BBB Accredited · Rating A — How can we help you? Talk with an expert today: (800) 200-9553
Reading everything that could move your money — central-bank language, Treasury issuance, inflation prints, energy and shipping disruption, mine supply, the assets you actually own — is several hours of work before breakfast. Then the world keeps moving all day. Nobody sustains that, which is why most people end up reacting to whichever headline shouted loudest.
Capstone Market Intelligence does that reading continuously, connects it to what you hold, and tells you plainly what changed, what it touches, and what is still unknown. Not a prediction. A briefing you can act on.
Written by the Capstone Metals research desk. Reviewed by Daniel Kenney, Chairman and Founder. Last updated September 21, 2026.
An original Capstone Metals editorial illustration with a brass research compass at its center. Around it are market charts and securities, a portfolio of gold and silver, the Federal Reserve building and a desk of current-event source documents. The connected composition represents separate intelligence layers challenging and informing one another before a conclusion reaches the research record. It describes the platform's scope, not a prediction or performance claim.
There has never been more market information available and never less time to make sense of it. The value is not in having the data — anyone can have the data. The value is in the assembly: knowing that this policy change, this supply disruption and this correlation shift all point at the same holding in your account.
A worked example of the kind of chain the system is built to draw: refined-fuel output is concentrated in a small number of facilities, so verified damage to one of them is a supply-side diesel question before it is anything else. Diesel is freight, and freight is the cost of almost every physical good, which is an inflation input, which is a rates input, which reaches bonds, industrial equities and metals in a particular order. Each link in that chain is checkable, each carries its own confidence, and the ones that cannot be verified are marked unknown rather than assumed.
That is the whole intent: knowledge you can apply, assembled while it still matters, instead of an archive you never read.
These are not six products. They are six passes over the same day, each answering something the others cannot, coordinated by one research orchestrator. Deterministic code owns every number; the machine-learning layers classify regimes, rank opportunities, calibrate probabilities, select features, flag anomalies and weight the ensemble. No model is permitted to calculate a result or decide anything on its own.
Layer 1
Price, trend, volatility, correlation and regime
Every covered asset is read in context rather than alone: how it is behaving, how volatile that behaviour is, what it is moving with, and which market regime we appear to be in.
Layer 2
The Fed, the Treasury, rates, inflation, issuance
A rate decision is not a headline here, it is an input. The system carries the mapping from a policy change to the assets and sectors it historically touches, so the question becomes what it means for what you hold.
Layer 3
Named sources, timestamps, contradictions, confidence
Events arrive with their source, their event time and their publication time attached, plus any reporting that contradicts them. A refinery strike, a shipping lane, an export ban — logged as evidence, not as a mood.
Layer 4
Concentration, overlap, drawdown, co-movement
The layer most people never get. Two holdings that look diversified can be the same bet wearing different names, and that only shows up when the whole portfolio is read at once.
Layer 5
Point-in-time data, locked holdouts, cost assumptions
Nothing becomes a rule because it looked good on a chart. A strategy family is tested on data as it existed at the time, against a reserved period it has never seen, with fees and spread charged on every trade.
Layer 6
Audit trail, kill switch, remembered dead ends
Every run is recorded — what was asked, what was answered, what it cost, whether it passed. Failed ideas are kept so the system stops rediscovering the same dead ends, and a human can halt the whole thing at any time.
What moved, why it may matter to what you own, and what remains unresolved — in plain sentences, with the sources attached.
Concentration, overlap and shared exposure across everything you hold, so diversification is measured rather than assumed.
You see what the tested rules would do, and when they would do it, without a single dollar at risk. What you do with that is your decision.
Every strategy's tested windows, costs, reserved holdout and result — published whether it flattered us or not.
There is no single best strategy, and any firm claiming one is selling a certainty that does not exist. What the literature does support is that certain families of behaviour have been documented across many markets and decades. We test those families, and we write down what they do not prove. The specific parameters and combinations we use are our own work and stay ours — the method is public, the recipe is not.
Persistence of an instrument's own past return over 1–12 month horizons has been documented across liquid futures markets (Moskowitz, Ooi & Pedersen, 2012).
What it does not prove: Documented in a sample period, before costs at the size we would trade, and with long flat or negative stretches. It is not a prediction.
Status: implemented
Relative-strength ranking effects appear across and within asset classes (Asness, Moskowitz & Pedersen, 2013).
What it does not prove: Crowding, crash risk in reversals, and sensitivity to the ranking window and rebalance frequency.
Status: implemented
Scaling exposure by recent realised volatility has been studied as a portfolio management overlay (Moreira & Muir, 2017).
What it does not prove: Reduces volatility, not loss. Adds turnover and therefore cost.
Status: implemented
Valuation spreads have explanatory power in cross-asset studies.
What it does not prove: Needs point-in-time fundamentals with reporting lags we do not yet ingest.
Status: research module
Short-horizon reversal and cointegration effects are widely studied.
What it does not prove: Highly sensitive to costs and to regime breaks; needs careful capacity analysis.
Status: research module
Scheduled macro releases move metals and rate-sensitive assets.
What it does not prove: Needs vintage-accurate release data, which is a later ingest module.
Status: research module
Implied volatility surfaces and skew carry information.
What it does not prove: No options data source. Deliberately not built.
Status: deferred
Order-flow imbalance explains short-horizon price moves (Cont, Kukanov & Stoikov, 2014).
What it does not prove: Requires book/trade data and realistic execution modelling. Deliberately not built — a fabricated feature set would be worse than none.
Status: deferred
Promising in simulation.
What it does not prove: Only meaningful once the simulator and data discipline are mature. Deferred.
Status: deferred
Daily research note
The daily pass is not an article mill. It is a disciplined reading loop: identify what changed, connect it to named assets and households, separate verified facts from analysis, and preserve what remains unknown. If a change deserves a public page, it is routed through the same editorial and technical checks as any other Capstone publication.
People ask for a win rate because it sounds like the whole answer. It is not. What decides an outcome is the size of the gains kept against the size of the losses taken — expectancy, not accuracy. A disciplined approach that is wrong more often than it is right can still come out ahead, and an approach that is right most of the time can still be ruined by one position it refused to cut.
Illustration only — hypothetical arithmetic, not our results and not a projection
Take ten decisions. Four go well and are allowed to run to roughly three units each; six go badly and are cut near one unit each. That is a 40% hit rate and, in this arithmetic alone, twelve units gained against six lost. Reverse the discipline — let the losers run to three units and take the winners at one — and the same 40% becomes a serious loss. Nothing about the accuracy changed. The rules around the position did.
Real markets add costs, gaps, slippage and stretches where nothing works. This example shows why we study position sizing, predetermined exits, diversification across uncorrelated holdings and risk scaling at least as carefully as we study signals.
That is also why we do not publish a headline win rate. We measure one internally — every directional claim is written down before the outcome is known and scored against our own stored price history, wins and losses alike — and we hold it to a deliberately severe standard before any figure appears on this site. It has not met that standard, so there is no number here. If that ever changes, the figure will be published with its sample size, its window, and the losses included.
Integrated market, policy, current-event, portfolio and quantitative-research access at $199, $299 and $599 per month. Cancel from your own dashboard at any time.
Research and education only. Nothing on this page is investment advice, a recommendation to buy or sell any security or metal, or a forecast. No outcome is promised or implied. Simulated and historical results do not indicate future results, and any strategy discussed here may lose money. Speak with us about your own circumstances before acting: (800) 200-9553.
