All Open Requests
All Open Requests
All Closed Requests
You have voted on this request. View results below.
R2G_Compare a 'custom group' of strat's to a custom 'peer group'
This is an R2G request along the lines of better manager comparison and selection tools. However, I would love to be able to do this with any of my live port's as well.
Background. See: Hedge Fund alpha persistence. It's fairly clear and simple paper.
The paper shows that for Hedge Funds, at the 6, 12 and 24 month levels (but especially the 6 month level), o-s-s strategy raw returns and strategy alpha vs. an appropriate bench are very effective (and statistically significant) selection methods for predicting Short and medium term excess returns and alpha going forward. The single best variable that can be added to these two is a 'strategy distinctiveness index' measure from a peer group.
I don't want to have to regularly do this in excel if I can avoid it.
So...For these measures to work, strategies users should be able to create 'groups' of strategies, custom bench's and custom 'style indexes' or 'peer groups' for each. That's the bulk of this request. It's a user interface / data analysis request.
P123 would add tools allowing users to run studies creating custom 'fund groups', custom benchmarks and custom 'peer groups' for any R2G's (the custom peer group could be the custom benchmark). P123 would then calculate alpha and performance stats for a single R2G or group of R2G's vs. this (user defined) custom peer group and also (ideally) create the Strategy Distinctiveness Index (SDI) number and a rank for the strategy or group of strat's vs. it's peer group and benchmark - using alpha and SDI and any other user specified existing features (fund flows might be one if the strat's being autotraded).
For example. I want 3 microcap systems. So, I click on Joe's 5 stock model. Joe's 10 stock model and Dave's 20 stock model as my 'port'. And I then click on and create a custom group of 15 microcap systems with similar turnover and at least 10 months of live o-s-s. This is also my custom benchmark.
And then, from P123, I run the study and I then get performance stats only on their o-s-s performance over the trailing 6 months. And I see how they did as a group vs. the underlying style index.
(Long-term a ranking function and backtesting function will ideally be added to these - so sim's can be run on 'regime switching' among models).
It would be great to be able to specify that only out-of-sample data is used in the studies. And what length of period the user wanted to include.
Again, I would think that this will likely work better at the manager level then at the 5 stock system level, due to the very high variance likely existing in 5 stock models. So...ideally we could compare one 'group' of strat's to a peer group and get all of the above. And, ideally, we could specific the time period for the study (i.e. trailing 6 months, trailing 1 year, etc).
, # of Votes:
Scores are calculated as (importance) × (# of votes), where importance ranges from 1 to 4.