Data Quality

What the system sees, where the gaps are, and what's been mitigated.

Universe overview

Companies

608

survivor universe

Start

2006-01-03

first source date

End

2015-12-31

last source date

Coverage timeseries

6696082006-01-312015-12-31
OHLCV countFundamentals countUniverse size

Survivorship-biased universe: all 608 companies are present every day in the source data.

Per-signal coverage

SignalWarmup daysFirst valid dateMean coverage
momentum2522007-01-04
100.0%
mean_reversion52006-01-10
100.0%
value02006-03-23
98.6%
growth2522007-01-04
96.0%
quality02006-03-23
98.6%

Universe composition

Exchange

NASDAQ608

Sector

Banks72
Semiconductors and Semiconductor Equipment52
Software36
Electronic Equipment, Instruments and Components30
Hotels, Restaurants and Leisure28
Health Care Equipment and Supplies27
Biotechnology24
Communications Equipment23
Specialty Retail20
Professional Services17
Health Care Providers and Services14
Commercial Services and Supplies14
Capital Markets13
Technology Hardware, Storage and Peripherals12
Insurance11
Ground Transportation11
Life Sciences Tools and Services11
Media9
Machinery9
Consumer Staples Distribution and Retail9
Oil, Gas and Consumable Fuels9
Food Products9
Diversified Telecommunication Services8
Entertainment7
Pharmaceuticals7
Household Durables7
IT Services6
Financial Services6
Energy Equipment and Services6
Diversified Consumer Services6
Electric Utilities6
Aerospace and Defense5
Interactive Media and Services5
Building Products5
Electrical Equipment4
Specialized REITs4
Beverages4
Metals and Mining4
Consumer Finance4
Chemicals4
Air Freight and Logistics4
Health Care Technology3
Hotel and Resort REITs3
Leisure Products3
Distributors3
Construction and Engineering3
Passenger Airlines3
Industrial Conglomerates3
Trading Companies and Distributors3
Automobile Components3
Household Products3
Textiles, Apparel and Luxury Goods3
Broadline Retail2
Personal Care Products2
Water Utilities2
Retail REITs2
Multi-Utilities1
Paper and Forest Products1
Real Estate Management and Development1
Health Care REITs1
Wireless Telecommunication Services1

Anomalies

Negative filing lags (fixed)

available_date = max(filing_date, period_end_date) ensures data is never accessible before quarter end.

CompanyFiling datePeriod endLag

Extreme daily returns (clipped)

returns clipped at +/-100% in run_backtest engine.

CompanyDateReturnCauseMitigation

Honest caveat

Survivorship bias

The source universe contains 608 surviving names. Known failed or acquired companies are absent, including:

Lehman Brothers

LEH · 2008-09-15

bankruptcy

Bear Stearns

BSC · 2008-03

acquired by JPM

Wachovia

WB · 2008-10

acquired by WFC

General Motors

GM (old) · 2009-06

bankruptcy

CIT Group

CIT · 2009-11

bankruptcy

Estimated return inflation

1-4% annualized

All reported returns are upper bounds; methodology is sound, data has a known limitation

Fundamentals freshness

Mean

44d

Median

39d

Max

620d

23520-301870330-60301760-907390-12034120-15023150-180118180+

Most fundamentals are filed within 90 days of period-end; we use available_date as the PIT key.

Reproduce: .venv/bin/python scripts/data_quality_report.py