OmniVest User Forums
OmniVest Pro Tools Beta
101 vs 25
0 replies, 880 viewings
|
OmniVest User Forums
OmniVest Pro Tools Beta
101 vs 25
|
Last Activity 10/7/2016 12:00 PM 0 replies, 880 viewings |
|
|
|
| Printer friendly version |
kmcintyre Elite    Posts: 890 Joined: 10/11/2012 Location: Portland, OR  User Profile |
1) It would allow PW to adapt to changing market conditions 2) It would allow large pools of strategies to be applied. (Larger than PW can handle at this point in time...) 3) My testing showed that it would help PW find magic beans. Also, I previously posted a spreadsheet showing results for an experiment where I ran 24 competitions using the 101 stock RTM strategies and standard account settings. I varied the simulation start date from 11/1/2011 - 10/1/2013. The duration of each PW run was 1 year. I ranked the EFs in an attempt to determine which EF was most predictive. Over the past two days I ran another experiment, leveraging the one just mentioned. I started by using OV to sort the 101 stock RTM strategies by CAR over 24 one year periods. The 24 rankings each preceded the simulation ranges used in my 24 competition experiment. Once I had the 101 strategies sorted by CAR, I took the top 25 performers and ran another PW run with the strategy universe limited to those 25 strategies. I repeated the process 24 times. So I used the top 25 RTM strategies from the previous year as the universe for a one year PW run. I did that 24 times. (redundant, but I wanted to be clear...) I'm attaching a spreadsheet showing the results. (I'm also attaching the spreadsheet that I used to capture and sort the strategy statistics for the prior years.) The "101 vs 25" tab attempts to summarize the difference between feeding PW 101 strategies vs feeding PW the top 25 strategies. It's not sexy, and I'm sure the data could be sliced in many ways. But here are my observations. 1) Using a CAR sort to filter the strategy pool increased the average CAR from 17.48% to 23.57%. Average Calmar and EE also improved. 2) The "best" stats pretty much stayed the same. 3) The "worst" stats improved dramatically. 4) The number of PW|EF combinations that resulted in lost money decreased from 28 to 2. I believe the experiment clearly shows that filtering the strategy universe has merit. I only had CAR available. (I've ask for one column of the Strategy page to be user selectable, but that idea was shot down as being way to hard to implement.) Hence I used a one year period to filter strategies. Would a different metric or time period be better? Quite possibly. But a simple CAR filter would sure be nice. (For the 3 reasons I sited at the top...) I also believe this experiment confirms what I have been saying for a while. PW doesn't find the magic beans. A CAR filter might incrementally improve results, but I believe there is a more fundamental problem with ECA and/or EFs that prohibits PW from achieving the kind of results we all desire. Please reconsider adding a CAR filter to PW. (Something more robust would be excellent too.) Keith  Attached file : EF Test - Stock RTMs.xlsx (130KB - 276 downloads) Attached file : Stock RTM performance.xlsx (199KB - 255 downloads) |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend(un)/Freeze thread | |
Toggle e-mail notification |
