Best Market Data Sources for Systematic Short Selling Research

A short-selling research stack needs prices, financing evidence and company context. Each answers a different question.

The data demands of short selling

A price history can show how a stock moved. It cannot establish whether shares were available to short, what the account would have paid to borrow them, or when a financing disclosure became available. Systematic short-selling research needs those constraints alongside the investment hypothesis.

Start by separating four inputs: market prices and quotes; reported or estimated short positioning; broker-specific locate and borrow information; and corporate disclosures or model outputs used to select candidates. Avoid treating similarly named fields as interchangeable.

Price data for short selling

Massive, formerly Polygon.io, documents stock bars, trades and quotes. Databento offers multiple normalized schemas, with available trade, quote and book detail depending on the dataset. Select the feed and entitlement required for the proposed test rather than assuming every subscription includes the same history or depth.

For small and micro-cap research, inspect bid-ask spreads, session coverage, halts, stale quotes, corporate-action handling and delisted symbols. Daily closing prices cannot establish an intraday fill. Compare your intended entry and exit conventions against the actual records delivered.

Short interest and securities lending

S3 Partners provides short-positioning, crowding and financing data products. Compare the methodology, distribution channel and required history for the particular product; a positioning measure is not a firm locate offer for your account.

ORTEX's live short-interest documentation describes estimates derived from lending information, alongside borrowed and returned shares and cost-to-borrow observations. Intraday transactions can change or be cancelled before end-of-day consolidation. Check whether the API or export you purchase contains the same timing and fields as the displayed platform product.

Fintel's short-interest pages distinguish intraday availability and borrow-fee fields from daily short-sale volume and less frequent reported short interest. Those are different measures: trading marked short is a flow, while short interest concerns outstanding positions. Record the source and observation date of each field.

Provider capabilities were reviewed October 28, 2026. Pricing, API access, export rights and historical coverage vary by product and agreement; obtain the terms for the exact workflow instead of relying on an old headline subscription price.

Broker-specific locate quote history

Alphanume's short-locate quote history adds a concrete feasibility input: what Alpaca quoted for hard-to-borrow stocks in a scheduled sweep. The series starts October 8, 2026, and has no earlier backfill. It does not compare brokers or represent all shares available in the market.

Each served row is keyed by trading date, am or pm slot, and ticker. Keep status, price_per_share, cost_per_100 and available_qty together. The dollar cost for 100 shares is a locate quote, not an annualized borrow rate. A quote is neither a reservation nor an executed order.

The scheduled sweeps are 11:45 and 17:30 UTC on trading days. snapshot_time_utc records the sweep-start minute, shared across the sweep, rather than each row's precise quote time or its public availability time. Scheduling is not a delivery guarantee. Save the response and your retrieval time before using it in a live decision.

no_shares has zero available quantity and can retain an indicative price. A price alone therefore cannot establish available shares. Null is not zero cost or free borrowing. The API excludes the pipeline's missing status, so absence from the response is not evidence that a ticker was available. The Python locate monitor works through these checks.

Premarket drop risk alongside financing constraints

The Premarket Drop Risk dataset supplies a separate candidate-selection input. Its prob_drop estimates the probability of an unadjusted regular-session open-to-close decline of at least 5%, for selected actively traded U.S. microcaps retained by the model's positive-class selection. It is not a forecast of the size of the return or a screen of every premarket mover.

Features use a 09:00 Eastern cutoff. That cutoff is different from publication and retrieval time. Rows before August 31, 2026 were generated retrospectively by walk-forward replay, rather than published live on those historical mornings. The public response does not include the internal publication timestamp or replay-source flag, so preserve your own retrieval evidence and label historical replay separately.

The model does not supply borrow availability, locate cost, spread or fill evidence. Join candidate and locate records on the intended date and ticker only after checking that each observation was available before the decision. Later return fields are outcomes for evaluation, rather than inputs to a morning screen. A high modeled probability cannot make an unavailable or uneconomic short executable.

Corporate events and capitalization context

SEC disclosures can support research into financing, dilution and distress. Alphanume's Dilution Events dataset structures filing records, while its corporate-default feed identifies default disclosures. Read the source filing and distinguish registration, effectiveness and subsequent issuance. A registration statement does not prove shares were sold or that the price will decline.

Historical market-cap and shares-outstanding data can help construct size-based groups. It is not universally point-in-time: the producer can carry a first filing's share count backward by up to 180 days. A closing-price-based value is also unavailable earlier in that session. Review the timing and share-count limitations before using it to reconstruct a historical micro-cap universe.

Match the research question to the input

Assembling a short-selling data stack

Define the candidate universe first, then record the information available at the decision time. Add broker-specific feasibility evidence and explicit cost and execution assumptions. Report how many candidates lacked a usable quote, how many passed the proposed constraints, and which costs were observed versus modeled.

Alphanume's Short Locates and Premarket Drop Risk API routes require Pro access and an API key; public landing pages describe the products. Review API documentation and current access terms for evaluation. For account selection, compare short-selling broker tools; for cost modeling, distinguish locates from daily stock-borrow fees.