◆System map
SuperEzio research / Evidence
Technical evidence is published in English.
An educational research system for learning how market data becomes a qualified decision. Its 14 mathematical modules report what worked, lacked data or failed on each run; six expert roles provide structured perspectives. The diagram below shows the real boundaries. This is support for human analysis, with no verified investment returns or autonomous trading claim.
The local cockpit collects dated market data and fundamentals, calculates risk and scenario measures, obtains six specialist views and lets Bob explain the evidence and its limits. The operator reviews any action. TD orders remain manual; a gated IBKR simulation path is separate from real-money execution, which is frozen.
MathHub registers 14 modules in src/core/math/math_hub.py. Its coverage object counts successful, skipped or insufficient-data, and error results for every analysis. Registration is not proof that all 14 produced a valid number in any given question. Tests validate calculations and selected edge cases with local fixtures.
| Module | What it calculates or checks | Evidence and condition |
|---|---|---|
| price_series | Price series, latest price and dated candles | MathHub; needs usable prices |
| technical_indicators | SMA, EMA, RSI, MACD, bands and volume measures | test_math.py; needs price/volume history |
| return_metrics | Observed and annualized return, volatility | test_math.py; needs at least two closes |
| sharpe_sortino | Risk-adjusted return estimates | test_math.py; needs return series |
| drawdown | Largest decline on the observed curve | test_math.py; needs price history |
| var_cvar | Historical loss quantile and tail average | test_math.py; depends on observed returns |
| kelly | Sizing estimate from wins and losses | test_math.py; skipped without both outcomes |
| dcf_proxy | Cash-flow valuation proxy | test_math_hub.py; skipped without cash flow or shares |
| crisis_aggregate | Aggregated crisis signals | test_crisis.py; reports input and status limits |
| return_bootstrap | Scenarios resampled from observed returns | test_math_hub.py; needs 10 observed returns |
| portfolio_hrp | Hierarchical risk-parity allocation | test_portfolio.py; requires return histories |
| f51_unified | Decision score, when supplied after expert stage | test_math_hub.py; skipped at pre-expert stage |
| penny_momentum | Momentum screen for small-cap series | test_math_hub.py; needs ticker and 20 dated candles |
| trading_losses | Expected versus actual return losses | tests/contracts/test_trading_losses.py; skipped without both series |
The six expert roles return structured outputs. The committee status now distinguishes complete, partial and failed participation. F51, neural assessment, SDCA and guardrails add constraints, but some remain observational; their presence is not evidence of a final stop in every path. The official TD list was compared with the local 108-symbol manifest on 4 October 2026; a broker order preview remains decisive for eligibility and fees.
A local automated full-suite run after the maintenance fixes passed; it tests code behaviour, not a live broker transaction or financial outcome. A separate authenticated end-to-end demonstration with external model and market-data services remains to be recorded.
Official list checked against the 108-symbol local manifest on 4 October 2026. Eligibility and fees require confirmation in the TD order preview.