MarketRadar Methodology

What this page is

This page explains, in plain language, how MarketRadar's numbers, levels and articles are produced. Every detail below is taken from the site's own code, data files, or already-published pages. Where a detail could not be verified in code or data, it is left out and listed at the bottom of this page as an open item — it is not guessed.

Timestamps and timezone

Article timestamps and analysis "Price as of" times are shown in UTC. A daily analysis opens with a line such as "Price as of 15:07 UTC on October 25, 2026". That line is the moment the analysis was generated. Analysis pages are snapshots taken at that time — they do not update live. If an article is changed more than five minutes after publication, the byline adds an "Updated HH:MM UTC" line.

Homepage ticker and market snapshot

The homepage ticker and the "Market snapshot" card are painted from the site's snapshot file, which carries its own updated time, and then from live data feeds:

Gold and silver: spot prices from gold-api.com.
Bitcoin and Ethereum: prices and 24-hour change from CoinGecko.
FX majors, WTI and indices: daily candles from Yahoo Finance, with percentage change measured against the candle nearest 24 hours earlier (timestamp-aware, so a missing candle does not skew the base).

Only gold, silver, Bitcoin and Ethereum refresh live (every 60 seconds) on the page. Euro / US dollar, pound / US dollar, the S&P 500 and the Nasdaq show the latest snapshot value, and the snapshot label says so explicitly. The label never claims all eight prices are live at once.

Analysis data sources

Daily analysis, the Levels Digest and the Weekly Outlook are built from candle (open / high / low / close) data:

Crypto and forex pairs (including gold): Kraken's public OHLC feed. Gold uses Kraken's PAXGUSD pair as a proxy for gold — it is not a direct spot-gold feed, and analysis pages do not claim otherwise.
US indices and stocks: Yahoo Finance's chart feed (for example, the S&P 500 as ^GSPC and the Nasdaq 100 as ^NDX).

Timeframes used are D1 (daily), H4 (4-hour, resampled from hourly candles), H1 (hourly) and M15 (15-minute). Data ranges in code are 1 year for D1/H4, 6 months for H1 and 1 month for M15.

Freshness gate

A level is only published if the latest candle is fresh: within twice the timeframe interval plus 30 minutes of the current time. Stale data fails the build instead of publishing an old price. One exception is coded for closed markets: a US index's Friday close, up to 4 days old, is accepted on a closed-market day and is labelled "Friday close - markets closed" — never presented as a live price.

How key levels are computed

Levels come from one deterministic calculation applied to the candles above (the same calculation feeds the article text and the charts, so the two cannot disagree):

Price is the last candle's close on each timeframe.
Trend context uses the 14-period Average True Range (ATR) and the 50-period simple moving average (SMA50).
Bias combines swing structure (whether recent fractal highs and lows are rising or falling, using 3-candle fractals) with price versus SMA50. A score of +2 or more reads bullish, −2 or less reads bearish, anything in between reads neutral. The daily bias is the digest's bias, with the 4-hour bias used only when the daily bias is neutral.
Resistance and support combine the last two swing highs/lows with a round-number grid (for example, every $25 for gold, every $1,000 for Bitcoin), limited to the three nearest levels on each side. A level essentially at the current price (within 5% of one ATR) is not counted as a level.
Supply and demand zones are centred on the most recent swing high above price (supply) and swing low below price (demand), each extending 0.4 × ATR on either side.
Liquidity pools are pairs of near-equal swing highs or lows (within 0.15 × ATR), limited to two on each side, and are only mentioned when such a pool actually exists.
Day and week context comes from the daily candles: previous day and previous week highs and lows, and the day range from the last 24 hourly candles.

Levels are checked before publication: the ATR is recomputed exactly (same 14-period lookback) and must match, and a second ATR check with a 20-period lookback must stay inside a wide band, along with resistance, support and fractal checks. A failed check blocks the article.

Scenarios and invalidation

Each daily analysis ends its scenario section with explicit triggers and failure conditions: a bullish view triggers on a 4-hour close back above the first resistance and fails on a 4-hour close below the first support; a bearish view triggers on a 4-hour close under the first support and fails if price reclaims the first resistance on a 4-hour close. The levels table lists the timeframe each level was spotted on (D1, H4, H1, M15), and scenario triggers use the 4-hour close unless a scenario states otherwise. Levels are zones, not exact lines, and waiting for a 4-hour close is how intraday noise is filtered out.

When a full analysis is published — and when it is not

Most days, the seven-asset Levels Digest (gold, Bitcoin, Ethereum, euro / US dollar, pound / US dollar, S&P 500, Nasdaq 100) is the analysis product. A separate full analysis for one asset is published only when that asset meets a threshold: either its hourly price has moved more than 1.5 × its daily ATR from the previous close, or a major catalyst less than 24 hours old directly concerns that asset (central-bank decisions, CPI/NFP releases, ETF/SEC decisions, exchange hacks or insolvency, major protocol events, or directly linked supply shocks). Otherwise the asset is covered by the Digest only, and the generator refuses a standalone analysis.

Fundamentals and sources

Fundamental backdrops in analysis are built from recent headlines that match the asset's keywords; unrelated headlines (for example, dividend or listicle stories for a forex pair) are filtered out, and fewer than two relevant headlines means no fundamentals links are shown for a weekly outlook. News articles name and link their original source (never just a homepage) in the source line, and quotes carry their outlet in the same sentence ("per <outlet>").

Content review before publication

As stated on our Editorial Policy: every article is read and verified against at least one other source or market data, rewritten independently in our own words and structure, credited and linked to its original source, and reviewed by a human editor before publication. On top of that, automated pre-publish checks run before any deploy: an originality check that fails on any shared 6+ word phrase with the source text, an AI-fingerprint heuristic check, and a publish QA gate covering broken internal links, title length, banned promotional phrases, pips arithmetic, dated Fed-odds figures, duplicated paragraphs across posts, links to removed posts, and internal/process wording leaking into reader-facing text. A failed check blocks the deploy; the article is not published in that slot.

Updates and corrections

Our approach is stated on the Corrections page and in the Editorial Policy: factual errors are corrected as soon as they are confirmed; material corrections are noted in the article with the date and a brief description of what was fixed, and are recorded in the Corrections log; minor fixes such as typos or formatting may be fixed without a note. At the time of writing, the Corrections log records no corrections yet. If you spot an error, email umairch204@gmail.com with the article link and the correction, or use the Contact page.

Pakistan hub data

The Pakistan Markets Hub uses verified figures, each labelled on the page with its date and source: the KSE-100 latest verified close (with day range, previous close and source), the SBP policy rate, gold and silver per tola with their source and date, and a macro table whose dates are shown in the table itself. USD/PKR comes from the Yahoo Finance snapshot. If no tola figure from the regular source is present, the page derives an indicative tola from the international gold price in US dollars multiplied by the USD/PKR rate, and labels it as derived. Older Pakistan stories on the page carry their own article dates; their headlines are not the current close.

What this page does not yet state

The following details could not be verified in the code or data read for this page, so they are deliberately not asserted above. They are open items for the owner to confirm before they are ever published as facts:

The refresh cadence of the snapshot pipeline (how often the snapshot file itself is regenerated).
Whether any data provider beyond those named above is used for any specific article (each article's own source line is the per-article record).
Any commercial or data-licence arrangement with a provider (see also the owner-review items on the Risk & Conflicts Disclosure page).