Quantitative Finance / Independent exploration
RiskLab
Volatility-targeted momentum and tail-risk experiments with explicit turnover costs and a reproducible subset of offline tests.
Problem
Risk targeting changes both market exposure and trading costs. RiskLab provides a small, inspectable research setting for studying that interaction.
Architecture
Python modules separate signals, volatility scaling, backtesting, metrics, tail estimation, and stress testing. A separate C++ Monte Carlo extension is present; it was not loaded or benchmarked in this review.
Implementation
The backtester multiplies supplied signal and volatility exposure, computes turnover from position differences, and deducts basis-point costs. Its contract expects signals and scaling to be shifted before entry.
The inspected accounting computes , where is the basis-point cost converted to a fraction. Neutral-date filtering limits how this expression maps to the full trading calendar.
Engineering decisions
Keeping the accounting function small makes cost arithmetic testable. Causality is a pipeline property: a prefix-stability test of one signal cannot certify all preprocessing and exposure calculations.
Results
VERIFIED: 13 existing tests passed in an isolated copy on 9 November 2026, covering backtest columns/accounting, signal prefix stability, volatility, and metrics. This is a selected offline subset, not a full research reproduction or C++ speed benchmark.
Limitations
The backtester removes neutral-signal dates before calculating turnover and wealth. Exit costs and calendar-consistent benchmark comparison therefore need review. The tests that passed do not establish profitable out-of-sample performance.
Reproducibility
Reviewed at commit 971422295d94 on 2026-11-09. Source inspection establishes implementation structure; repository-authored results are not independent measurements. Four isolated accounting, signal, volatility, and metrics test modules passed (13 tests). The full repository and performance benchmarks were not rerun.
Next investigation
Resolve the limitations above, preserve the reviewed commit and exact experiment configuration, and reproduce correctness checks before promoting performance or deployment claims.