2f}")

# Share counts are automatically adjusted too
ttm_shares = nvidia.get_ttm("WeightedAverageNumberOfSharesOutstandingBasic")
print(f"Weighted Avg Shares: {ttm_shares.value / 1e9:.2f}B")

Behind the scenes, get_ttm() calls the internal method _get_split_adjusted_facts() which: 1. Gets all facts for the company 2. Detects splits using detect_splits() 3. Applies adjustments using apply_split_adjustments() 4. Returns normalized facts for TTM calculation

---

Complete Workflow: NVIDIA 10-for-1 Split Example

NVIDIA executed a 10-for-1 stock split on June 10, 2024. Let's build a complete workflow that detects this split, normalizes historical EPS, and validates the adjustments.

from edgar import Company
from edgar.ttm import detect_splits, apply_split_adjustments

# Step 1: Get NVIDIA data
nvidia = Company("NVDA")
facts = nvidia.get_facts().get_all_facts()

# Step 2: Detect splits
splits = detect_splits(facts)
print(f"Found {len(splits)} splits:")
for split in splits:
    print(f"  {split['date']}: {split['ratio']:.0f}-for-1")

# Step 3: Filter to EPS facts
eps_facts = [f for f in facts
             if 'EarningsPerShareBasic' in f.concept
             and f.period_end
             and f.numeric_value is not None]

# Sort by period
eps_facts.sort(key=lambda f: f.period_end)

# Step 4: Apply split adjustments
adjusted_facts = apply_split_adjustments(eps_facts, splits)

# Step 5: Compare pre-split periods
print("\nEPS Comparison (split-adjusted):")
print(f"{'Period':<12} {'Original':>12} {'Adjusted':>12} {'Cumulative Ratio':>18}")
print("-" * 58)

for orig, adj in zip(eps_facts[-8:], adjusted_facts[-8:]):
    # Calculate cumulative ratio from the adjustment context
    context = adj.calculation_context or ""
    if "split_adj_ratio" in context:
        ratio = float(context.split("_")[-1])
    else:
        ratio = 1.0

    print(f"{orig.period_end!s:<12} ${orig.numeric_value:>11.2f} "
          f"${adj.numeric_value:>11.2f} {ratio:>17.1f}x")

Expected output:

Found 2 splits:
  2021-07-20: 4-for-1
  2024-06-10: 10-for-1

EPS Comparison (split-adjusted):
Period       Original     Adjusted   Cumulative Ratio
----------------------------------------------------------
2022-01-30   $    4.44   $    0.11             40.0x
2022-05-01   $    1.36   $    0.03             40.0x
2022-07-31   $    0.51   $    0.01             40.0x
2022-10-30   $    0.58   $    0.01             40.0x
2023-01-29   $    0.88   $    0.02             40.0x
2023-04-30   $    1.09   $    0.03             40.0x
2023-07-30   $    2.70   $    0.07             40.0x
2024-01-28   $    5.16   $    0.13             40.0x

Periods before both splits get cumulative adjustment of 40x (4 × 10). Periods between the splits would get 10x adjustment. Periods after June 2024 need no adjustment.

---

Common Analysis Patterns

Track Split History Across Multiple Companies

from edgar import Company
from edgar.ttm import detect_splits

tech_stocks = ["AAPL", "MSFT", "GOOGL", "AMZN", "NVDA", "TSLA"]

for ticker in tech_stocks:
    try:
        company = Company(ticker)
        facts = company.get_facts()
        splits = detect_splits(facts.get_all_facts())

        if splits:
            print(f"\n{ticker} ({company.name}):")
            for split in splits:
                print(f"  {split['date']}: {split['ratio']:.1f}-for-1")
        else:
            print(f"\n{ticker}: No splits detected")

    except Exception as e:
        print(f"\n{ticker}: Error - {e}")

Build Split-Adjusted Time Series

Compare EPS over multiple years with consistent per-share values:

from edgar import Company
from edgar.ttm import detect_splits, apply_split_adjustments
import pandas as pd

company = Company("AAPL")
facts = company.get_facts().get_all_facts()

# Get splits and adjust facts
splits = detect_splits(facts)
adjusted_facts = apply_split_adjustments(facts, splits)

# Extract EPS time series
eps_data = []
for f in adjusted_facts:
    if 'EarningsPerShareBasic' in f.concept and f.fiscal_period == 'FY':
        eps_data.append({
            'fiscal_year': f.fiscal_year,
            'period_end': f.period_end,
            'eps': f.numeric_value
        })

# Create DataFrame
df = pd.DataFrame(eps_data).sort_values('fiscal_year')
print(df)

# Calculate growth rates on split-adjusted data
df['yoy_growth'] = df['eps'].pct_change() * 100
print(f"\nAverage EPS growth: {df['yoy_growth'].mean():.1f}%")

Validate Split Adjustments Against Company Reports

Companies publish adjusted historical data after splits. You can validate the adjustment logic:

from edgar import Company
from edgar.ttm import detect_splits, apply_split_adjustments

company = Company("NVDA")
facts = company.get_facts().get_all_facts()

splits = detect_splits(facts)
adjusted = apply_split_adjustments(facts, splits)

# Find the same period reported before and after split
# Pre-split: Filed in 2023 10-K, reports 2023 EPS (pre-split basis)
# Post-split: Filed in 2024 10-K, reports 2023 EPS (post-split basis)

eps_2023_filings = [f for f in adjusted
                     if 'EarningsPerShareBasic' in f.concept
                     and f.fiscal_year == 2023]

# Group by filing date
from collections import defaultdict
by_filing = defaultdict(list)
for f in eps_2023_filings:
    by_filing[f.filing_date].append(f)

# Compare values across filings
for filing_date in sorted(by_filing.keys())[:3]:
    facts_list = by_filing[filing_date]
    if facts_list:
        f = facts_list[0]
        print(f"Filed {filing_date}: 2023 EPS = ${f.numeric_value:.2f}")

Calculate Split-Adjusted Market Cap History

Combine share counts and stock prices to build historical market cap:

from edgar import Company
from edgar.ttm import detect_splits, apply_split_adjustments

company = Company("AAPL")
facts = company.get_facts().get_all_facts()

splits = detect_splits(facts)
adjusted = apply_split_adjustments(facts, splits)

# Get split-adjusted shares outstanding
shares_facts = [f for f in adjusted
                if 'CommonStockSharesOutstanding' in f.concept
                and f.fiscal_period == 'FY']

for f in sorted(shares_facts, key=lambda x: x.period_end)[-5:]:
    shares_b = f.numeric_value / 1e9
    print(f"{f.period_end}: {shares_b:.2f}B shares (split-adjusted)")

---

API Quick Reference

Detection and Adjustment Functions

Function	Parameters	Returns	Description
detect_splits(facts)	facts: List of FinancialFact	List of dicts with date and ratio	Find all stock splits in facts
apply_split_adjustments(facts, splits)	facts: List of FinancialFact splits: List of split dicts	List of adjusted FinancialFact	Apply retroactive adjustments

Company Methods with Automatic Split Handling

Method	Returns	Description
company.get_ttm(concept, as_of=None)	TTMMetric	TTM for any concept (split-adjusted)
company.get_ttm_revenue(as_of=None)	TTMMetric	TTM revenue (split-adjusted)
company.get_ttm_net_income(as_of=None)	TTMMetric	TTM net income (split-adjusted)

Split Detection Parameters

The detection logic uses these constants from edgar.ttm.calculator:

Constant	Value	Purpose
MAX_SPLIT_LAG_DAYS	280	Maximum days between split date and filing date
MAX_SPLIT_DURATION_DAYS	31	Maximum period duration for split facts

---

Things to Know

Splits apply retroactively to historical data only. Facts from periods after the split date don't need adjustment -- they're already reported on a post-split basis.

Filing date matters for restated facts. If a fact was filed after a split date, the company has already adjusted it. The adjustment logic checks filing_date to avoid double-adjusting.

Multiple splits compound. A company with a 4-for-1 split in 2021 and a 10-for-1 split in 2024 requires a cumulative adjustment of 40x for pre-2021 data.

Reverse splits work automatically. A 1-for-10 reverse split has ratio=0.1. Per-share metrics get divided by 0.1 (multiplied by 10), which correctly increases the adjusted historical EPS.

Balance sheet items don't need adjustment. Assets, liabilities, and equity are not per-share values. Total stockholders' equity stays the same regardless of share count.

Not all facts have filing dates. The adjustment logic handles None filing dates by assuming the fact needs adjustment if it's from before the split.

Instant facts are preferred. Split events are moment-in-time occurrences. The detector accepts instant facts (no period_start) or short-duration facts (≤31 days) but rejects quarterly/annual durations.

8-K filing timing varies. While Item 8.01 is common for split announcements, check Items 5.03 and exhibit press releases. Not all companies follow the same disclosure pattern.

Weighted average share counts need special handling. These represent time-weighted averages over a period. For Q4 EPS derivation with splits, use the formula: Q4 shares = 4 × annual_shares - 3 × YTD_9M_shares.

---

Related

• Financial Data - Extract financial statements and metrics
• 8-K Current Reports - Parse material event filings
• Company Facts API - Access XBRL facts programmatically