VOL. 07 / Synthetic demonstration

The Sealed Model

Allocation models arrive as a pie chart with no visible link between the assumptions and the answer, so nobody can ask what happens if a view is wrong. This one runs in your browser on assumptions written into the source, and publishes every one of them.

Illustrative system study

Quick read

Demonstration · synthetic data

Synthetic demonstration

The problem: An allocation you cannot interrogate. Invented assumptions, shown in full.
Who it is for: Anyone evaluating whether I can build a quantitative tool that a non-specialist can actually use. It is a demonstration of method.
What you receive: A solver you can argue with, and the assumptions table behind it
Evidence status: Demo evidenceComputed from synthetic inputs. Not an observed investment result.
Main limitation: Every input is invented. Nothing here is a market observation, a backtest, or a track record.

Discuss your project

Volume VII · Demonstration · synthetic data

The Sealed Model

An allocation you cannot interrogate. Invented assumptions, shown in full.

Synthetic demonstration

Built for this site on invented data to show a method. Not an engagement, not a client result, and not in use by anyone.

The problem

Allocation models are usually presented as an answer: a pie chart with no visible link between the assumptions and the output. Anyone being shown one has no way to ask what happens if a view is wrong.

Who it is for

Anyone evaluating whether I can build a quantitative tool that a non-specialist can actually use. It is a demonstration of method.

Context

Built for this site. It is a synthetic demonstration, not an engagement and not a product: no client commissioned it and no one is using it to allocate money.

My role

Designed and built entirely by me — the solver, the simulation and the interface.

Data provenance

None. Every input is an invented long-run assumption written into the source, not a market feed: eight asset classes with stylised expected returns, volatilities, a correlation matrix and benchmark weights. No live prices are fetched, and there is no market data anywhere in this tool.

Constraints

It has to run in a browser with no server, and it has to stay honest about being synthetic while still behaving like the real method.

Approach

Start from benchmark weights and reverse-engineer the equilibrium returns they imply, as Black–Litterman does.
Let the visitor state views as scenarios, blend them with the equilibrium prior at a confidence they control, and re-solve.
Run Monte Carlo paths from the resulting allocation to show the spread of outcomes rather than a single expected return.

Approach

Expose the assumptions and the formulas in an advanced view, so the numbers can be checked rather than trusted.

Decisions that mattered

Kept the assumptions visible and editable instead of hiding them behind the chart.: The point of the tool is that the output follows from stated inputs. Hiding them would make it decoration.
Show a distribution of outcomes, never a single projected return.: A single number invites being read as a forecast. It is not one.
Everything runs client-side.: No server means nothing a visitor types is transmitted or stored anywhere.

Figure

Synthetic illustrationInvented data, used to demonstrate a method. Not a real result.

Figure An allocation across eight asset classes shown against benchmark weights, beside a fan of simulated outcome paths widening over time. The fan shows a broad range of outcomes rather than a single line, with the median path in the middle and progressively wider bands around it.

Allocation and outcome spread produced by the optimizer from its built-in assumptions. Synthetic illustration — the inputs are invented and no market data is involved.

Baseline

Each allocation is shown against the benchmark weights it started from, so the effect of a view is visible as a difference.

How it was evaluated

The solver is covered by unit tests that check its mathematical properties — that weights sum to one, that a neutral view reproduces the benchmark, that the covariance matrix stays positive definite. There is no accuracy claim to evaluate, because there is nothing being predicted.

Results

No performance figure is published for this project, so none is claimed.

What was delivered

A client-side Black–Litterman solver with a Monte Carlo simulator.
Simple and advanced views over the same computation.
Regression tests over the financial logic.

What it changes in practice

It shows the working. The method is the same one used on real mandates; only the numbers are invented, and the tool says so on its face.

Limitations

Every input is invented. Nothing here is a market observation, a backtest, or a track record.
Not investment advice, and it produces no recommendation about any real security.
Long-run assumptions with no regime changes, no transaction costs and no taxes.

Human review

Not applicable — it is a demonstration, and its output is not used for any decision.

Interactive calculation

Synthetic illustrationComputed in your browser from the stated assumptions. No external data.

Have a problem shaped like this?

Tell me what you are trying to decide and I will tell you whether this approach fits. Related service: Forecasting & financial analytics.

Discuss your project

Nothing on this site is accounting, audit, tax, legal or investment advice, and no figure here is a promise of a result. Where a number appears, its source and its limits appear with it.