تشغيل أي مهارة في Manus بنقرة واحدة

fetch-holdings

النجوم١

التفرعات٢

آخر تحديث١٧ يوليو ٢٠٢٦ في ١٣:١١

Use when fetching, loading, or matching ETF holdings data for Tuleva portfolio analysis. Covers data sources, ISIN matching, and common pitfalls.

التثبيت

التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.

تشغيل في Manus

المصدر

TulevaEE

TulevaEE/reporting-engine

فتح مستودع GitHub عرض مستودعات المنشئ

تنزيل

تشغيل في Manus

المهن ذات الصلةSOC

استنادا إلى تصنيف SOC المهني

مطوّرو البرمجياتمهن الحاسوب والرياضيات·SOC 15-1252

SKILL.md

readonly

تشغيل أي مهارة بنقرة واحدة

name	fetch-holdings
description	Use when fetching, loading, or matching ETF holdings data for Tuleva portfolio analysis. Covers data sources, ISIN matching, and common pitfalls.

Fetching & Matching ETF Holdings

Golden rule: use ISIN as the primary stock identifier

ISIN is the only universal, unambiguous equity identifier. Never use Ticker|Location as a primary key — tickers differ across providers (e.g. HDFCB vs HDFCBANK, BRK/B vs BRKB, 0Y3K|Ireland vs EATON|United States).

Data sources

ACWI benchmark — use SPDR, not iShares

SPDR MSCI ACWI ETF (SPYY GY) publishes daily holdings WITH ISINs:

https://www.ssga.com/uk/en_gb/intermediary/etfs/library-content/products/fund-data/etfs/emea/holdings-daily-emea-en-spyy-gy.xlsx

Skip first 5 rows (metadata), row 6 is header
Columns: ISIN, SEDOL, Security Name, Currency, Number of Shares, Percent of Fund, Trade Country Name, Local Price, Sector Classification, Industry Classification, Base Market Value
~2,300 equities, 100% weight coverage, 49 countries
Filter equities: df[df['ISIN'].str.match(r'^[A-Z]{2}[A-Z0-9]{9}[0-9]$')]
Has duplicate ISINs (preferred + common shares) — aggregate with groupby('ISIN') before using as lookup
No ticker column — only ISIN, SEDOL, Security Name

Do NOT use iShares SSAC — its CSV omits ISINs (only Ticker, Name, Location). SSAC is still useful for iShares-vs-iShares comparisons (e.g. ESG screening analysis) since Ticker|Location is consistent within iShares.

TKF model portfolio ETFs

ETF	Source	Has ISINs?	Notes
Xtrackers USA/Canada	Downloaded XLSX	Yes	Skip 3 rows, column C is ISIN
Amundi USA	Downloaded XLSX	Yes	`engine='calamine'`, skip 20 rows, column B is ISIN
BNP Europe/Japan/Pacific	Downloaded XLSX	Yes	`engine='calamine'`, skip 22 rows, filter `Asset_Class == 'Equity'`
Invesco EM	companiesmarketcap.com	Yes	Web scrape, ~93% coverage (missing ~7% inflates active share 3-5pp)
iShares SAWD	iShares CSV	No	Has Ticker+Location only. Need name→ISIN bridge via SPDR
Vanguard ESG NA	Downloaded XLSX	No	Has Ticker+Name only. Need name→ISIN bridge via SPDR

Proposed equity model portfolio ETFs

ETF	ISIN	Source	Has ISINs?	Notes
Amundi MSCI USA Screened	IE000F60HVH9	Amundi XLSX	Yes	openpyxl fails on stylesheet — parse via `zipfile` + raw XML (see below)
Xtrackers MSCI USA Screened	IE00BJZ2DC62	DWS XLSX	Yes	Skip 3 rows, header row 4, column C is ISIN, column K is Weighting (0–1 scale, multiply by 100)
iShares MSCI USA Screened (SASU)	IE00BFNM3G45	iShares CSV	No	Standard iShares format. Bridge to ISIN via SPDR name matching
Amundi MSCI World Ex USA Screened	FR0013209921	Amundi XLSX	Yes	Same XML parsing as Amundi USA
iShares MSCI EM IMI Screened (SAEM)	IE00BFNM3P36	iShares CSV	No	~2,650 stocks (IMI includes small-caps). Many won't match ACWI — expected

TKF100 "Alt B" model portfolio ETFs (locked 2026-07-13)

The chosen redesign: 4 MSCI World Screened + 1 Vanguard global all-cap + Invesco EM. Weights: World funds 19.27% each, Vanguard 12.00%, Invesco EM 10.92% (EM pinned to 12.18% look-through). Benchmark = MSCI ACWI, plus ACWI IMI (SPDR SPYI) for the all-cap Vanguard sleeve.

ETF	ISIN	Source	Has ISINs?	Notes
iShares MSCI World Screened (SAWD)	IE00BFNM3J75	iShares CSV	No	Standard iShares format. Name→ISIN bridge via SPDR
Amundi MSCI World Screened	IE000QWCYQT0	Amundi XLSX	Yes	`parse_amundi_holdings()` raw-XML method (broken stylesheet). Launched Oct 2025
Xtrackers MSCI World Screened 1C	IE000I9HGDZ3	DWS XLSX	Yes	Skip 3 rows, header row 4, column C is ISIN, column K is Weighting (×100). German (`de-de`) file parses fine — read by column position
BNP Easy MSCI World (ESG Filtered) Min TE	IE000W8HP9L8	BNP XLSX	Yes	`engine='calamine'`, skip 22 rows, filter `Asset_Class == 'Equity'`. Sampled — some large names missing (genuine underweights)
Vanguard ESG Global All Cap	IE00BNG8L278	Vanguard XLSX	No	Ticker+Name only → bridge via SPDR. All-cap (incl. small + EM) → small-caps miss standard ACWI; score vs ACWI IMI. ~10.5% EM internally
Invesco MSCI EM Universal Screened	IE00BMDBMY19	Invesco XLSX	Yes	Prefer provider XLSX over companiesmarketcap scrape (scrape ~93% coverage inflates AS 3–5pp)

Resolved download URLs (these portals geoblock / JS-render — exact locale matters; the file itself downloads via an in-page button, not a static link):

iShares (SAWD, prod 305419): https://www.ishares.com/uk/individual/en/products/305419/ishares-msci-world-esg-screened-ucits-etf-usd-acc-fund → Holdings → Download CSV
Amundi: https://www.amundietf.nl/en/professional/products/equity/amundi-msci-world-screened-ucits-etf-acc/ie000qwcyqt0 → Documents
Xtrackers/DWS: https://etf.dws.com/en-no/IE000I9HGDZ3-msci-world-screened-ucits-etf-1c/ → Downloads → Index constituents. (en-no = English; en-ie/en-gb → 404, en-lu → 403, de-de → German but parses fine)
BNP: https://www.bnpparibas-am.com/en-lu/fundsheet/equity/bnp-paribas-easy-msci-world-min-te-ucits-etf-c-ie000w8hp9l8/?tab=documents → Documents (holdings are under Documents here, not Portfolio data; use en-lu, NO investor-type segment, short slug; fallback slug inserts -esg-filtered)
Vanguard: https://www.vanguard.co.uk/professional/product/etf/equity/9470/esg-global-all-cap-ucits-etf-usd-accumulating → Portfolio → holdings
Invesco EM: https://www.invesco.com/lu/en/financial-products/etfs/invesco-msci-emerging-markets-universal-screened-ucits-etf-acc.html → Holdings XLSX
Benchmark ACWI IMI (SPDR SPYI): same SSGA path as SPYY, swap spyy→spyi in the filename

Parsing Amundi XLSX files

Amundi XLSX files have broken XML stylesheets that crash openpyxl. Parse via raw XML extraction:

import zipfile, re, html

def parse_amundi_holdings(fpath):
    """Parse Amundi fund holdings XLSX via raw XML (bypasses broken stylesheet)."""
    with zipfile.ZipFile(fpath) as z:
        # Read shared strings
        ss = []
        if 'xl/sharedStrings.xml' in z.namelist():
            for si in re.findall(r'<si>(.*?)</si>',
                    z.read('xl/sharedStrings.xml').decode('utf-8'), re.DOTALL):
                ss.append(''.join(re.findall(r'<t[^>]*>([^<]*)</t>', si)))
        # Read sheet rows
        sheet_xml = z.read('xl/worksheets/sheet1.xml').decode('utf-8')
        rows = []
        for rxml in re.findall(r'<row[^>]*>(.*?)</row>', sheet_xml, re.DOTALL):
            d = {}
            for ref, attrs, val in re.findall(
                    r'<c\s+r="([^"]+)"([^>]*)><v>([^<]*)</v></c>', rxml):
                col = re.match(r'([A-Z]+)', ref).group(1)
                if 't="s"' in attrs:
                    d[col] = ss[int(val)] if int(val) < len(ss) else ''
                else:
                    try: d[col] = float(val)
                    except: d[col] = val
            if d: rows.append(d)
    # Find header row, parse holdings
    hi = next(i for i, r in enumerate(rows) if 'ISIN code' in r.values())
    # Columns: B=ISIN, C=Name, D=Asset class, E=Currency, F=Weight (0-1), G=Sector, H=Country
    holdings = {}
    for r in rows[hi+1:]:
        isin = html.unescape(str(r.get('B', '')))
        if len(isin) == 12 and r.get('D', '') == 'EQUITY':
            holdings[isin] = float(r.get('F', 0)) * 100  # Convert 0.09 → 9%
    return holdings

Prebuilt mapping files

Two mapping files in reports/adhoc/data/ speed up cross-provider matching:

isin_to_stockid.json — 1,698 ISIN → Ticker|Location mappings. Built from isin_ticker_map.json + matched Amundi/Xtrackers holdings. Use when bridging ISIN-based funds to Ticker|Location-based benchmarks (e.g. legacy SSAC).
provider_name_mappings.json — 705 fuzzy name mappings per provider (Amundi USA, Xtrackers USA, Amundi World Ex USA). Maps provider-specific names to SSAC stock_id. Use as fallback when ISIN matching isn't available.

For ISIN-based active share (preferred), neither file is needed — join directly on ISIN against SPDR ACWI.

Building a name→ISIN bridge for iShares/Vanguard

For the 2 ETFs without ISINs, use SPDR ACWI as a bridge:

Load SPDR (has ISIN + Security Name + Trade Country Name)
Build lookup dicts: {normalized_name: ISIN} and {(name_prefix, country): ISIN}
For each iShares/Vanguard stock, normalize its name and look up the ISIN
Use two-pass matching: first with share class info preserved (keep_class=True to distinguish Class A/C), then without
Fall back to prefix+country matching (15→10→8→6→4 chars) for abbreviated names

Bridging results (as of Mar 2026): ~1800 of 2600 ticker-only stocks bridged, 1.26% weight unmatched (mostly Vanguard small caps genuinely not in ACWI).

Matching rules

Always join on ISIN first — this matches 95-100% of holdings for ISIN-bearing ETFs
For residual unmatched: use normalized name matching with class-aware two-pass bridge
Never match on ticker alone across providers — same ticker can be different companies (e.g. CFR is Cullen/Frost in US, Richemont in Switzerland; ALV is Autoliv in US, Allianz in Germany)
Preferred vs common shares: for active share purposes, treat as the same stock (e.g. Samsung Electronics KR7005931001 pref = KR7005930003 common)
Multi-class shares (Class A/B/C): preserve class info when bridging to avoid collapsing different ISINs (e.g. Alphabet Class A US02079K3059 ≠ Class C US02079K1079)

Common pitfalls

OpenFIGI returns wrong exchange tickers: Maps ISINs to local exchange tickers that don't match iShares conventions. Avoid using OpenFIGI for cross-provider matching.
Name normalization must handle cross-provider differences: iShares uses "INC", SPDR uses "Incorporated"; iShares "AMAZON COM INC", SPDR "Amazon.com Inc."; iShares "MCDONALDS CORP", SPDR "McDonald's Corporation". Strip dots, apostrophes, hyphens early.
Abbreviated vs full names: "LVMH" vs "LVMH Moet Hennessy Louis Vuitton SE", "GSK PLC" vs "GLAXOSMITHKLINE" (rebrand). Short prefixes (4-6 chars) with country constraint catch abbreviations; rebrands need manual mapping.
"THE" prefix and "/The" suffix: Vanguard uses "TJX Cos Inc/The", SPDR uses "The TJX Companies Inc." — strip both leading and trailing "THE".
Single-letter sequences from dot stripping: "U.S. Bancorp" → "U S BANCORP" after dots removed. Collapse r'\b([A-Z]) ([A-Z])\b' → joined letters.
Country name differences: iShares uses "Korea (South)", SPDR uses "South Korea". Use prefix+country as a hint, not a hard requirement.
BNP Min TE funds use sampling: BNP doesn't fully replicate — some large stocks (e.g. Toyota Motor) may be missing. This creates genuine underweights, not matching bugs.
pensionikeskus.ee III pillar page does NOT include TKF — it's a separate fund category.
Validate matching quality with decomposition: Always check weight excluded/added/reduced. "Weight added" > 2% signals matching failures. Current notebook achieves 1.54% weight added.

Name normalization function

def normalize_name(name, keep_class=False):
    """Aggressively normalize stock names for cross-provider matching."""
    n = str(name).upper().strip()
    n = n.replace('.', ' ')    # "Inc." → "Inc "
    n = n.replace("'", '')     # "McDonald's" → "McDonalds"
    n = n.replace('-', ' ')    # "Freeport-McMoRan" → "Freeport McMoRan"
    n = n.replace('/', ' ')    # "TJX Cos Inc/The" → "TJX Cos Inc The"
    n = re.sub(r'^THE\s+', '', n)    # leading "The"
    n = re.sub(r'\s+THE\s*$', '', n)  # trailing "The"
    n = re.sub(r'\s*\([A-Z]+\)\s*', ' ', n)
    n = re.sub(r'\s+(KRW|TWD|INR|CNY|HKD|ZAR|SAR|BRL|MXN|THB|IDR|MYR|PHP|PLN|CLP|QAR|AED|KWD|COP|PEN|CZK|EGP|HUF|TRY|USD|NPV)\s*[\d.]*', '', n)
    # Strip voting designations (but optionally keep CLASS X)
    n = re.sub(r'\s+(?:SUBORDINATE VOTING|NON VOTING|PFD|PREF)\s*', ' ', n)
    if not keep_class:
        n = re.sub(r'\s+(?:CLASS [A-Z]\b|CL [A-Z]\b)\s*', ' ', n)
        n = re.sub(r'\s+CLA?\s*$', '', n)
    # Collapse single-letter sequences: "U S" → "US", "S P A" → "SPA"
    n = re.sub(r'\b([A-Z]) ([A-Z])\b', r'\1\2', n)
    n = re.sub(r'\b([A-Z]) ([A-Z])\b', r'\1\2', n)
    for _ in range(3):
        n = re.sub(r'\s+(AKTIENGESELLSCHAFT|INCORPORATED|CORPORATION|CORP|COMPANIES|COMPANY|COS|LIMITED|LTD|INC|PLC|CO|SA|AG|NV|SE|SPA|AB|ASA|OYJ|TBK|BHD|PCL|BERHAD|SH|ADS|REIT|DIVIDEND RIGHT CERT|PAR|N|SOCIETE ANONYME.*)\s*$', '', n)
    n = re.sub(r'\s+SPA\s*$', '', n)
    n = re.sub(r'\s+A/S\s*$', '', n)
    n = n.replace(' & ', ' AND ')
    n = re.sub(r'\s*(AND|&)\s*$', '', n)
    n = re.sub(r'\s+', ' ', n).strip()
    return n

Two-pass bridge pattern

# Build lookups from SPDR
spdr_name_class = {}   # WITH class → ISIN (precise)
spdr_name_to_isin = {} # WITHOUT class → ISIN (fallback)
spdr_prefix_country = {}  # (prefix, country) → ISIN
for _, r in spdr_acwi.iterrows():
    norm_cls = normalize_name(r['Security Name'], keep_class=True)
    norm = normalize_name(r['Security Name'], keep_class=False)
    country = r['Trade Country Name']
    spdr_name_class[norm_cls] = r['ISIN']
    if norm not in spdr_name_to_isin:
        spdr_name_to_isin[norm] = r['ISIN']
    for plen in [15, 10, 8, 6, 4]:
        key = (norm[:plen], country)
        if len(norm) >= plen and key not in spdr_prefix_country:
            spdr_prefix_country[key] = r['ISIN']

def bridge_to_isin(ticker, name, country):
    norm_cls = normalize_name(name, keep_class=True)
    if norm_cls in spdr_name_class:
        return spdr_name_class[norm_cls]
    norm = normalize_name(name, keep_class=False)
    if norm in spdr_name_to_isin:
        return spdr_name_to_isin[norm]
    if country:
        for plen in [15, 10, 8, 6, 4]:
            if len(norm) >= plen:
                key = (norm[:plen], country)
                if key in spdr_prefix_country:
                    return spdr_prefix_country[key]
    return None

Data refresh workflow for active share analysis

When running a new active share analysis, download fresh data on the same day to ensure dates match:

1. Download ACWI benchmark

Download SPDR MSCI ACWI (SPYY) daily holdings XLSX:

https://www.ssga.com/uk/en_gb/intermediary/etfs/library-content/products/fund-data/etfs/emea/holdings-daily-emea-en-spyy-gy.xlsx

Updated daily. Check row 4 for "Holdings As Of:" date.
Save to working directory (not committed to repo — large file, changes daily).

2. Download fund holdings (same day)

Fund	Where to download
Amundi USA Screened (IE000F60HVH9)	amundietf.com → product page → Documents → Fund Holdings XLSX
Xtrackers USA Screened (IE00BJZ2DC62)	etf.dws.com → product page → Documents → Constituent XLSX
iShares USA Screened (SASU, IE00BFNM3G45)	ishares.com → product page → Holdings → Download CSV
Amundi World Ex USA Screened (FR0013209921)	amundietf.com → product page → Documents → Fund Holdings XLSX
iShares EM IMI Screened (SAEM, IE00BFNM3P36)	ishares.com → product page → Holdings → Download CSV

For the TKF100 "Alt B" model portfolio funds (iShares/Amundi/Xtrackers/BNP World Screened + Vanguard ESG Global All Cap + Invesco EM), use the resolved download URLs in the Alt B data-source table above — plus the ACWI IMI benchmark (SPDR SPYI) for the all-cap Vanguard sleeve. Download all on the same day as the benchmark.

3. Match and compute

Preferred: ISIN-based matching (no mapping files needed)

Load SPDR ACWI — pd.read_excel(..., skiprows=5), filter equities by ISIN regex, groupby('ISIN') to aggregate
Load Amundi XLSX — use parse_amundi_holdings() (see above) — returns {ISIN: weight}
Load Xtrackers XLSX — pd.read_excel(..., header=3), ISIN in column C, weight in column K (×100)
Load iShares CSVs — bridge names to ISINs via SPDR (see two-pass bridge above)
Build portfolio: for each fund, scale stock weights by allocation percentage, aggregate by ISIN
Compute: Active Share = ½ × Σ|w_portfolio[i] − w_benchmark[i]| over all ISINs

Fallback: name-based matching (when using SSAC benchmark)

Use provider_name_mappings.json for Amundi/Xtrackers → SSAC stock_id matching
Use isin_to_stockid.json to bridge remaining ISINs to Ticker|Location

4. Validate

Total matched weight should be >99% for ISIN-bearing funds (Amundi, Xtrackers)
iShares bridging: >98% for USA (SASU), ~40-50% for EM IMI (SAEM) — the rest are small-caps not in ACWI
Check benchmark date matches fund holdings dates (within 1-2 days)
"Weight added" (portfolio-only stocks) should be <0.5% — higher signals matching bugs

name	fetch-holdings
description	Use when fetching, loading, or matching ETF holdings data for Tuleva portfolio analysis. Covers data sources, ISIN matching, and common pitfalls.