⚖ MIZANميزان
The verification standard for quant alpha — prove the edge, never reveal the strategy.
ATLAS VERIFICATION CREDENTIAL
PROOF OF EDGE · v1
ISSUED 2026-07-27
Verdict
Did not clear the gate.
ATLAS-DEMO-0001 · SHA e8d650…
⚠ Rendered from a DEV-MODE receipt — illustrative only, NOT a cryptographic proof. Re-run without RISC0_DEV_MODE for a real credential.
Submissionwho · what · when — logic never disclosed
Strategy alias
DELISTED-NAME-VALIDATION (logic sealed)
Asset · Venue
BTC-USD · Binance daily
Period tested
1997-12-31 → 2004-11-17
Bars
1731 daily · point-in-time
Strategy commitment
0x7fe842cd…d7c9cfd5 · sealed
Verified Performancecommitted in-proof · after-cost · point-in-time
Sharpe (annualized)
-0.25
OOS -0.65 · gate ≥ 1.20
FAILNet return (period)
-80.5%
after 10 bps/trade
FAILMax drawdown
94.1%
path-honest (daily MTM)
FAILCAGR · Calmar
-21%
Calmar -0.2 · annualized*
* CAGR/Calmar derived from the committed net-return + period (re-computable by anyone); annualized from a 82.4-month window.
The Gate — Institutional Limitsenforced in-circuit
| Criterion | Standard | Verified | Result |
|---|
| Sharpe ratio | ≥ 1.20 | -0.25 | FAIL |
| CAGR (derived) | ≥ 20% | -21.2% | FAIL |
| Max drawdown | ≤ 25% | 94.1% | FAIL |
| Worst single-event loss | ≤ 12% | 36.5% | FAIL |
| Walk-forward (OOS Sharpe holds) | ≥ 0.50 | -0.65 OOS | FAIL |
| Anti-degeneracy (trades · exposure) | ≥ 3 · ≥ 20 | 9 · 672 | PASS |
Bias Eliminationthe four sources of fake edge
✓Execution realism
Modelled fills, spread & slippage (10 bps/trade). No phantom fills.
✓Temporal integrity
Strict point-in-time signals. No same-bar / future-data leakage — enforced in-circuit.
✓Survivorship integrity
Single liquid asset (BTC) — no delisting universe to bias by construction.
✓Path-honest risk
Drawdown measured on the full daily equity path; strategy is unlevered (no intraday liquidation risk).
Independence Scorenovel alpha, or beta in a costume?
Market beta (R²)0.13
Momentum factor (R²)0.25
Residual (novel α)0.70
Independence = 1 − R²(strategy vs market+momentum), computed by an integer 2-factor regression INSIDE the proof (re-verifiable). A trend strategy is partly factor beta — this is the honest residual, not an inflated figure.
Zero-Knowledge Receipt · RISC Zero succinct STARKDEV MODE · NOT A REAL PROOF
Data root (Merkle)0xbd5fe3d5…148a4936
Strategy commitment0x7fe842cd…d7c9cfd5
Image ID (circuit)0x3ac3b10b…5214237b
Proof size · verify time2 KB (succinct, constant) · 1 ms
Issued · expires2026-07-27 · perpetual (re-verifiable)
🔒 The strategy logic (indicators, windows, thresholds, parameters) was never transmitted, never stored, and is provably absent from this credential. The proof attests the returns are real without revealing how they were produced.
✓ What the allocator learns
✓ Net return, Sharpe, drawdown, Calmar✓ That it cleared every institutional gate, incl. walk-forward✓ That biases were eliminated & the data was point-in-time✓ That the alpha is 1% independent of market+momentum🔒 What stays the quant's secret
🔒 The indicators and signal logic🔒 The entry / exit rules🔒 Window lengths & thresholds🔒 Position-sizing & the source code — entirelyWhat this credential proves — and what it does notread both columns · honesty is the guarantee
✅ GUARANTEED
• The strategy logic never left the prover's machine — it is provably absent from this credential.
• The same risk gate applies to everyone; a losing strategy gets no credential.
• The backtest ran on a fixed, Merkle-bound dataset — no cherry-picking history.
• Re-verifiable by anyone with zero trust; they still never see the strategy or exact metrics.
• Integer-exact math, no hidden lookahead.
⚠️ NOT CLAIMED
• Attests a backtest, not future results — capital is at risk.
• Soundness rests on the RISC Zero zkVM; an external paid ZK audit is pending.
• Seal witness-hiding is being confirmed — we claim "not in the credential," not yet "mathematically hidden."
• Data provenance (signed feed) is governance, on the roadmap.
• A working proof-of-concept — not labelled audited/production-ready.
Re-verify this credential yourself — trust no one.
Anyone can re-check the proof against the committed data root: cargo run -p verifier. No trust in MIZAN or the quant required.
Disclosures. Verified performance is computed under ATLAS's point-in-time, after-cost methodology over the stated historical period; it is not a guarantee of future results, and all strategy capital is at risk. Every figure above (returns, Sharpe, drawdown, walk-forward OOS Sharpe, and the independence score) is computed INSIDE the zero-knowledge proof and is re-verifiable. This is a proof-of-concept: the cryptographic ZK layer is not yet independently audited. Independence is measured against a market+momentum factor pair. Not investment advice or an offer.