Data methods & trust
How the data is built, where it is thin, and what we deliberately do not serve.
13 pages
- MethodologyHow official sources become datasets: parsing, joins, adjustment and point-in-time rules.
- Data qualityCoverage depth per dataset, which sources publish irregularly, and what we withhold and why.
- GlossaryThe terms you need to read Taiwan market data correctly, each with the mistake it prevents.
- knowledge_dateThe earliest date a figure could have been known in the real world — the day it was actually announced, not the period it describes.
- point-in-time (as-of query)Query a past date and see only what was known on that date — not today's revised values.
- Adjusted pricePrices with the gaps from dividends, rights issues and splits stitched back in, so figures from different dates are comparable.
- Filling the dividend gapThe price climbing back to where it was before the ex-dividend drop.
- The three institutional investorsForeign investors, investment trusts and dealers — the three institution types Taiwan publishes net buy/sell for after each close.
- Margin buying and short sellingRetail leverage: margin is borrowing money to buy, short selling is borrowing stock to sell.
- Securities lending vs. lent-and-soldInstitutions borrowing stock. Borrowing it is not the same as selling it — only the sold part is short pressure.
- Open interest (OI)Contracts still open in the futures or options market — a stock, not a flow like volume.
- Survivorship biasLooking only at companies that are still listed, which systematically overstates historical returns.
- data_gapsMissing data is marked as missing — never interpolated or forward-filled.