Backtest with forward testing

SmartRetailTradingRobot
26 Apr 2023, 10:52

How would a forward-test look like? do you mean something like a monte carlo simulation?

jim.tollan
04 May 2023, 12:47 ( Updated at: 08 May 2023, 17:33 )

OMG - I literally navigated to the suggestions section to suggest this very idea.

Here's how i see it working:

1. Define the DateRange of the optimisation run

2. Define a sliding window (percentage) of which data is used for the backtest and which data is used for forward testing - i.e. 80/20, 90/10 etc

Having a combined *window* on the data that accommodated backtesting with out of sample data would truly be a gamechanger.

-----------------

FYI - I currently do this in a number of ways:

1. Select a daterange (say 01/01/2017-01/05/2023)

2. Set the Optimization daterange from 01/01/2017-30/04/2022

3. set the Backtest daterange to 01/05/2022-01/05/2023

4. Run the optimisation

5. From the optimisation, select the candidate set and apply it to the backtest window

6. rinse and repeat

----------------

It's quite cumbersome, but works. Combining this behaviour into the optimisation window would be amazing and offer huge insights.

Hope you get the gist. Still revelling at the weirdness of the timing of this request