How do I tell whether Taiwan market data is missing something?

Two kinds of missing, handled differently

• The market was closed: weekends, public holidays, typhoon days. Not a gap — your series should not have those dates.
• The data did not arrive: the market traded and you have no row. That is a gap, and left unhandled it quietly distorts calculations.
• Conflating the two usually ends in forward-filling, which invents a flat day that never happened.

Prices: use meta.market_status to separate closed from missing

The price response's meta carries last_trading_day and market_status, a list of date and status pairs where status is open or unknown. Open means the market traded; unknown means not yet determined. Compare that against the dates you received and the distinction resolves itself.

import requests

r = requests.get(
    "https://api.twmarketdata.com/v2/datasets/twse-daily-price",
    headers={"X-API-Key": "sk_live_your_key"},
    params={"symbol": "2330", "start_date": "2026-06-01", "end_date": "2026-07-31"},
    timeout=20,
)
r.raise_for_status()
payload = r.json()

print("rows", payload["count"], "data as of", payload["data_as_of"])
print("last trading day", payload["meta"]["last_trading_day"])

have = {row["date"] for row in payload["rows"]}
for entry in payload["meta"]["market_status"]:
    if entry["status"] == "open" and entry["date"] not in have:
        print("market traded but no row:", entry["date"])

Every price row explains how it was produced

Each row carries price_method and price_confidence. When a value looks odd or a stretch is uncertain, read those two before deciding whether to use it — they tell you more about trustworthiness than the close itself does.

for row in payload["rows"][:5]:    print(row["date"], row["close"], "method", row["price_method"], "confidence", row["price_confidence"])

Some datasets carry data_gaps, and it is per row

Margin and short balance data, for instance, carries a data_gaps field on every row. In practice it is usually an empty list, meaning no known gap for that row; when populated it describes that row's gaps. Note that it is per row, not one summary for the response.

m = requests.get(    "https://api.twmarketdata.com/v2/datasets/margin-short",    headers={"X-API-Key": "sk_live_your_key"},    params={"symbol": "2330", "limit": 20},    timeout=20,).json() for row in m["rows"]:    if row["data_gaps"]:        print(row["trade_date"], "gap markers:", row["data_gaps"])

Elsewhere data_gaps describes the dataset, not a day

On some datasets the same name holds named coverage limitations rather than dates — statements about what the dataset never covered, such as being limited to the listed market. That is not telling you a day is missing; it is telling you a scope boundary exists, which is worth reading before any cross-market analysis.

There is no single site-wide gaps field

Worth saying plainly: data_gaps is not present on every endpoint. The price endpoint has none, and offers market_status, count, data_as_of and per-row price_method and price_confidence instead. So the right move is not to hunt for data_gaps, but to check which signal the endpoint you are calling actually provides.

Whatever the signal, do not patch the hole yourself

• Do not forward-fill. It manufactures a flat session, understating volatility and hiding drawdown.
• Do not substitute zero. Zero is meaningful in prices and in net flow, so afterwards you cannot distinguish 'was zero' from 'was absent'.
• Leave the gap and skip it explicitly in calculations. A series repaired to look complete will flatter a backtest — the same problem as look-ahead, one layer down.

What we do not have

• No uniform data_gaps field across all datasets, as above.
• No real-time quotes, so an intraday missing tick is out of scope here — the day's data exists after the close.
• In market_status, unknown means undetermined, not closed. Do not treat it as a holiday.

Common questions