US Equities Trade Only Adjusted Minute Bar Excluding FINRA/TRF Trades

How to use it

Times are local US Eastern clock times (DST-aware), not UTC. Treat TRADE NB / NBBO-flagged rows as the backtest-grade subset. Cancellations reference earlier trades and can arrive hours later, so aggregate only after the session closes. Prices are unadjusted unless the dataset name says Adjusted; join us-equities-*-adjustment-factors on SecId for adjustment.

Typical use

Backtesting, execution and microstructure research, feature engineering for US equities.

Limitations

Delivered as recorded from the source feed: AlgoSeek does not remove crossed or locked quotes, bad prints, or out-of-sequence packets unless contracted to. Consumers own cleaning. Price columns are NOT raw: divide by PriceAdjFactor to approximate as-traded prices (inference; the factor's convention is undocumented). Cross-dataset check: the industry-standard adjusted sample for the same AAPL day shows an adjusted/raw ratio of about 0.998651, the same value as PriceAdjFactor here. Whether Volume is raw or adjusted cannot be determined: VolumeAdjFactor is 1 for every sample row, and Volume equals the unadjusted bar's volume. For reference, AlgoSeek's Daily Cumulative Adjustment Factors guide multiplies price by its cumulative price factor and divides volume by its cumulative volume factor; whether this table follows that is unconfirmed. The adjustment is a single factor for the whole day (0.998651458110918 for all 100 rows), consistent with a per-trade-date cumulative backward factor. Restatement behavior (rebuild on new corporate events) is documented for the sibling adjusted datasets, not for this one.

Coverage

US exchange-listed equities (common and preferred stock, ETFs, ETNs, ADRs, warrants, units) from the SIP, with FINRA/TRF reports excluded per the page's first paragraph and the sample's volumes. The page's second paragraph says the data includes off-exchange TRF trades (see discrepancies). History length is not stated.

Answers to common questions

What time zone are the timestamps in?

US Eastern local time with daylight saving (the guides label it EST/ET).

Are prices adjusted for splits and dividends?

Only in datasets whose name contains Adjusted; all others are as traded.

Usage constraints

• Licensed vendor data: never export raw rows to a tenant-local destination or include them verbatim in a public artifact; serve only through entitlement-checked platform reads.
• Do not treat the vendor's `Ticker` as a stable identity across time; join on an AlgoSeek security key (SecId or ASID, which are distinct) or the platform symbol mapping.

Known issues (8)

• Primary key is inferred; AlgoSeek's guide does not name one.
• Documentation discrepancySchema mismatch with the linked guide: the page links the 10-column base Minute Bar guide, but the sample has 13 columns. TradeDate (yyyy-mm-dd) and BarDateTime (timestamp) replace the guide's Date (yyyymmdd) and TimeBarStart (HH:MM); ASID, PriceAdjFactor and VolumeAdjFactor appear in no supplied guide. Date and TimeBarStart were mapped to those columns rather than added as guide…
• Documentation discrepancyPage says the data 'follows the same format as' the Trade Only Adjusted Minute Bar (17 columns: SecId, raw prices, *Adjusted columns), yet the sample has no SecId and no *Adjusted columns; it has one price set plus two factor columns.
• Documentation discrepancyPage says the dataset delivers raw OHLC/volume plus adjusted OHLC/volume, but the sample only has one set of price columns and they are already adjusted: for all 100 rows, price = unadjusted ex-FINRA sample price x PriceAdjFactor (max abs diff 5e-5, 4-decimal rounding).
• Documentation discrepancyPage description says both that off-exchange TRF trades are excluded and that the data includes 'off-exchange trades reported to TRF' (second paragraph looks copied from the standard adjusted page). The sample's Volume/TotalTrades match the ex-FINRA unadjusted bars (663,233 / 1,075 at 09:30), not the standard bars (848,436 / 7,183).
• Documentation discrepancyScraped bundle fields page_category and page_name hold description sentences (page has an extra description paragraph that shifted the scrape); the dataset name here is taken from the page heading.
• Documentation discrepancyLinked guide's VolumeWeightPrice formula has the same spurious NBBOMidpoint term; see the base Minute Bar specs.
• Documentation discrepancySample comes from the ArdaDB database_table channel, unlike the other six datasets in this batch whose samples are S3 cloud_storage CSVs.