Executive summary

A backtesting is a static snapshot of what could best have been done in the past. Such results would only have value if you believe the future will more or less match the past. Backtesting requires a far from negligible number of choices, and often ends up in an overfit of the data. The performance you observe as a result of backtesting is probably the best you could have achieved in the past, had you known the future in advance.

Humans manage the unknown by applying trial and error, adjusting their opinions as experience from new evidence accumulates. What if we could build an Artificial Intelligence to replicate human learning principles without the biases nor limitations?

Encelade Ltd is a Swiss company founded in 2008. The Company’s activities result in the application of state of the art research in statistics. As a company formerly active in the field of Alternative Asset Management, Encelade has carried out investments through careful automatized and systematic implementation of short term quantitative trading strategies.

Following a period of thoughtful research, we developed a sequential allocation decision tool, AGNOSTIC, which is model-free, parameter-free and able to:

  • Profit from opportunities early, well before even the best statistical tools could identify them.
  • Deal with a large set of possible choices, without falling into the trap of overfitting.
  • Avoid the corollary binary 0/1 decisions resulting from frequentist statistical testing.
  • Narrowly match the performance of the best available choice(s), i.e. the performance you usually get in your backtesting and never obtain in the future.
  • Dispense with assuming anything about the choices, i.e. be oblivious to the characteristics and detailed implementation of each particular choice.
  • Only consider strategies’ realized performance.

AGNOSTIC can be used as a tool to come close to the best:

  • trading strategy, among a selection of trading strategies;
  • asset allocation model, among a set of acceptable ones;
  • hedge funds to invest into, from a long-list of approved hedge funds;
  • factors / risk premia / alternative beta, from a broader choice of implementable ones;
  • parameters, among all possible combinations of parameters, for a particular strategy;
  • portfolio construction methodology, among all relevant ones, like mean variance, risk parity, etc.

You may think of AGNOSTIC as a tool that will carry out a reality check at each time step. Starting from day one, updating its trust in each available choice by encompassing relevant history. In doing so, AGNOSTIC will generate a portfolio whose updated weights will help to match, as closely as possible, the performance of the best choice(s), only to be known with the benefit of hindsight.

AGNOSTIC does not rely on any assumption or choice of parameter. The algorithm is the exact same for all clients, whatever the data: asset classes, trading strategies, hedge funds, etc. It does not assume any particular behaviour for the investment, e.g. momentum or mean-reversion.

To cover the needs of finance practitioners, AGNOSTIC can handle missing data, investments with a deferred starting date or different ending date, i.e. investments with different history lengths are handled in a consistent way.

A typical use case for AGNOSTIC is related to asset allocation decisions. Naturally, a promise to keep up with the performance of the best possible asset allocation is tempting. Below we present the result obtained from a diversified set of potential asset classes for a Swiss institutional investor, including investments in equities, bonds, real estate, commodities and hedge funds, together with portfolio constraints. The gray shaded area represents the extent of possible performance that can be achieved without constraints, and the turquoise area is the extent of possible performance achievable accounting for the constraints. AGNOSTIC’s performance is given in red.

AGNOSTIC is available as Software as a Service (SaaS) with access through a web interface for manual handling or through an API for automated processes. Encrypted communications with the server ensure total confidentiality of your data. A plug-in to handle portfolio constraints is available as an option.