How SVET
Rating works.

Provenance and authorship

Source theory versus implementation.

The owner-controlled manuscripts SVET World Index v3 and SVET World Index 3.0 (with Comments) define normalized Shannon histogram entropy, A–D grades, the seven-observation Metric direction sign, the 4×4 hierarchy and broad proxy families. The manuscripts do not completely define the bin count, windows, minimum samples, exact public series, missingness, aggregation weights, aggregate direction or validation design.

Those missing rules are explicitly labeled versioned mvp-0.1 decisions. They are not retroactively attributed to the manuscripts. Evernomics supplies the motivation of durable, intergenerational wealth stewardship; it is not proof that any asset grows forever and does not turn a rating into a recommendation.

Four-Composite architecture

Sixteen views of one system.

The Telegram headings Nature, Science/Politics, Markets and Tribe/Crypto are a compact business-report crosswalk. They do not replace or recalculate the research Composites.

Operational data dictionary

Every Metric and its evidence.

Raw provider responses are retained with retrieval times and checksums. Where provider terms restrict redistribution, the public layer exposes definitions, aggregate observations, scores and provenance rather than republishing unrestricted raw feeds.

Formal calculation

From observations to one rating.

Entropy grades

A is lower dispersion and D is higher dispersion. Neither is automatically good or bad. Stable progress, stable decline and inactivity can all be low entropy; productive diversity and destructive turbulence can both be high entropy.

Direction

A Metric direction is +1 when the current transformed value is above the arithmetic mean of the preceding seven valid native observations, −1 when below, and 0 for an exact tie or insufficient history. The current observation is excluded from that mean. Composite direction is the sign of its four-Metric direction vote; total direction is the sign of the four Composite directions.

Activity percentile

Each Metric declares an interpretable activity value—such as flux, fatalities, price, funding, TVL or passenger ratio. Its tie-aware midrank against prior observations is:

High activity means unusual activity relative to that Metric's own history, not a beneficial outcome.

Alignment

Cross-Composite alignment is A = (dI + dII + dIII + dIV) / 4, ranging from −1 to +1. Alignment strength is its absolute value. all_up and all_down require unanimity across all four Composite directions.

Research hypothesis

What the Index may predict—and what it cannot yet claim.

The owner's theory is that a broad regime is more credible when natural conditions, institutions/knowledge, established markets and community/industrial systems move in alignment. Synchronized high activity and persistent positive alignment may mark a mature regime; depressed activity followed by persistent positive alignment may mark recovery. These statements are hypotheses, not buy/sell rules.

The proposed primary test asks whether lagged four-Composite alignment improves 90-day forward-return forecasts for BTC, Nasdaq and the S&P 500 beyond ordinary market predictors and Composite III alone. Secondary outcomes include 30/180-day returns and maximum drawdown.

• Chronological walk-forward testing only; no random time-series shuffling.
• Features lagged to information available at each prediction origin.
• Historical-mean, market-only and Composite-III-only baselines.
• Untouched holdout, block-bootstrap uncertainty and multiplicity correction.
• Negative, rejected and inconclusive results preserved with favorable ones.

Daily production and audit trail

One calculation, two interpretations.

1. The runner checks every free public source and stores response bytes, URL, retrieval time and checksum.
2. The canonical score uses the latest completed UTC day; the 6:00 a.m. Pacific report normally uses that completed day.
3. Each source is parsed and transformed according to the frozen data dictionary.
4. Sample and freshness checks attach quality flags; source failures never become invented zeros.
5. Deterministic code calculates 16 Metric scores, four Composite scores and the total. This numerical step uses no LLM and zero OpenAI tokens.
6. SQLite, CSV, JSONL and immutable JSON preserve formula/code versions, run ID and result hash.
7. The Telegram model writes only the narrative. Code inserts and validates the exact stored score lines before posting.
8. The final post is archived with model usage, response ID, Telegram ID, Index run and hash.

Freshness

Service-level ages are 3 days for calendar-daily, 5 for business-daily, 45 for monthly and 120 for delayed-release series. Values within the limit may be carried with their date; values outside it become unavailable.

The operational pipeline can still calculate when all historical series exist but current observations are stale. It retains the last Metric score and labels the total calculated_with_unavailable_inputs. Such output is a provisional business view, not a quality-passing confirmatory research observation.

Validity and interpretation limits

What could make the signal wrong.

1. Construct validity: one proxy cannot exhaust a philosophical Metric. Wikipedia attention is not truth or politics; TSA travel is not global emotion; Bitcoin is not all crypto.
2. Histogram sensitivity: entropy changes with windows, bin count, outliers and the rolling minimum/maximum.
3. Publication/revision bias: weather, economics, SEC and conflict sources arrive on different schedules and can be revised.
4. Geographic/platform bias: multiple proxies are U.S.-only or English-web measures.
5. Selection risk: attractive proxies and baskets can be chosen to fit a story unless frozen before outcomes.
6. Temporal leakage: latest-revised values may differ from what was public at a prediction origin.
7. Interpretation risk: entropy is not direction, desirability, causation or return.
8. Statistical power: unanimous regimes may be too rare for decisive testing.
9. Propagation risk: attention is not truth, authority activation is not objective danger, and a direct-jump story must not become a validated forecast without historical event tests.

Change control

mvp-0.1 is deterministic and immutable. Corrections append new source/formula versions; rejected runs remain visible. A more complex model is accepted only if it improves a preregistered out-of-sample criterion and remains interpretable. The language model has no numerical role.

Plan 3 safeguards

1. Do not confuse attention with truth.
2. Do not confuse official announcement with objective danger.
3. Do not force every event into a universal reading.
4. Do not rewrite mvp-0.1.
5. Do not claim mind-reading.
6. Test the direct-jump rule with historical event studies.

Bibliography and research record

Primary sources and reproducibility.

1. Shannon, C. E. (1948), A Mathematical Theory of Communication, Bell System Technical Journal 27, 379–423 and 623–656.
2. Owner-controlled SVET World Index v3, SVET World Index 3.0 (with Comments), and local Plan 3 source /home/svetrating/eversvet/svetindexplan3; full source text not redistributed here.
3. Natural Resources Canada solar radio flux.
4. Wikimedia Analytics Pageviews API.
5. NASA POWER Daily API.
6. GDACS API and WHO Disease Outbreak News.
7. Crossref REST API.
8. UCDP documentation.
9. World Bank Commodity Markets.
10. SEC Form D datasets.
11. TSA checkpoint travel numbers.
12. OFR Financial Stress Index.
13. DefiLlama API documentation.
14. Coinbase Exchange documentation.
15. U.S. Census Business Formation Statistics.

AI-assistance disclosure

The SVET/Evernomics concepts, research direction, four-Composite framework and owner decisions belong to the project owner. Codex assisted with methodology recovery, operationalization, source mapping, code, tests, documentation and prose. The deterministic calculation does not use an LLM. Any future academic submission must preserve an accurate disclosure and follow the receiving institution's current policy.