Quantitative Architecture & Mathematical Scoring
Technical Specification of Stochastic Telemetry Generation, Directional Scoring, and Null-Hypothesis Disclosure.
01. Single-Process Pipeline & Ingestion Protocol
Spectre Finance employs a single-process architecture. Directional telemetry is generated in-process by the application on a five-minute cadence, published to the billboard, and scored against exchange-quoted prices (via a public market-data feed) once the corresponding bar has closed.
[CSPRNG] --(scheduled cycle: /api/cron/generate, T = 300 s)--> [Spectre Billboard] --(Web UI)--> [Subscriber]02. Entropy Score Definition
The reported entropy score is a pseudorandom scalar drawn uniformly from [50, 75] and is statistically independent of the selected direction and of all market covariates. It is a presentation parameter of the stochastic source and carries no information about the selected direction or the realised outcome; hit rates are expected to converge to the null in every score band (see §04).
Larger values indicate a larger draw from the uniform pseudorandom source; they carry no information about the realised outcome.
Compare score bands with observed hit rates on untouched future data; under the null, every band converges to a 50% hit rate as n → ∞.
03. Directional Scoring & Verification
Each forecast is verified ex post: once its five-minute bar closes it is scored correct, incorrect, or push against the observed price. An absolute log return within a threshold (default 0.5 bps) is a push and is excluded. The dashboard reports directional accuracy over settled forecasts:
Under the null hypothesis the expected hit rate is exactly one half (see §04).
04. Stochastic Process Specification & Null-Hypothesis Disclosure
The directional output dt ∈ {+1, −1} is realised as an independent, identically distributed Bernoulli(½) trial (i.e., a fair-coin process) sampled once per asset per bar from the operating system's cryptographically secure pseudorandom number generator (CSPRNG). The sampling procedure is memoryless and takes no market data as input; the price feed is consulted solely to settle outcomes after the fact.
I(dt; rt+1) = 0 E[hit rate] = ½
Consequently the signal sequence is a martingale difference sequence with respect to the price filtration, and under the efficient-market null hypothesis (H0) the expected directional accuracy is exactly one half. Observed deviations from 50% are attributable to sampling variance, distributed approximately N(½, 1/4n), and should not be interpreted as alpha, skill, or predictive power.
No neural network, gradient-boosted ensemble, or other learned estimator participates in signal generation. This publication is a demonstration of stochastic telemetry presentation and is provided for satirical and educational purposes. Legacy materials from earlier maintainers are kept in the archive.