User guide: Setting up a backtest

Summary

Conduct a backtest via "Build new portfolio"

About

The backtest page is available once a new portfolio has been correctly configured. A backtest runs a historical simulation of the portfolio based on the portfolio configuration.

Sequence of steps

1. Users must first enter a name e.g. "My Test Portfolio"
2. Users must enter a start date, backtest since 2015 are supported by default.
3. Users must specify a rebalancing frequency.
   1. By default this value is set to Quarterly.
   2. By opening the gear icon we arrive at a further configuration menu.
   3. In particular we see at the bottom of the configuration menu, the dates of the next three rebalances, e.g. 1 April 2015, 1 July 2015, 1 October 2015
   4. We can also set specific months for rebalances, specific days within those months, specific weekdays, specific weeks (in a month), as well as apply any offsets.
   5. We can also specify what happens when the dates fall on holiday, whereby we can rebalance on the next business day or the previous business day.
4. Users can further specify a starting NAV which represent the envisaged starting size of the portfolio. The units are specified in the currency of the portfolio itself.
5. Users can further set a broker fee, the default value is set to 1 basis point (i.e. 1/100th of a percent). We recommend setting this value to at least 10.
6. Users can finally see a validation box which provides both suggestions on the configuration and whether or not the configuration is valid.
7. Users can finally begin the backtest and will receive a notification once it completes.

Technical details

The backtest is constructed to be as realistic as possible. In terms of the underlying models and data, there is no "look-ahead bias" – all the forecasts and models are constructed using only data that was available at each rebalance date in the past. As such there is no "information leakage" or so-called time-travelling.

Additionally the following assumptions are made:

• trading at daily close prices
• 10 basis point transaction costs
• Market impact model to account for slippage from large illiquid trades
• 1 business day implementation/execution delay
• No dealing in tiny or fractional shares
• Proper simulation of daily cash reconciliation