Markets do not wait for the earnings press release to start repricing. In the days before a report, news flow accelerates: supply chain whispers, analyst notes, partner disclosures. Prices often drift in the direction of the eventual surprise. Understanding that drift helps you contextualize positioning, not front-run insider information.
Drift studies are easy to botch. Look-ahead bias, vague event windows, and unadjusted benchmarks produce charts that impress in slides and fail replication. The framework below is intentionally modest: reproducible windows, explicit data sources, and interpretation guardrails.
Using Vantafin news via GET /news, the N WebSocket channel, get_news in the MCP server, and earnings dates on the calendar, we outline a reproducible pre-earnings drift study you can run on your own coverage list. Vantafin does not provide options flow data.
Define the event window
For each earnings date on the calendar, mark T-10 through T-1 trading days. Exclude the announcement day from the pre-window so you do not mix pre-release drift with post-release gap. Count news items per day from GET /news or get_news as a coarse measure of attention; volume z-scores from GET /quotes/{ticker} flag unusual trading without requiring derivatives data.
Abnormal return versus a sector ETF or custom basket from screener exports keeps comparisons honest. A semiconductor name should not be judged only against SPY if its peers moved together on the same supply chain headline.
Pull price history with get_price_history via the historical OHLCV guide and news via the REST API. Store event metadata: report date, fiscal quarter, surprise outcome when known, so you can segment results later.
Interpret carefully
Drift is common; alpha from drift is not guaranteed. Separate routine product rumors from material claims that would trigger 8-K disclosure. If the news is immaterial, drift may reflect positioning games rather than information leakage.
Vantafin does not offer options flow. Focus on price drift, volume, and news count instead. A steady drift higher into earnings with rising news count and volume deserves a different preview note than flat price with scattered headlines.
Add names to watchlists and configure email alerts for filing and news events, not just price. News alerts prompt you to read the article before the move becomes irreversible.
Report hit rates with confidence intervals if you publish internally. A 58% directional hit rate across forty events is interesting; one memorable anecdote is not research.
Scale the study
Run the panel across quarters; store accession IDs for news items you want to audit. Replication starts with archived sources, not screenshots. The MCP server helps prototype event windows in Cursor ("pull news for MSFT T-10 to T-1 for the last eight earnings dates" via get_news) before you codify the pipeline in Python.
Limits: pricing. Drift research is most valuable when shared with compliance-aware language: contextualize positioning, do not claim illegal edge. Used that way, it improves earnings preview notes and post-mortems alike.
Measuring setup before the print
Stocks often trend into earnings as positioning builds. Measuring pre-earnings drift (return and volume versus a baseline window) helps separate momentum positioning from fundamental repricing. Mark earnings dates from the calendar, define a window such as T-10 through T-1 trading days excluding the reaction day, and pull marks with bulk quote snapshots at consistent clock times so your windows stay comparable quarter to quarter.
Compare drift to sector peers from a screener export so a rising tide does not look like single-name alpha. Normalize by beta or sector return over the same window, and document which normalization you used so results compare across quarters. Pull headline counts and article text via REST news routes and get_news in the MCP server; subscribe to N.{ticker} when you want live headlines during the drift window.
Cross-check drift direction against recent management guidance revisions. Drift with downward guidance can signal denial rather than strength. That tension is exactly what pre-print risk meetings should surface before the desk sizes into a binary event.
Watchlists, alerts, and post-earnings follow-through
Score every watchlist name reporting this week; sort by drift magnitude for desk prioritization when twelve prints land in three days. Configure email news and price alerts when sentiment or price moves exceed thresholds three days before the event so researchers review filings before the crowd moves.
Store drift and actual gap direction to study whether pre-print momentum predicted post-print continuation or reversal. get_price_history over MCP supplies split-adjusted history for event studies without maintaining a separate vendor file. Fire alerts when drift exceeds a sigma threshold you define, but report hit rates with confidence intervals if you publish internally) one memorable anecdote is not research.
Ask the MCP server: "Which holdings reporting Friday show the largest positive drift this week?" for a Thursday risk huddle. Cited answers beat manually sorted spreadsheets when the calendar is crowded.
Integrating drift into broader process
Drift screens pair naturally with batch quotes at fixed clock times and with email filing alerts when 8-K items arrive late in the drift window. A filing can invalidate momentum that looked clean on price alone. Combine drift with screener volume z-scores rather than importing datasets outside your equity stack.
Confirm entitlements on pricing before automating drift scores across a full index universe each earnings season. Cache baseline volatility estimates locally rather than recomputing from scratch nightly unless your universe changes frequently.