Transparency on how QuantMedia sources, processes, and presents market data, research metrics, and financial news. This page documents our data pipeline, analytical methodology, editorial standards, and platform disclosures.
QuantMedia aggregates financial data from multiple third-party providers. No proprietary trading data is generated or stored. All data is sourced via publicly available APIs under their respective free-tier terms of service.
| Provider | Data Type | Coverage | Latency / Notes |
|---|---|---|---|
Yahoo Finance (via yfinance) | End-of-day OHLCV bars; index and futures closes | 180-name signal universe; S&P 500, Nasdaq, Dow, VIX, gold, WTI, DXY | End-of-day only. Captured once per weekday after the US close. |
| Finnhub | News headlines | General market newswire | Server-side only, once per weekday. The API key is held as a repository secret and is never sent to the browser. |
| CoinGecko | Cryptocurrency prices | BTC, ETH, SOL, and major altcoins | Captured once per weekday by the pipeline; not streamed to the browser |
| Open Exchange Rates | Foreign exchange rates | USD crosses (EUR, GBP, JPY, etc.) | Indicative mid-market rates; updated periodically |
| TradingView | Embedded ticker tape widget | Major indices, equities, FX, commodities | Real-time via TradingView embed |
With one exception, nothing on this site is real-time. Equity, forex, crypto and commodity figures are end-of-day snapshots captured once per weekday by an automated pipeline that runs at 23:30 UTC, after the 16:00 ET US close. They are stored as static JSON files and served from this domain; the browser makes no live market-data calls. The single exception is the embedded TradingView widget, which streams its own data independently.
Every page derives its freshness label from the timestamp of the snapshot it is displaying, so a stalled pipeline shows “delayed” or “data feed paused” with the real capture date rather than continuing to claim it is current. If a data source fails, the previous snapshot is retained with its original timestamp instead of being replaced by blank cells. None of this data is suitable for live trading decisions.
Headlines on the News page are retrieved server-side once per weekday from the Finnhub general news endpoint, de-duplicated, and classified into categories using word-boundary regular expressions (an earlier version matched bare substrings, which mis-filed any headline containing the letters “ai” — gain, against, remain — as technology news). Categories are:
Classification is purely algorithmic. No editorial judgment or human curation is applied to news categorization. Headlines are displayed as received from source feeds with attribution to the originating publisher.
Runnable implementations of the VPIN and HRP research, with example data and tests, are indexed at research reproducibility.
Two proprietary metrics are derived from this same scan and published daily: QuantMedia Signal Breadth and QuantMedia Sector Confluence.
The daily US stock signals dashboard is produced by a single deterministic script, scripts/daily_update.py, which runs after each US close. Its full behaviour is summarised here so the figures quoted across the site can be checked against one definition.
| Parameter | Value | Detail |
|---|---|---|
| Universe | 180 US-listed equities | A fixed liquidity list, not an index. Includes large-cap S&P 500 names alongside high-volume tickers that are not index constituents. |
| Eligibility | ≥ 60 sessions | Completed daily OHLCV bars required within the trailing 1-year window. Names failing this are skipped for the day rather than scored on partial history. |
| Signals | 30 binary checks | 10 trend/moving-average, 4 momentum oscillator, 3 Bollinger, 3 volume, 4 rate-of-change, 2 52-week position, 2 stochastic, 2 volatility-regime. |
| Decision rule | score ≥ 22 → BUY | At least 73% agreement. Everything below the threshold is WATCH and is not displayed. |
| Reference levels | ATR(14) multiples | Stop = close − 2.0 × ATR; target = close + 1.5 × ATR. Volatility-scaled, not fixed percentages. |
| Schedule | Mon–Fri, 23:30 UTC | One run per weekday. No intraday updates, no weekend runs. |
The threshold is not optimised against historical returns, so it is not fitted to any particular market period — and equally carries no claim of being an optimal cut-off. The engine uses no fundamental data, no machine learning and no discretionary override. QuantMedia publishes no track record for it: there is no live or audited performance history, no Sharpe ratio and no trade log, because no verified out-of-sample record exists to report. Requiring broad multi-factor agreement also biases output toward trend continuation, which means signals thin out in choppy or falling markets by construction.
The research section presents quantitative concepts and models used in market microstructure analysis. Below are brief descriptions of the core metrics referenced throughout the platform.
VPIN is a metric for estimating order flow toxicity, originally developed by Easley, Lopez de Prado, and O'Hara. It measures the imbalance between buy-initiated and sell-initiated trade volume within equal-volume buckets, providing a proxy for the probability of informed trading. Values approaching 1.0 indicate highly toxic (informed) order flow, while values near 0 suggest balanced, uninformed activity. QuantMedia presents VPIN as a heuristic research metric -- not a live trading signal.
HRP is a portfolio optimization technique introduced by Marcos Lopez de Prado. Unlike mean-variance optimization, HRP uses hierarchical clustering to group correlated assets and allocate risk in a top-down manner. This approach avoids matrix inversion, making it more numerically stable and resistant to estimation error in covariance matrices. Research papers on the platform discuss HRP in the context of quantitative portfolio construction.
PSR quantifies the probability that a strategy's observed Sharpe ratio exceeds a given benchmark, accounting for the non-normality and serial correlation of returns. It addresses the common problem of backtest overfitting by providing a statistical test for strategy performance. A PSR below 0.95 suggests insufficient evidence that a strategy's performance is genuine rather than a product of data mining.
The Hurst exponent measures long-range dependence in time series. Values above 0.5 indicate persistent (trending) behavior, while values below 0.5 indicate anti-persistent (mean-reverting) behavior. QuantMedia references Hurst exponent estimation on equity order flow as a regime detection tool.
QuantMedia operates as an algorithmic aggregation platform. Content is generated, sourced, and displayed through automated pipelines with no manual editorial intervention in the following areas:
Research papers and analysis articles are authored by QuantMedia Research and represent independent academic research. All quantitative models, code examples, and infrastructure descriptions reflect the author's research environment and are published for educational purposes.
QuantMedia is a non-commercial, independent research platform operated by QuantMedia. The platform does not execute trades, manage assets, or provide personalized financial advice. All content -- including market data, research papers, and analytical tools -- is provided for informational and educational purposes only.
Not financial advice. Nothing on this platform constitutes investment advice, a recommendation to buy or sell any security, or a solicitation for any financial product. All market data is sourced from third-party providers and may be delayed, indicative, or subject to error. Users should consult qualified financial professionals before making investment decisions.
No affiliation. QuantMedia is not affiliated with, endorsed by, or connected to any exchange (NYSE, NASDAQ), broker-dealer, or financial institution. References to exchanges, data centers (Equinix NY4/NY5), or specific technologies describe the author's research interests and do not imply any commercial relationship.
Data accuracy. While efforts are made to ensure data accuracy, QuantMedia makes no warranties regarding the completeness, reliability, or timeliness of any information presented. All data is provided "as is" without guarantee.