MORTIMER: Multi-asset Optimisation and Risk Targeting with Integrated Monthly E6 Rebalancing

MORTIMER is a six-asset equal-risk-contribution (ERC) strategy that dynamically scales exposure to forecast volatility, delivering systematic downside-risk mitigation with low return drag after costs.

Hypothesis

MORTIMER: Multi-asset Optimisation and Risk Targeting with Integrated Monthly E6 Rebalancing is a portfolio of six exchange-traded funds (ETFs) allocated by equal-risk contribution (ERC), with causal volatility scaling. The hypothesis is that persistent volatility permits lower portfolio risk after trading costs; see the research specification.

Data

Use QQQ, IWM, HYG, TLT, GLD, and DBC adjusted daily prices from Yahoo, with initial-release ALFRED (Archival Federal Reserve Economic Data) DGS3MO observations for cash accrual. Restore the hash-bound inputs into ignored data/:

uv run --env-file .env --locked python tools/snowflake_store.py fetch --manifest research/archive/erc-control/erc-market-manifest.json --source data --path 'cache/erc/*'
uv run --env-file .env --locked python tools/snowflake_store.py fetch --manifest research/archive/erc-control/erc-market-manifest.json --source data --path 'cache/erc-final/*'

Methodology

Estimate monthly constrained ERC weights from 126 sessions of Ledoit–Wolf covariance. Target 10% annual volatility using the larger of two exponentially weighted moving average (EWMA) forecasts, with spans of 20 and 63 sessions. Cap gross exposure at one, and rebalance exposure at a five-percentage-point change. Execute close-derived orders at the following close and charge five basis points on risky purchases and sales; run the ten-basis-point cost stress on development data.

Results

Net metric	Training	Validation	Out-of-sample (OOS) test
Annualised return	5.85%	7.56%	9.52%
Annualised volatility	6.37%	7.93%	7.47%
Excess-return Sharpe	0.859	0.660	0.714
Maximum drawdown	−12.36%	−16.63%	−7.00%
Annual half-L1 turnover	0.958	1.569	1.119

The held-out out-of-sample period spans 2 October 2024 through 2 October 2026. Results are reported separately for training, validation, and out-of-sample test.

Risk and capacity

Use a 35% base weight limit per fund, long-only positions, and gross exposure at most one. Volatility scaling reduces exposure when forecast risk rises; it does not impose a drawdown stop. Evaluate participation against lagged dollar volume and apply square-root impact scenarios before selecting capital.

Reproduction

uv sync --locked
uv run --locked python tools/reproduce_mortimer.py
uv run --locked jupyter lab

Run the research notebook after restoring the inputs. The reproduction command writes metrics, returns, exposures, and provenance to ignored results/mortimer-submission/ without acquiring data or selecting parameters. Configure Snowflake credentials in ignored .env using .env.example; see shared assets.

Repository

Path	Contents
notebooks/	Reproducible MORTIMER research notebook
src/	Frozen MORTIMER strategy implementation
tools/	Acquisition, reproduction, and reporting commands
data/	Ignored observed inputs, reference material, and caches
research/	Specifications, provenance, frozen source, and trial ledger
results/	Ignored generated evidence and build outputs
report/	Quant note source, included assets, and quant-note.pdf

References

Read the quant note for methodology, evidence, and source references. Volatility targeting and equal-risk contribution are established methods; MORTIMER combines them under a fixed universe, bounded exposure, and explicit execution accounting.

Contribute

Follow CONTRIBUTING.md and AGENTS.md. Keep additional worktrees outside this repository, inputs in ignored data/, generated evidence in results/, and credentials in ignored .local/credentials/ or .env.

Licence

Project code is available under the MIT Licence. Market data and cited third-party materials retain their source terms.