ticker-reference-data

US ticker reference data, rebuilt automatically every day.

Datasets built from public exchange notices and regulatory filings, plus our own corrections (splits the sources missed, ticker renames resolved by CIK, halt reconciliation across feeds):

File	What it is	Size
data/fresh_universe.csv	"fresh" tickers: IPO or reverse split within the last 365 days	~95 KB
data/ticker_changes.json	ticker renames resolved by CIK	~1.1 MB
history/halts/	trading halts, one CSV per year, 2019→today	~9.8 MB

Files under data/ are a snapshot rewritten daily; those under history/ are a cumulative series that only grows.

Consuming it

Every day a release tagged with the snapshot's as_of is published. Assets on a tag are immutable, so the URL identifies the content unambiguously:

# A specific date's snapshot. 200 = that exact day, guaranteed. 404 = not published
# yet (and your job knows to keep using its previous copy).
curl -sSfL https://github.com/Quant-Lodge/ticker-reference-data/releases/download/2026-07-24/fresh_universe.csv -o fresh_universe.csv

# Whatever was published last.
curl -sSfL https://github.com/Quant-Lodge/ticker-reference-data/releases/latest/download/fresh_universe.csv -o fresh_universe.csv

Halts live in their own tag namespace, halts-YYYY-MM-DD, because the two pipelines run at different hours and each dataset derives its tag from its own data:

curl -sSfL https://github.com/Quant-Lodge/ticker-reference-data/releases/download/halts-2026-08-07/halts_2026.csv -o halts_2026.csv

That release is not marked as latest: that pointer belongs to fresh_universe.csv, which is what the bots consume. For halts, use the dated tag.

Prefer releases for automated consumption. The files under data/ are also there, but they are served from raw.githubusercontent.com, which points at a mutable branch and therefore caches (max-age=300): right after a push you can get the previous version. Measured serving content 218 s old. Fine for eyeballing in a browser; not fine for a job that needs today's data.

https://raw.githubusercontent.com/Quant-Lodge/ticker-reference-data/main/data/fresh_universe.csv
https://raw.githubusercontent.com/Quant-Lodge/ticker-reference-data/main/data/ticker_changes.json

fresh_universe.csv

A ticker qualifies as fresh if rs_days ≤ 365 or ipo_days_eff ≤ 365.

ticker,rs_days,ipo_days_eff,fresh_via,as_of_date
AAAC,,225,ipo,2026-07-24
AAHYX,228,,rs,2026-07-24
ABTC,18,324,both,2026-07-24

Column	Type	Description
ticker	string	symbol
rs_days	int | empty	days since the most recent reverse split; empty if none
ipo_days_eff	int | empty	days since the effective IPO; empty if it couldn't be determined
fresh_via	ipo | rs | both	which condition qualifies it
as_of_date	YYYY-MM-DD	snapshot date, identical on every row

One column may exceed 365 days when the other one is what qualifies: ABTC gets in on its 18-day-old reverse split even though its IPO was 324 days ago.

Typical cut: ~3,800 tickers (~2,700 by IPO, ~1,000 by reverse split, ~110 by both).

The universe is not filtered by price, volume or tradability — it includes OTC, funds and units. Filter for your own use case.

ipo_days_eff is effective, not the literal listing date: it accounts for the start of continuous trading, so a ticker recycled after a delisting doesn't show up carrying the age of the previous issuer.

ticker_changes.json

A ticker → rename resolution map, using the SEC CIK as stable identity.

{
  "FTFT": { "cik": "0001066923", "old_ticker": "SPU" },
  "MBOT": { "cik": "0000883975", "old_ticker": "STEMD" }
}

Field	Description
cik	SEC CIK (zero-padded string); null if unresolved
old_ticker	previous symbol; null or absent if there was no rename
flat_file_only	true = the symbol shows up in the flat files but its identity wasn't resolved

It is a lookup cache: it also records negative results so the query isn't repeated. Of ~15,300 entries, ~1,150 carry a usable rename; the rest are lookups that came back empty. An entry can be null (queried, no answer). When consuming it, keep the ones with a non-empty old_ticker.

Typical use case: on day 1 after a rename, the previous close lives under the old symbol.

history/halts/

US equity trading halts, one CSV per year. 69,256 events, 2019→today.

A halt isn't published by the exchange: it is disseminated over the SIP as an administrative message, and every venue receives the ones from all tapes. That's why NYSE, Nasdaq and Cboe each carry the full universe, and here they are reconciled: one event is one row, and the fuentes column says which ones saw it.

symbol,symbol_raw,halt_date,halt_time,halt_ms,reason_code,reason_raw,name,listing_market,resume_date,resume_quote_time,resume_trade_time,duracion_seg,pause_threshold_px,px_antes,px_resume,gap_pct,fuentes,ingested_at
BKSY.WS,BKSY WS,2026-08-07,16:05:26.515,515,H11,Regulatory Concern,BlackSky…,NYSE,,,,,,,,,cboe|nasdaq_rss|nyse,2026-08-07T22:29:11

Column	Description
symbol	canonical symbol (share class separated by a dot: BKSY.WS)
symbol_raw	as sent by the winning source, for auditing
halt_date / halt_time	halt start, New York time; milliseconds only when Nasdaq saw it
reason_code	canonical: LUDP (LULD), T1/T2 (news), T12, H10 (SEC suspension), UNKNOWN
resume_quote_time	start of the quote-only period — this is where the reopening price forms
resume_trade_time	actual resumption; empty = never resumed, the close is not imputed
duracion_seg	halt duration in seconds
px_antes / px_resume / gap_pct	price before and after, and how far it moved while untradable
fuentes	which feeds saw the event, |-separated

Column names are kept in Spanish where they already ship that way — renaming them would break every consumer downstream.

File	Events
2019.csv	1,212
2020.csv	14,895
2021.csv	5,595
2022.csv	6,834
2023.csv	9,175
2024.csv	10,529
2025.csv	12,370
2026.csv	8,646

One name can be halted many times in a single day — the max measured in the series is 60. Event identity is (symbol, halt_date, halt_time) to the second; deduplicating by (symbol, date) would delete 59 of those 60, precisely on the parabolic days.

An old year can change any day. The feeds return halts that are still OPEN, and some have been open for years (SVA has been halted since 2023-06-13): every row is routed to the file for its year, not the run date.

Hard floor at 2019-02-22: that's as far back as the NYSE history goes, the only source with range backfill. And be careful crossing May 2012 in any statistic: LULD didn't exist before that, so you'd be mixing regimes.

name is the current name, not the one at halt time. For renamed securities the symbol isn't necessarily the one that traded that day either.

Update schedule

data/ is rebuilt daily around 02:15 ET. fresh_universe.csv is rewritten in full; ticker_changes.json grows incrementally.

history/halts/ is updated Monday to Friday at 22:00 ET, once the day has closed and the three sources have converged (news halts are published well into the afternoon and many resume the following day).

Commits only happen when something actually changed — the last commit date is the date of the last real change, not of the last run.

as_of_date is the source of truth for snapshot freshness. If your job depends on having today's data, validate that column instead of trusting the file mtime.

Sources

Built from public exchange notices (NYSE, Nasdaq, Cboe) and regulatory filings, reconciled against each other. Published as a convenience for our own consumption, with no guarantee of accuracy, completeness or availability. Verify against the primary source before using it for any decision.