Add data range parts (OOS, IS) module into Automatic Retest task in Custom Project

Voted for this task.

it is a feature request, not a bug, and it works differently. Automatic retest uses IS/OOS settings from the given strategy, and adds newly added data to the OOS part.

I don't see reason why to allow customizing IS/OOS settings in automatic retests.

Hi Mark,
all strategies have own parameters given like TF, MM, Tradiong options etc., and all can be changed in Automatic retest. Except IS/OOS settings. I think switch for chose between Manual/From_strategy settings is perfect and Automatic retest in generel is most valuable thing in SQX.

Our use case is simple. It's well know issue strategies from SQ often lose theirs performance in future. We have literally hundred thousands of strategies from few years of using SQ and one of biggest issue is determine which strategies are still OK. Strategy on picture is made in 2017. It would be great if we can compare (avg. Profit per year, or RDD) original data (2003-2017) as IS with current data (2017-now) as OSS in Automatic retest.

but if you are using automatic retest on the newest data it will add the newest data as OOS, so i think it works how you need

the task stays refused. we will not change it, it works exactly how you want it - new data are added to OOS part.

IMHO not. In AutomaticRetest you can swich between From_strategy, which is last one retest of strategy, or manualy set up data range, which is concider allways as IS. As we have all strategies stored retested to all data.

Anyways this wouldn't solve our issue. Maybe I wrote it not so correct as easiest example of loosing performance of AOS phenomenon. We need compare performance od strategies made sometimes 2015-2018 with actual markets data, from as example 2019-Now (let's say at least 50% avg.YearProfit, RDD etc.) and original strategies has OSS started around 2015. So with feedom of data ranges of IS/OOS. We store old strategies retested to all data available in those years so as example 5.5.2003 - X.X.2018 so just IS. Or even retested to only 1 day long data range to minimize strategies size, because are fastest for loading and retest for new data range is fast anyway.
This is also only one options which cannot be use as in normal retest modul. Especially if comparing IS/OOS is fundamental feature of SQ.

vlada is right, IS/OOS distribution could be also in automatic retest task, i could be very usefull for every kind of tests

Voted for this task.

Voted for this task.

One of the major benefits for adding this will be to do an auto-retest of optimised strategies to add an additional OOS part built on top of optimised strategies e.g. build from 2013 to 2018, optimise from 2013 to 2018, then auto retest from 2013 to 2020 with OOS from 2019 onwards.

This will go nicely with the other feature request I've asked for: https://roadmap.strategyquant.com/tasks/sq4_6750

Voted for this task.

Voted for this task.

Voted for this task.

Voted for this task.

I have been having exactly the same issue for years. This would solve a common problem with strategies which are slowing down. 

Voted for this task.