Independent research
Quantitative finance · Research tooling · 2026 · Researcher and product architectA portfolio laboratory that keeps model risk visible
InsightThe useful output is not a single optimal allocation. It is a way to see which assumptions produced it — and how quickly it changes when the benchmark, factor model or risk objective moves.
Try it
Move an assumption
Choose an objective, then alter benchmark pull and diversification. The allocation moves because the answer is conditional, not absolute.
Portfolio laboratory
One universe. Three different questions.
Return against risk. The cyan frontier marks the portfolios producing the most return per unit of volatility.
How tightly every candidate stays anchored to the selected market portfolio.
Higher values distribute risk across more independent holdings and exposures.
- Return
- 12.4%
- Volatility
- 11.3%
- Sharpe
- 1.09
- Active share
- 69%
The frontier moves as benchmark gravity and diversification change the feasible set. Demonstration only; no market data or investment recommendation.
01
Context
A continuing set of browser-based research tools — Alpha-Forge, AlphaMetric Core and FactorLab — built to move portfolio theory out of a notebook and into an interface where its assumptions can be inspected.
02
Problem
Portfolio optimisers are very good at producing precise weights from uncertain inputs. Presented as one answer, that precision can conceal sensitivity to expected returns, covariance, benchmark choice, frequency and the risk-free rate.
03
Insight
The model becomes more useful when the interface preserves disagreement. A maximum-Sharpe portfolio, a minimum-variance portfolio, an equal-weight baseline and a factor-aware allocation should be comparable in the same frame rather than collapsed into one recommendation.
Benchmark selection is itself an assumption. FactorLab therefore supports both a named market benchmark and an internally adaptive universe, then exposes market, size and value factor contributions alongside specific risk.
04
Strategy
Build a family of focused tools rather than a single opaque terminal. Each one starts with a different question — asset selection, portfolio efficiency or factor attribution — but shares the same demand for visible inputs, interpretable diagnostics and reversible choices.
05
System
Alpha-Forge calculates covariance, asset beta and alpha, correlation structure, tracking error, information ratio and portfolio R², then compares maximum-Sharpe, minimum-variance and equal-weight allocations.
AlphaMetric Core runs Monte Carlo portfolios and a deterministic solver side by side, with backtests that show whether the solver found a meaningful efficiency gain or only false precision. FactorLab constructs market, SMB and HML factors from the supplied universe, estimates factor betas and separates factor from specific variance before solving the allocation.
06
Outcome
Five working research tools now form a compact portfolio-construction lab. The value is not a claim to predict markets; it is a disciplined environment for asking how much of an allocation survives a change in assumptions.