Understand the research behind the screen.

Dates, freshness & missing observations

An observation date describes the period measured. A release date describes when the publisher made it available. A capture time describes when this app stored it. These can differ substantially; a successful refresh does not mean the observation is new.

Fresh, delayed, stale, missing, sample, manual and admin-reviewed labels describe data state or provenance, not accuracy. Delayed can be normal for a scheduled publication. Freshness windows differ by series; there is no universal real-time feed. Dates use Australian display conventions; use the timestamp’s stated timezone.

A blank value or dash is not zero. Some charts need imported data or enough usable history. Sample data and model defaults must not be interpreted as independently sourced market observations. Historic revisions mean stored history is not guaranteed to reproduce what an investor knew at the time.

Prices, units & chart comparisons

AUD price conversion: USD per troy ounce divided by USD per AUD gives AUD per troy ounce. For example, US$100 / 0.50 = A$200. This is a unit example, not a market quotation. A kilogram and a troy ounce are different product units; compare matching weights, purity, dates and prices.

Dealer buy and sell-back prices describe different sides of a transaction. Spreads and premiums depend on the comparison basis shown in each dashboard. Confirm product specification, fees, delivery, storage, availability and a current executable quote with the dealer.

A gold/silver ratio compares gold USD/oz with silver USD/oz. A metal/oil ratio compares metal USD/oz with WTI USD/barrel; these are descriptive ratios, not fair-value estimates. Indexed chart series compare changes from a common starting level, not their original price levels.

Long archives can use monthly observations. Moving averages or short-horizon comparisons are meaningful only for the actual observation frequency. Futures-based historical references can differ from spot through contract terms, roll effects and observation timing.

Composite scores are not all calculated the same way

Selected currency and stress composites rank an input within its own trailing 365-day observed minimum and maximum: 100 × (latest − minimum) / (maximum − minimum). This is a range position, not a statistical percentile. At least five observations and a non-zero range are required; unavailable ranked inputs are omitted and remaining weights are re-normalised.

Commodity Currency uses Brent 50%, gold 30% and silver 20%. AUD Weakness uses inverse AUD/USD 40%, inverse AUD trade-weighted index 40% and broad US dollar index 20%. Banking Stress uses St. Louis stress 40%, Chicago conditions 30% and high-yield spreads 30%. Geopolitical Risk uses the overall index 50%, acts 25% and threats 25%.

Other composites still use fixed bands, not dynamic trailing ranges: real rates, inflation pressure and liquidity. Components are clamped to 0–100 and blended across available inputs. A high score has the meaning given by its label, not a universal investment interpretation. A score made from fewer inputs is not directly equivalent to one with full coverage.

Real Rate combines inverted 10-year TIPS yield (60%) with inverted policy rate minus 10-year breakeven (40%), mapped over −2.5 to +2.5. Inflation Pressure maps breakeven and 5y5y expectations over 1–4, and Australian CPI over 1–6, with weights 40/30/30. Liquidity maps US M2 growth over −2–10, Australian M3 growth over 0–12 and inverted reverse repo over −1500–0, with weights 45/35/20. Available weights are re-normalised. See the individual chart for units and input status.

Dedicated signal model: version 2.0

The dedicated signal, vehicle-selection and model-history pages are administrator-only in the interface. Publishing this explanation does not grant access or certify the underlying permission boundaries. This is a rule-based research model, not a trained probability model, portfolio optimiser or suitability assessment.

Gold starts at 50 and applies fixed increments for real yields, AUD/USD, inflation breakevens, volatility, credit spreads, policy rates and the broad dollar index; available money-supply growth and reported official holdings changes add further context. The result is clamped to 0–100. This separate model has not been converted to the trailing-range method used by selected composites.

Silver starts at rounded 85% of the gold score, then adds 5 when the gold/silver ratio exceeds 85, zero above 75, or subtracts 5 otherwise. Vehicle comparisons use fixed adjustments to the gold/silver scores, including oil-price and volatility penalties for miners. They are not security-specific valuations or executable allocations.

The displayed confidence is a coverage heuristic: 92 minus 8 for each defaulted input, 6 for each sample input, 4 for each stale/missing-status used input and 1 for each delayed used input, rounded and bounded to 20–95. It is not a probability of profit, forecast accuracy or statistical confidence interval. A run using tracked defaults or samples is flagged as sample.

The model’s money-supply input annualises the change between available stored endpoints when at least 60 days apart. It is not necessarily a matched-year-on-year observation. Model validation, consistent input coverage and point-in-time performance testing remain separate work.

Market alarms & historical replay

Alarm rule version robust-move-v1 compares the latest usable observation with the previous distinct observation, in percentage or absolute units depending on the indicator. Source gaps, freshness limits and positive-value requirements for percentage calculations can prevent evaluation.

At least 30 valid prior release-to-release changes are required. The robust boundary is median absolute historical change + 3 × 1.4826 × median absolute deviation of those absolute changes. An alarm requires the current absolute change to meet the indicator’s floor and exceed that robust boundary. The displayed threshold is the greater of the floor and robust boundary.

Floors differ by indicator, for example 3% for gold history, 5% for silver history and 0.20 percentage points for selected yields. Active alarms expire after 48 hours under this rule version. These screening choices are not calibrated probabilities or buy/sell instructions.

Historical replay reconstructs threshold events, not a trading strategy or return series. It does not establish out-of-sample performance and can reflect later data revisions. Backtested events are labelled separately and are not live dashboard alarms. User-created alert rules are a separate management interface, not a promise of email, SMS or mobile push delivery.

Data rights, institutional use & research limits

This guide is a source inventory and explanation of current calculations, not a legal clearance, financial recommendation or independent audit. No institutional licence, exchange entitlement or right to redistribute raw or derived data is granted with an account.

Before government, central-bank or other institutional deployment, establish documented rights for display, storage, automated capture, derived outputs, AI use, subscriber access and white-label redistribution for each source. Attribute original providers and verify their current terms or obtain written permission where required.

Institutional readiness has not been certified. Security and entitlement hardening, independent access testing, licensing clearance, operational recovery evidence and institutional requirements assessment remain prerequisites. Do not upload confidential reserve, portfolio or personal financial records on the assumption that this research interface is approved for them.

General research does not consider personal objectives, financial circumstances or needs. Historical relationships can break down. Seek appropriately licensed advice where needed and verify the original release before relying on any figure.