Skip to main content

Checking access...

Portfolio Strategy

ARIA

Walk-forward

Adaptive Risk Investment Allocation

Rebalancing
Monthly
Universe
5 ETFs
As of
...
--
Annualized Return
--
Max Drawdown
SPY: -33.7%
--
Sharpe Ratio
SPY: 0.89
--
Avg. Exposure
Rest in cash
--
Current Position

Backtest 2010 to present. Walk-forward test active since April 1, 2026.

The Strategy

ARIA (Adaptive Risk Investment Allocation) applies risk management independently across four equity segments: US large cap, US technology, US small cap, and international developed markets. Each segment is monitored using its own set of macroeconomic indicators. When conditions for a given segment deteriorate, that position exits to cash while the others remain invested.

The result is a portfolio that scales down total equity exposure as aggregate risk rises and scales back up when conditions are favorable. This granular approach can hold one or two segments while exiting others, rather than making an all-or-nothing equity call.

Walk-Forward Test Active Starting April 1, 2026, this strategy is being tested with market data in a walk-forward framework. All historical data before this date represents backtested performance.

How It Works

At each month-end, the model evaluates each of the four ETFs (SPY, QQQ, IWM, EFA) against macroeconomic data. Positions are either fully invested or fully in cash on a per-segment basis. When invested, each active segment receives equal weight in the portfolio.

The dashboard displays both backtested and walk-forward performance. The shaded area marks April 1, 2026, the start of the walk-forward test. All data to the left represents historical backtesting; all data to the right represents hypothetical out-of-sample performance.

4
Equity ETFs
Binary
Long / Flat per Segment
Monthly
Rebalancing

Simulation Settings

Set your start date and initial capital. ARIA applies independent risk management across four equity segments with monthly rebalancing.

Model Performance

Drawdown

Peak-to-trough decline

Equity Exposure

Total equity allocation over time

Current Positions

SPY
S&P 500
--
QQQ
Nasdaq 100
--
IWM
Russell 2000
--
EFA
Int'l Developed
--
Model Confidence Score
--

Walk-Forward Test

-- days live
--
Model Return
--
SPY B&H
--
vs Benchmark
--
Data Points

ARIA vs SPY Buy & Hold

Full backtest period comparison:

Metric ARIA SPY B&H
Total Return----
CAGR----
Volatility----
Sharpe Ratio----
Max Drawdown----
Exposure--100%

Crisis Performance

Strategy behavior during major market downturns:

Crisis ARIA SPY
2020 Covid---34%
2022 Bear---25%

Detailed Metrics

Model Value
--
--
YTD Return
--
SPY: --
Volatility
--
Sharpe Ratio
--
Sortino Ratio
--
Max Drawdown
--
Calmar Ratio
--
Beta vs SPY
--
VaR (95%)
--
Exposure
--
Up Capture
--
Down Capture
--

Statistical Validation

Before deployment, ARIA was subjected to a rigorous multi-phase validation framework based on peer-reviewed statistical methods. All tests were conducted on the fixed production strategy without post-hoc parameter tuning.

Permutation Test passed
p < 0.001

10,000 random position sequences with identical average market exposure were compared against the strategy. Only 1 in 10,000 matched the observed risk-adjusted return. Based on White (2000).

Block Bootstrap passed
CI: 1.10 – 2.13

10,000 block bootstrap samples (6-month blocks) produce a 95% confidence interval for the Sharpe ratio that lies entirely above zero. Based on Politis & Romano (1994).

Cross-Validation passed
100%

Combinatorial Purged Cross-Validation with 12-month embargo across 230 paths. Every single path produced a positive out-of-sample Sharpe ratio (median: 1.60).

Benchmark Attribution passed
6 / 6

Statistically significant alpha (t > 2.0) against all six benchmarks tested: buy-and-hold, balanced 60/40, equal weight, inverse volatility, momentum, and trend following.

Tail Risk passed
−4.7%

Conditional Value at Risk (95th percentile) of the strategy is −4.7% per month, compared to −8.6% for the SPY benchmark. Worst observed month since 2010 was −7.4%.

Data Integrity verified
No revisions

All underlying data sources are market-based and final upon release. Revision-prone series were systematically excluded. Publication lag is applied to ensure no look-ahead bias.

Validation framework based on White (2000), Politis & Romano (1994), and Bailey & Lopez de Prado (2014). All tests were conducted on the fixed production strategy without post-hoc parameter tuning. Past statistical performance does not guarantee future results.