Would it be feasible to allow /strategy/{id} calls for designer models?
My use case is that I’m trying to incorporate one or more designer models in my book allocation to improve risk adjusted returns, as I mentioned in a previous post. Instead of manual adding and tuning weights one at a time in my book, I use the /strategy endpoint to pull weekly returns, and use mean variance optimization to calculate an optimal strategy mix from the estimated mean returns and covariances.
This works fine for my strategies, but the /strategy endpoint calls fail for designer models, even if I’m subscribed to the model. For unsubscribed models, I’d argue the call should still work but with holdings omitted. I don’t believe this would reveal any information not currently available to users. For example, I can browse unsubscribed designer models and manually download their simulated daily returns in CSV format.
Allowing this via the API would very handy and hopefully drive more designer model adoption.
Sounds good. I added it in the “Priority Backlog” which is the bucket where we pick future tasks.
Thanks.
NOTE: This is inline with our priorities to improve the Designer Model marketplace. Some of the key improvements we plan are:
Allow for discretionary models where designer have control: either deviate from the recommendations or entirely hand picked trades (no intraday allowed).
Automated rebalances with a Linked Account (set it and forget it)
While we’re on the topic, there’s one other missing feature that’s making it harder than it should be for me to test out and adopt designer models: leverage at the book level. For example, say I’m trying to implement a book that’s 120% long equities consisting of my strategy A and designer strategy B. Since I can’t control the leverage of the designer strategy, nor can I control leverage at the book level, then I’m forced to control leverage in a roundabout way on strategy A. If I want a 60%/60% mix, then I need to generate a new sim for strategy A at 0.6/(1-0.6) = 1.5x leverage, and then generate a book weights of 0.4 and 0.6 for A and B, respectively. If I want to evaluate a 70%/50% mix, then I need to run another sim for A levered at 0.7/(1-0.7) = 2.33x. This doesn’t scale well for 2 strategies and is even harder to manage as your number of strategies increases. This is one reason why I’m finding it necessary to tackle this problem externally using a portfolio optimizer.