Data & Methodology

Why this page exists

A financial figure without its provenance is close to useless. The same “current price” can mean a live quote, a fifteen-minute-delayed quote, or yesterday's close, and each one supports a different decision. This page documents the origin, refresh cadence, and calculation of everything Portfolio Watch displays, along with the gaps we know about.

We would rather publish an uncomfortable limitation than let you over-trust a number. Where our data is stale, approximate, or hard-coded, it is said so plainly below.

Data sources

Portfolio Watch aggregates public market data. We are not a market data vendor, an exchange, or a broker, and we do not hold any proprietary price feed.

Some schedules for recurring macroeconomic events, notably Federal Reserve FOMC meeting dates, are maintained as a fixed internal calendar rather than fetched live. These are published well in advance by the issuing body, but if a date is rescheduled our calendar will not reflect the change until we update it.

Refresh intervals

Nothing on Portfolio Watch is a real-time quote. Every figure has a defined staleness window, and the windows differ substantially between features.

The practical consequence is that intraday readings can lag the market by several minutes, and any figure produced by a scheduled job reflects conditions at the time that job ran, not the moment you loaded the page. Do not use this site for execution decisions that depend on a current price.

How the main figures are calculated

Each signal has a dedicated guide covering both the formula and its failure modes. In brief:

• Zacks strategy performance simulates investing $100 every market day in the most recent Bull of the Day, holding while a pick remains among the 30 most recent, against an identically funded $100-per-day SPY position. Prices are daily closes; dividends, commissions, and slippage are excluded. Full detail in the Zacks backtest guide.
• Fear index reports the latest close of the VIX for the S&P 500 track and the VXN for the Nasdaq track, mapped onto labeled tiers. Thresholds and caveats in the VIX and VXN guide.
• Options income screens out-of-the-money cash-secured puts and covered calls on liquidity, delta band, and an annualized yield of (premium / capital) × (365 / days to expiry). Delta is computed locally with Black-Scholes. Full parameters in the options income guide.
• Implied earnings move is the at-the-money straddle price divided by the current share price, using the first expiry on or after the earnings date. Detail in the implied move guide.
• Precious metals pairs front-month futures with the corresponding physically-backed ETF and derives the gold/silver and gold/platinum ratios, relative volume, and a nearest-expiry call/put ratio. Detail in the precious metals guide.

Use of AI-generated text

Two features on Portfolio Watch produce text using large language models, and we label them here rather than leaving you to guess.

Company profiles — the short descriptions of what a company does, its business model, and its typical customers are generated by a language model working from the company's own published business summary and its sector and industry classification. Each profile is generated once and cached.

Email digest summaries — the sentence or two describing what changed for a ticker since the previous digest is generated by a language model from the underlying metric changes.

Known limitations

An honest inventory of the weaknesses we are aware of:

• The forward-return table on the fear index is static. The historical forward 1-month, 3-month, 6-month, and 1-year figures shown for each fear tier are fixed reference numbers held in the application, not results recomputed from live data. Read them as rough historical context, not as a current backtest.
• The Zacks pick history begins when we started recording. There is no data from before our first stored pick, so the performance series covers a short and arbitrary window that has not been tested across a full market cycle.
• We do not store the price at the moment a pick was published. The backtest therefore enters at that day's close rather than at the price a reader could have acted on.
• No dividends, commissions, or slippage are modeled anywhere. This flatters the strategy relative to the SPY benchmark, because the benchmark's distributions are a real part of its return.
• Option quotes can be stale. Outside market hours, and on thinly traded contracts, we fall back to the last traded price, which may be hours old. Implied volatility in that case is back-solved rather than quoted.
• Data is sourced from free public endpoints. These occasionally return missing fields, delayed values, or corrections after the fact, and they carry no accuracy guarantee. Where a source fails, a panel may show incomplete data rather than an error.
• Scraped content can break. The Zacks picks are read from page markup. If that markup changes, collection can silently miss a day until we notice and fix it.
• We cover U.S.-listed securities. Coverage of other markets, small caps, and recently listed companies is inconsistent.

Editorial position

Portfolio Watch is an independent project. We do not accept payment to feature, rank, or favorably describe any security, and no company can pay to appear in our data. Which tickers appear is determined by the data sources and by the watchlists our users configure.

The site is funded by display advertising and voluntary reader contributions. Advertising is served by a third-party network and its content is not selected, endorsed, or reviewed by us. Advertisers have no input into our editorial content, our calculations, or which securities appear. Where a link earns us a commission, we disclose it.

We are not registered investment advisers, brokers, or dealers, and we do not provide personalized financial advice. Nothing on this site is a recommendation to buy or sell any security.

Corrections

If you find a figure you believe is wrong, we want to hear about it. Email portfolio-watch@gmail.com with the page, the figure, and what you expected to see. We investigate reported data errors and correct the underlying calculation where the report is confirmed. Material corrections to a methodology are reflected on this page with an updated revision date.