1. Abstract
ZARQ is a crypto risk intelligence platform that rates digital assets using two complementary models: a Trust Score measuring relative quality, and a Distance-to-Default (DtD) score measuring proximity to critical failure. Together they form the basis for crash prediction, contagion mapping, and a conviction-ranked long/short trading strategy.
The structural collapse detection system identified 113 out of 113 tokens that subsequently died, with 98% precision, validated out-of-sample from January 2024 through February 2026. A portable alpha strategy using these signals produced risk-adjusted returns between Sharpe 2.02 (Conservative) and 5.56 (Aggressive) out-of-sample.
Paper trading launched March 1, 2026, with SHA-256 hash-chained audit trail. Every signal is logged before market open. No hindsight. No adjustments.
2. Trust Score — Five Pillars
The Trust Score assesses relative quality and creditworthiness of crypto entities (tokens, exchanges, DeFi protocols) on a 0–100 scale, mapped to Moody's-style letter ratings (Aaa through D).
| Pillar | Weight | Measures |
|---|---|---|
| Security | 30% | Audits, hack history, contract risk, reserves, ATH recovery |
| Compliance | 25% | Regulatory status, social presence, supply transparency |
| Maintenance | 20% | GitHub activity, volume activity, development cadence |
| Popularity | 15% | Market cap rank, volume, community size |
| Ecosystem | 10% | Multi-chain presence, DeFi integration, category breadth |
Rating Scale
| Rating | Score Range | Grade |
|---|---|---|
| Aaa | 95–100 | Highest quality |
| Aa1–Aa3 | 80–95 | High quality |
| A1–A3 | 65–80 | Upper medium |
| Baa1–Baa3 | 50–65 | Investment grade floor |
| Ba1–Ba3 | 35–50 | Speculative |
| B1–B3 | 20–35 | Highly speculative |
| Caa–C | 5–20 | Substantial risk |
| D | 0–5 | In default |
3. Distance-to-Default (DtD) — Seven Signals
Adapted from Merton's structural credit model used in corporate bond analysis, DtD measures standard deviations from a default barrier. Scale 0–5 where 5 = safe, 0 = imminent collapse risk. DtD is recalculated daily from on-chain and market data.
| # | Signal | Weight | Measures |
|---|---|---|---|
| S1 | Liquidity Depth | 10% | Turnover, volume trend, volume stability |
| S2 | Holder Concentration | 5% | Whale activity, Gini coefficient, distribution |
| S3 | Ecosystem Resilience | 30% | Drawdown, volatility, momentum, acceleration |
| S4 | Fundamental Activity | 10% | Volume trend, panic detection, price/volume divergence |
| S5 | Contagion Exposure | 25% | BTC correlation, downside beta |
| S6 | Structural Risk | 5% | Flash crash frequency, spread, token age |
| S7 | Relative Weakness | 15% | Performance vs BTC (7d/14d/30d) |
Weight Calibration (v3.1)
Signal weights were calibrated through correlation analysis against historical crashes. Key adjustments: S1 (Liquidity) reduced from 20% to 10% — volume often increases during crashes, making it counterproductive as a distress signal. S5 (Contagion) increased from 10% to 25% — strongest single predictor with -1.14 mean differential between crash and stable tokens. S7 (Relative Weakness) added at 15% — captures rapid underperformance relative to BTC.
Alert Levels
| Level | DtD Threshold | Interpretation |
|---|---|---|
| SAFE | ≥ 4.0 | Low risk |
| WATCH | ≥ 3.0 | Elevated monitoring |
| WARNING | ≥ 2.0 | Active risk |
| CRITICAL | ≥ 1.0 | High distress |
| STRUCTURAL COLLAPSE | < 1.0 | Imminent failure risk |
4. Structural Collapse Detection
The STRUCTURAL COLLAPSE / STRUCTURAL STRESS alert system combines DtD signals with trend analysis to identify tokens at elevated risk of permanent failure. This is the core predictive system and has been validated with strict in-sample / out-of-sample separation.
Out-of-Sample Validation (January 2024 – February 2026)
Target Tracking
| Target | Definition | Hit Rate |
|---|---|---|
| 1st Target | -30% from alert price | 92% |
| 2nd Target | -50% from alert price | 65% |
5. Crash Probability Calibration
Crash probability estimates the likelihood of a ≥30% decline within 90 days, calibrated on 393 historical crash cycles with per-crisis-type betas achieving 3.4 percentage point median error.
| Trend | WARNING | CRITICAL | WATCH | SAFE |
|---|---|---|---|---|
| FREEFALL | 43% | 34% | 28% | — |
| FALLING | 37% | 30% | 25% | — |
| SLIDING | 33% | 28% | 20% | — |
| STABLE | 33% | 30% | 18% | 3% |
| IMPROVING | 20% | 18% | 12% | 2% |
Stresstest Engine
The portfolio stresstest uses the same calibrated model with per-crisis-type betas: contagion events (SOL -67.6% predicted vs -67.9% actual), liquidity crises, and stablecoin depegs. Median prediction error of 3.4 percentage points across historical scenarios. Evidence banners on the Risk Scanner and Contagion Map reference specific validated predictions.
6. Contagion Model
The contagion model maps dependency relationships between tokens through shared liquidity pools, ecosystem membership, bridge exposure, oracle dependencies, stablecoin reliance, and exchange concentration. Each token receives a Contagion Score (0–10) representing vulnerability to cascading failures.
Scenario Analysis
Three pre-built scenarios (major exchange collapse, stablecoin depeg, ecosystem contagion) plus custom portfolio input. Retroactive case studies validate the model against FTX, LUNA/UST, and 3AC — all using the same calibrated betas applied to current data.
7. Portable Alpha Strategy
The Portable Alpha strategy generates conviction-ranked long/short pairs using Trust Score spreads and DtD differentials. It operates within the investment-grade universe (A1–A3 rated tokens), applies bear market detection, and holds positions for 90-day cycles.
Strategy Rules
Universe: ~85 major tokens, excluding stablecoins. Rating class: IG_MID (A1–A3) — empirically the highest-alpha class. Pair selection: top-5 monthly by conviction score (40% composite spread + 60% DtD differential). Maximum 2 pairs per token. Bear detection: BTC monthly return < -15% triggers skip (Alpha) or defensive allocation (Dynamic/Conservative). Return cap: ±100% per leg.
Backtest Results (April 2021 – December 2025)
| Variant | Period | Return | Sharpe | Max DD | BTC Return | Alpha |
|---|---|---|---|---|---|---|
| Conservative | OOS (24mo) | +314.7% | 2.02 | -11.0% | +100.8% | +213.9% |
| Growth | OOS (24mo) | +842.2% | 3.86 | -12.1% | +100.8% | +741.4% |
| Aggressive | OOS (24mo) | +1,872.9% | 5.56 | -18.8% | +100.8% | +1,772.1% |
Allocation Variants
| Variant | Pairs Weight | BTC Weight | Cash Weight | Bear Adjustment |
|---|---|---|---|---|
| Alpha (Pure L/S) | 100% / 0% | — | — | Skip entirely |
| Dynamic | 30% (bear) / 40% | 10% (bear) / 40% | 60% (bear) / 20% | Defensive |
| Conservative | 20% (bear) / 30% | 5% (bear) / 30% | 75% (bear) / 40% | Defensive |
8. Retroactive Case Studies
FTX (November 2022)
Trust Score: 5.0/100 (Grade F). No proof of reserves, opaque corporate structure, circular collateralization through FTT token, and Alameda Research conflicts of interest. The score reflects systematic detection of structural red flags through publicly available data. Losses: $8.0B.
Terra/LUNA (May 2022)
Trust Score: 57.2/100 (Grade C+) for LUNA, 29.9/100 (Grade D) for UST. Algorithmic stablecoin with no real reserves, unsustainable 20% APY via Anchor Protocol, and known death-spiral risk mechanism. Combined losses: $58B.
Celsius Network (June 2022)
Trust Score: 5.0/100 (Grade F). Losses: $4.7B. Below-platform scoring across all five pillars.
9. Track Record
Backtest (April 2021 – December 2025)
60-month backtest period with strict IS/OOS separation at month 36. Full results available at /track-record. All signals are reproducible from the published methodology and publicly available market data.
Live Paper Trading (March 2026 →)
Three portfolios tracked live since March 1, 2026. All entries SHA-256 hash-chained in an append-only database. No UPDATE or DELETE operations permitted. Microsecond-precision timestamps. Signals logged before market open.
Current regime: BEAR (BTC -46.7% from 365-day ATH). Alpha Fund in bear skip (100% cash). Dynamic and Conservative running defensive allocations.
Dashboard: /paper-trading. Audit trail: GET /v1/crypto/paper-trading/audit.
10. Machine-First Architecture
ZARQ is built API-first. The dashboard is a showcase — the API is the product. Every data point visible on the website is available through a documented JSON endpoint. All endpoints are free during beta with a 1,000 calls/day rate limit.
API Surface (37 endpoints)
Ratings, DtD scores, early warnings, trading signals, safety checks, portfolio stresstest, contagion mapping, paper trading NAV, and audit trails. Full documentation at /docs.
AI Discoverability
Optimized for autonomous AI agent consumption: llms.txt, robots.txt with AI bot directives, Schema.org markup, AI-citable summary comments in HTML source, and MCP (Model Context Protocol) server integration. Bulk data available as JSONL under CC BY 4.0.
Infrastructure
Self-hosted on dedicated hardware. SQLite for crypto data (zero-overhead reads), PostgreSQL for agent data. Cloudflare Tunnel for TLS termination and CDN. Automated daily pipeline: data collection → scoring → signal generation → NAV calculation → audit logging. System health monitored every 5 minutes with automatic self-healing.