live · binance spot · ARM64 edge node · 68d running

Unbiased return, engineered.

AURA is a fully synchronous, parameter-driven trading engine for liquid crypto markets. It executes deterministically, gates risk before capital moves, and treats every live incident as a structural fix — not a manual override.

View live performance Architecture deep-dive

live Beating Gold over the current window

deterministic flow orbit · cadence · click any node to inspect

Live dashboard · last 68 sessions

At-a-glance.

Drill in →

Cumulative return

-4.03%

since 2026-03-06

TWR (annualized)

-19.56%

time-weighted · cash-flow-neutral

Max drawdown

-14.95%

peak-to-trough, intra-window

Sharpe (ann.)

-0.29

146 fills · rf = 0

Benchmark deltas · current live window

AURA return vs. equally-anchored passive holds of each asset.

2026-03-06 → 2026-05-14

Bitcoin (BTC)

trails

AURA

-4.03%

BTC

+17.17%

-21.19 pp

Gold (PAXG)

beats

AURA

-4.03%

Gold

-9.55%

+5.53 pp

S&P 500 (SPY)

trails

AURA

-4.03%

SPY

+11.16%

-15.19 pp

Two tracks

Research-grade backtest. Live deployment. No daylight between methodology and execution.

research · walk-forward 2025-11-14 → 2026-05-13

Backtest, post-cost · last validated slate

The walk-forward optimizer evaluates each parameter combination across multiple historical windows under successive-halving. Only knobs that beat both Gold and SPY survive promotion.

AURA total: +141.3%
Annualized: +497%
Sharpe: 3.64

vs. SPY: +130.8 pp
vs. Gold: +126.1 pp
Max DD: 14.95%

live · binance spot 2026-03-06 → 2026-05-14

Live execution · current window

Real capital, real fees, real slippage. The system absorbs deposits/withdrawals via time-weighted return, so performance %s never reflect cash flows — only investment skill.

Return: -4.03%
TWR (ann.): -19.56%
Sharpe: -0.29

vs. BTC: -21.2 pp
vs. Gold: +5.5 pp
vs. SPY: -15.2 pp

Backtests are run by a fully independent walk-forward optimizer. The same hot-reloadable parameter file feeds both research and live — there is no manual translation step.

What systematic execution actually requires

Three properties we never compromise on.

Benchmark-Beating Alpha

Every position is the deterministic output of a pre-committed parameter set, tuned against rolling historical windows. Walk-forward optimization only promotes knobs that beat both Gold and SPY across the validation slate.

Risk-Adjusted Returns

Three concentric gates — kill-switch, per-position SL/TP, portfolio-level cost-benefit — must clear before any capital moves. Annualized Sharpe and bounded drawdown sit at the heart of the score, not raw return.

Regime-Aware Allocation

A two-cycle regime hysteresis suppresses bull / sideways / bear flips. In low-conviction tape, capital is rotated defensively (tokenized gold / Bitcoin) under regime-specific utilization caps.

What this solves

A discretionary trader's failure modes, removed at the protocol level.

Emotional bias: Pre-committed thresholds executed by code, not a human at 3 a.m.
Execution latency: Single-threaded determinism with idempotent retries — no event-loop races, no double-fills.
Risk containment: Kill-switch on rolling drawdown plus per-position SL/TP. Losses are bounded, not improvised.
Parameter drift: Walk-forward optimization promotes only parameters that survive across multiple historical windows.