Feature Request: Add "Out-of-Sample" Statistics to Live Portfolios

Like the current Statics tab, the out-of-sample statics tab would contain Performance, Trading and Risk Measurements for all trades completed (i.e. open to close) since the portfolio was launched. I’m interested in studying the performance of my long-lived ports.
View the feature request here.

You got my vote :slight_smile:

Can you explain what this is specifically? Isn’t a live strategy an out of sample performance anyway?

A simulation can be converted to a live strategy. When that’s done, the simulation results are part of the presented data and separating the simulation results from the “live” portion is cumbersome.

+1.

Crucial +1

You can save a simulation as a new live strategy (not convert) then the strategy is live from the date you save it and everything is OOS.

Alternatively, you can load your live strategy into a book as a single asset, and then get the stats for any period for this strategy.

So I don"t really see the necessity for this request.

Like in DMs, it would be useful for comparing in-sample vs out-of-sample results. Maybe that’s the presentation format that should be used.

I like the concept but … what if:

Instead of providing stats since a port is launched, provide an option to recalculate the stats starting from any desired date. Do it for sims as well as live ports.

We could choose to see stats from the live date or some other date that is also meaningful, such as when a change was made that affected the model, its universe, or its ranking system, or a major market change. From a user perspective, it might be a simple matter of choosing a start date for all calculations now presented under the Statistics tab.

This approach would probably mean some assets already purchased but not sold were in hand at the start date chosen, rather than just cash. So I don’t know if that would fully meet your desired goal.

If there were a button to start from launch or any arbitrary date, that would work for me. But I’ve yet to see p123 implement a user feature request. So let’s set our expectations accordingly.

Another thought:

Perhaps allow both the stats start and end dates (defaulting to what they do now) to be overridden. That would allow some analysis of model behavior during specific market periods.