How we build
the Trinity score
Six intelligence engines. One composite reading synthesized from on-chain, macro, cycle, narrative, price, and exchange data. Open methodology. Public data only. EU-native infrastructure. No black box.
The cardinal idea
The crypto market cannot be read along a single dimension. On-chain tells what happened on the network. Macro tells the global monetary context. Cycle tells where we are in Bitcoin's long timeline. Narrative tells what is capturing public attention. Price tells what the market is paying. Exchange tells flows and positions. Each one read alone is partial ; all of them read together draw a picture nobody can interpret in minutes, unless someone synthesizes them.
The Trinity composite score is that synthesis. Six independent intelligence engines, each calibrated on public data, measured by four statistical methods whose results are published as measured, continuously audited, and combined into a single 0-100 reading. This score is not a prediction. It is an instant reading of the market state, like an aircraft cockpit that aggregates dozens of instruments into a few synthetic dials.
The six intelligence engines
Each engine analyzes a distinct market dimension and produces its own indicators. Their synthesis into a composite score captures information no single source can provide.
On-Chain Bitcoin
Holder conviction read at the source, not in the narrative.Direct reading of the Bitcoin blockchain : supply, UTXOs, cohort behavior, distribution, miner flows, cointime economics. MVRV, NUPL, SOPR, Realized Price, Cointime Price and the full reference on-chain indicator set are computed from raw UTXOs on a self-hosted Bitcoin node, no third-party aggregator proxy.
Macro Intelligence
The global macro backdrop that drives every cycle, multi-central-bank.Global monetary and financial context : sovereign yield curves, real yields, DXY, liquidity conditions across the Fed, ECB, BoJ, and BIS, monetary aggregates, emerging markets, gold and commodities. Official public series going back to the deepest available archive, methodology reproducible from primary sources.
Cycle Intelligence
Position in the Bitcoin cycle, measured, not guessed.Position in Bitcoin's long timeline : halving-anchored cycle framework, overlay comparison of successive cycles, Pi Cycle Top, Spiral, Block Progress, seasonal reading. Answers the question 'where are we' without being trapped in price-only reading.
Narrative Intelligence
Public attention before it translates into price.What is capturing public and institutional attention : Google Trends, Wikipedia, GitHub (developer activity), SEC EDGAR (US official filings), GDELT (global news), BTCMap (merchant adoption). Attention indicators, editorial momentum, regulatory readings, all from public sources.
Price Intelligence
Cross-asset structure and rotation, read as a single breath.Price reading across every dimension : BTC, ETH, SOL, cross-asset, reference technical indicators, multi-timeframe momentum, volatility structure. Price is never read alone, it is cross-referenced with the other engines to distinguish what the market is paying from what fundamentals justify.
Exchange Intelligence
Smart money in motion, reserves, net flows, derivatives.Behavior of platforms and participants : exchange reserves, net inflows and outflows, derivatives open interest, funding rates, smart-money positions on perpetual L1s. Public exchange APIs feed these series, no social media scraping, no third-party aggregator.
EDCD, Education-Driven Chart Design
The cardinal principle : a reader who only reads a chart's education must be able to anticipate what they will see. And conversely, a reader who only sees the chart must be able to anticipate what the education describes. Any education promise without a matching visual primitive is a violation. Any visual primitive not mentioned in the education is a violation.
Concretely, every Trinity chart passes five orthogonal audits before going live : DOM structure, data integrity, multimodal vision (a sub-agent verifies the pixel-by-pixel rendering against the educational contract), industry-source cross-check, and backend pipeline reliability. No chart reaches production if a single one of the five fails.
What EDCD guarantees
- Every chart teaches what it shows.
- Every educational promise is rendered visually.
- No surprise for the reader.
- Verified at every deployment.
What EDCD rejects
- Decorative charts without an educational contract.
- Marketing education decoupled from actual rendering.
- Metrics that self-cite without a definition.
- Any chart certified without human visual audit.
TAE, Trinity Audit Engine
Trinity Audit Engine is a permanent, autonomous, self-maintained system that guarantees the fidelity of every data point served by the platform against its official source or industry standard. Seven independent audits run continuously, daily, weekly, hourly, across every on-chain, macro, cycle, narrative, price, and exchange metric.
Every audit produces an immutable, append-only SHA-256 signature. If a single one of the seven audits fails, the system aborts the cycle and alerts ; no invalidated metric is served to users. External sources used for cross-checks are all public : FRED, ECB, BoJ, BIS, SEC EDGAR, public Bitcoin services, official archives. No commercial aggregator inside the audit loop.
Why TAE matters
A composite score is only as good as the fidelity of the data feeding it. TAE is the guarantee that the Trinity reading is never an average of accumulated errors. It is the institutional trust layer that a research desk, a family office, or a quant can audit at any moment, every signature, every accepted tolerance, every comparison with the official source is traceable.
Walk-Forward validation
Walk-Forward is the industry standard for validating quantitative models. It is a robustness test that prevents overfitting, the main pitfall of financial indicators.
The process
- Training, the indicator is calibrated on a historical window (e.g., 2012-2018).
- Out-of-sample test, the indicator is evaluated on data it has never seen (e.g., 2019-2020). No readjustment.
- Roll forward, the window advances and the process repeats, so the measurement covers several successive periods rather than a single one.
- Grading, the grade reflects out-of-sample performance as it was measured, from S+ to F. A low grade is not erased: the indicator stays readable with its grade, the date of its campaign and the perimeter tested.
Why this matters
Most indicators on the market are optimized on past data, they create the illusion of working because they were fitted AFTER the fact (overfitting). Walk-Forward eliminates this bias by systematically testing on unseen future data. The method guarantees nothing on its own: it produces a grade, and it is that grade, published on each indicator page with the date of its measurement, that says whether the indicator held out-of-sample or not.
Popper falsifiability
Inspired by philosopher of science Karl Popper, every Trinity indicator must be falsifiable. If an indicator cannot fail in a measurable way, it provides no useful information.
What we test
- Does the indicator produce outputs that can be objectively true or false?
- Does it have measurable analytical value on out-of-sample data?
- Is the result reproducible with the same inputs?
What we reject
- Vague indicators that claim to “explain everything” (non-falsifiable).
- Metrics that are not reproducible or depend on opaque proprietary data.
- Indicators that failed Walk-Forward testing (past performance ≠ future).
Additional robustness tests
Beyond Walk-Forward and Popper, two additional tests measure robustness. Their result is published per indicator, including where it is negative or where the measurement has not yet been made.
LOOCV (Leave-One-Out Cross-Validation)
The reference market events are removed one at a time, and recall is recomputed at each removal. The measure answers one precise question: does the result rest on a pattern, or on a single lucky episode? A removal that drops recall past the accepted threshold marks that dependency.
Perturbation tests
The aggregated score is re-drawn under ±10% random noise over several dozen trials, and the degradation of both correlation and recall is measured. A reading that flips on a rounding is reported as fragile, and that fragility is published alongside the indicator rather than hidden.
Multi-validation grading system
Each indicator receives a grade based on the combined Walk-Forward + Popper + LOOCV + Perturbation results. This grade determines the weight it carries in the composite score and whether it is admitted into the library.
| Grade | Meaning | Library status |
|---|---|---|
| S+ / S | Excellence, stable performance across all periods and all tests. | Maximum weight in the score. |
| A+ / A | Very strong, robust performance with minor variations. | Substantial weight. |
| B+ / B | Solid, valid contributions to score diversification. | Moderate weight. |
| C | Contextual, useful as a complement, reduced weight. | Marginal or contextual weight. |
| D / F | Insufficient, failure on at least one critical test. | Published with its grade visible, carrying no weight in the composite score. |
Not every indicator satisfies all four methods, and those that fail one remain visible with their result. Each indicator page carries the status obtained per method, the date of the campaign that measured it and the perimeter tested, including where the measurement has not yet been made.
Public data only, zero black box
All data used by Trinity is public, verifiable, and legally accessible. No commercial aggregator inside the audit layer, no opaque proprietary source, no unauthorized social media scraping. Every datapoint served by the platform is traceable to its official source, and the reader can reproduce the method from that source.
On-Chain Bitcoin
A self-hosted Bitcoin full node parses the raw blockchain directly. UTXO-level computation for every on-chain metric (MVRV, NUPL, SOPR, Realized Price, Cointime Price, cohort behavior). Public mempool services provide real-time fee and block analysis.
Macro Intelligence
Fed (FRED), ECB, Bank of Japan, BIS, US Treasury : sovereign yield curves, real yields, DXY, monetary aggregates, liquidity conditions, gold, commodities. Official public archives going back to the deepest series available.
Cycle Intelligence
Direct derivation from the Bitcoin blockchain (halving heights, block progression) and from the same on-chain layer that feeds the on-chain engine. Cycle framework, Pi Cycle Top, Spiral and Block Progress are all computed from public blockchain data.
Narrative Intelligence
Google Trends (search attention), Wikipedia (editorial attention), GitHub (developer activity), SEC EDGAR (US filings), GDELT (global news), BTCMap (merchant adoption). All accessed via public APIs or open datasets, no social media scraping.
Price Intelligence
CoinGecko public API for spot prices, volumes, dominance and market caps across BTC, ETH, SOL and cross-asset. Standard technical indicator references computed locally, not relayed from any third-party indicator service.
Exchange Intelligence
Public exchange APIs for reserves, net flows, derivatives open interest, funding rates. Real-time on-chain data from the Hyperliquid L1 perpetual DEX : liquidations, smart-money positions, funding history. All public, all traceable.
EU-native, regulatory posture
Trinity Insights operates from the European Union with infrastructure fully within the EU. GDPR, MiCA and AMF compliance is not a cosmetic layer added on top of a US product, it is the foundation from which the platform has been built. Every API response and every dashboard page ships its institutional disclaimer ; every dataset respects European requirements for sovereignty and traceability.
MiCA (EU Regulation 2023/1114)
Trinity Insights is not registered as a Crypto-Asset Service Provider (CASP). The platform is an analytical educational and informational tool, not a brokerage, asset management, or investment advice service. This posture is published in the Terms and reinforced at every API response.
AMF, institutional disclaimer
The platform's vocabulary avoids any term that could be construed as an investment recommendation. Composite scores are presented as readings and indications, never as advice. This editorial discipline applies to every public description, every indicator name, every interpretation.
GDPR, minimum user data
No wallet connection, no balance reading, no position extraction. The platform is entirely read-only on the market, it never touches user funds. Usage data is minimized and EU-resident.
AI Act, algorithmic assistance disclosed
Certain analyses are produced with algorithmic assistance (Article 50 of the EU AI Act). This assistance is explicitly disclosed wherever relevant. The platform remains an analytical tool, the user retains final responsibility for any decision.
What Trinity is NOT
- Trinity does not provide investment advice. It is an observation and analysis tool.
- Trinity does not predict prices. It measures market conditions and their historical context.
- Trinity never accesses your funds. The application is entirely read-only. No wallet connection is required.
- Trinity makes no performance promises. Past indicator results do not predict future outcomes.
Explore the scores yourself
Free 7-day Pro trial. Card required, nothing charged during the trial. Cancel anytime before billing.
Start 7-day free trial