Approach

A method you can check

The important difference between a quantitative fund and a traditional one is not that a computer decides. It is that the decision is made before the opportunity appears, and that it is written down.

Principles

Four rules we do not break

01

Decide before the situation

Rules are written and tested before the market asks the question. When volatility arrives there is no room for improvisation, there is a procedure.

02

Test instead of conviction

An idea that cannot be falsified against data is not a strategy, it is an opinion. A bad hypothesis rejected early is cheaper than a good story.

03

Cost is part of the result

Commissions, spread and slippage are counted in every test. A strategy that only works at zero cost does not work.

04

Risk before return

First we set how much can be lost and how fast. Only then do we look at how much can be earned inside those bounds.

Architecture

The five layers of the process

Each layer has its own input, its own measure and its own condition for going further. A layer that fails its test stops the process.

  1. 01

    Data

    We work with point-in-time data, the version that was known on the day. The model cannot see the future, because that is the most common reason a strategy that looks brilliant in a test loses money on a live account.

    Input
    Prices, fundamentals and macro data, point-in-time
    Condition to pass
    No access to anything unknown on the day of the decision
  2. 02

    Research

    Hypothesis, out-of-sample test, walk-forward, stress test. A strategy that does not survive that path does not enter the portfolio, however good the chart looks.

    Input
    A hypothesis about market behaviour
    Condition to pass
    The result holds out of sample and under stress
  3. 03

    Portfolio construction

    Concentration limits, diversification, and transaction cost priced into the decision from the start rather than added at the end. The portfolio is built to be durable, not impressive.

    Input
    Signals that passed validation
    Condition to pass
    The portfolio sits inside concentration and risk limits
  4. 04

    Execution

    Orders are placed automatically and measured. We report the gap between the assumed price and the one we got, because that is a real cost that eats returns quietly.

    Input
    The target portfolio
    Condition to pass
    Execution cost stays within the test's assumption
  5. 05

    Risk

    A layer independent of the strategy, with hard limits and daily measurement. Risk is a parameter agreed with the investor before launch, not a side effect.

    Input
    Positions and results from every day
    Condition to pass
    No limit is breached
Risk

What we do when the model stops working

Every quantitative strategy eventually loses its edge. The advantage the data found fades once someone else finds it too. The question is not whether, but how quickly we notice.

01

Hard limits

Maximum exposure, concentration and permitted drawdown are written as numbers and enforced automatically, with no discretionary step in between.

02

Drift monitoring

Every day we compare the strategy's behaviour against its profile from testing. A deviation outside the expected range triggers a review, not a debate.

03

Retiring a strategy

A strategy that behaves off-model for a set period is switched off, following a procedure written in advance.

04

The capacity ceiling

We take capital only up to the level at which the strategy still works. Once it is reached, the fund closes to new subscriptions.

Transparency

What the investor actually receives

The scope and frequency of reporting is one of the survey questions. What follows is what we treat as the minimum.

  • A monthly report with valuation, exposure and risk limit usage
  • A quarterly commentary on what worked, what did not, and why
  • A full cost statement in one place, not split into lines that are hard to compare
  • Notice of every material change to the model or to risk parameters
  • Annual financial statements audited independently

Have a view on how this should look?

The survey asks about exactly these parameters: liquidity, reporting, drawdown tolerance and fee structure.

Take the survey