How SigmaX Proves What It Claims.
SigmaX is built so that nothing can be misrepresented. Every claim carries an evidence class. Performance is reported net of cost or not at all. Green is earned through a gate — never assigned. This page is the canonical explainer for how that discipline works.
AI analyzes. Deterministic systems validate. Humans authorize. Capital remains governed.
Five Levels. Each Earned, Not Assigned.
Every strategy and claim carries an Evidence Level. A node cannot move up a level without passing the gate that defines the next one. Nothing skips ahead.
- EL1In-sample / backtest
A hypothesis tested against the data it was developed on. EL1 is the weakest evidence: it can describe an idea but cannot confirm it generalizes. Backtest results are labeled as such and never presented as live performance.
- EL2Out-of-sample tested
The hypothesis is tested on data it has not seen. Out-of-sample survival is the first gate where an idea earns the right to advance. Failure here is preserved as evidence, not deleted.
- EL3Paper-live / shadow observed
The strategy runs in paper or shadow mode against live conditions with no capital at risk. Fills, costs, and slippage are observed rather than assumed.
- EL4Internal-capital live
Operator or company capital only — never outside or customer capital. EL4 carries the full net-performance, drawdown, capacity, and kill-rule disclosure and a documented proof packet.
- EL5Customer access reviewCounsel-reviewed track
Reserved for counsel-approved customer-eligible structures. A strategy can reach this level only after legal pathway, eligibility, compliance review, and operator approval are explicitly implemented and verified.
Performance Is Reported Net — or Not at All.
Gross numbers describe a fantasy. SigmaX reports the number a real participant would have kept. Performance is never shown gross-only, and it is never invented.
Every cost component is modeled with realism — fee schedules, fill assumptions, and slippage are stated, not waved away. If a cost cannot yet be measured, the result is labeled pending or blocked rather than reported as if it were complete.
The Demo Board Ships Green: 0.
An empty success column is a feature. A green result means an out-of-sample gate was earned — and until one is, the honest count is zero. A demo board that showed manufactured wins would be the failure. The illustrative board below shows the failures on purpose.
Zero validated wins until an out-of-sample gate is passed. This is the honest default, not a placeholder waiting to be filled with optimism.
Items that have cleared their stated gate in this demo board — shown for shape only.
Currently under test. Not yet promotable. Outcome unknown until the gate resolves.
Refuted or fell short out-of-sample. Failures are preserved on the board, not hidden.
Cannot advance without an input the platform does not yet have — data, approval, or infra.
Sample, Demo, and Not Connected.
Illustrative figures used to show how a surface behaves. Sample data carries the “Illustrative sample — not client data” marker and is never upgraded to real evidence. The current web surface is demo-only with live connectors disabled.
No live data source is wired to this surface. “Not connected” is the honest default — SigmaX shows it rather than fabricating numbers to fill the space. A value appears only when a real, governed source is present.
What Atlas Can — and Cannot — Do.
Atlas is an analysis assistant. It works on evidence and prepares decisions; it does not make them and holds no capital authority.
- Explain the evidence behind a claim and its evidence class.
- Summarize a thesis and its supporting and opposing nodes.
- Draft a Briefing for human review.
- Forecast with explicit labels, assumptions, and confidence.
- Surface contradictions and counter-evidence.
- Prepare a Decision Record for a human to authorize.
- Trade.
- Move money.
- Authorize capital.
- Publish without approval.
- Guarantee an outcome.
- Replace human review.
Most Risk Comes from Other Agents.
Most risk does not come from static numbers — it comes from other agents responding. SigmaX models incentives, beliefs, counterparty behavior, and venue behavior before treating a claim as decision-ready. Every model shows its assumptions, solution concept, sensitivity, evidence, and limitations.
Read-only analysis. No capital authority. No execution.
What the model takes as given, stated explicitly.
The equilibrium or reasoning standard used to resolve the interaction.
How conclusions move when assumptions move.
The evidence class supporting each input and conclusion.
What the model does not capture and where it can be wrong.
SigmaX provides capital decision intelligence, scenario analysis, and authority workflow software. SigmaX does not provide investment advice. SigmaX does not place trades, move funds, route customer capital, or manage client capital. Shadow/simulated/paper/backtested results are not live trading results. No result is a guarantee of future performance. Human approval is required before any capital-linked action.
Where SigmaX uses AI, it is governed as assistive software: truthful capability claims, human oversight, traceable records, disclosure of AI assistance, and no capital authority. SigmaX designs AI controls with reference to applicable securities obligations, FINRA existing-rule guidance, SEC AI-claims discipline, NIST AI RMF-style risk management, and EU AI Act transparency readiness. This is compliance-readiness design, not a claim that every deployment or jurisdiction-specific use has completed legal review.