A risk number is only as good as the method behind it. This section explains how Alterscope produces the numbers it serves — what each model measures, the shape of the math, the inputs it consumes, and how we validate it — at a level a Head of Risk can judge for rigor and a quant can sanity-check. We publish the methodology and the formulas. We withhold the proprietary calibration — the exact factor weights, tuned parameters, and training specifics that represent the moat. Where a specific parameter is sensitive, we publish the formula and mark that parameter calibrated internally. The goal is transparency you can verify, not a recipe a competitor can clone.Documentation Index
Fetch the complete documentation index at: https://docs.alterscope.org/llms.txt
Use this file to discover all available pages before exploring further.
What we measure
Alterscope risk analysis runs at several layers, each documented on its own page:Risk factors
A seven-category factor model that scores a protocol or position 0–100, with per-factor attribution so you can see why a score moved.
VaR & Monte Carlo
Monte-Carlo Value-at-Risk and Expected Shortfall over correlated factor shocks, plus liquidity exit simulation.
Oracle classification
How price-feed configurations are categorized and scored for manipulation and staleness risk.
Pool risk scores
Independent lending-pool risk scores, each computed with a published peer methodology and surfaced side by side.
Graph intelligence
Network analysis over a knowledge graph: contagion cascades, concentration, communities, and capital flow.
Limitations
What we do not claim, the known boundaries of each model, and where coverage is still rolling out.
How to read this section
Each methodology page follows the same shape:- What it is — the question the model answers, in plain language.
- How it works — the method and the formula, with the inputs it consumes.
- What it means for risk — how to read the output and where it shows up in the API.
- What we withhold — the parameters that are calibrated internally, stated explicitly.
- Estimates, not guarantees. Every output is a model estimate — not a prediction of loss, and not investment advice. The Limitations page is required reading, not a footnote.
- Trust is a first-class field. Every API response states how fresh its inputs are and whether it passed an automated quality check — see Freshness & quality. A methodology is only credible if its outputs carry their own caveats, so ours do.