HomeReady-2-GoPortfolioRankingScreenerStockETFToolsCommunityHelp
ForumsEmail UserPollsFeature RequestsGroups


  Index  | Recent Threads  | Who's Online  | User List  | Search
  Search  
Quick Go »
Thread Status: Normal
Total posts in this thread: 3
[Request a Feature] [Post new Thread]
[Add To My Favorites] [Watch this Thread]
Author
Previous Thread This topic has been viewed 853 times and has 2 replies Next Thread
wolfi123
Member


SWITZERLAND
Joined: Dec 2, 2005
Posts: 7
Status: Offline

Weekly Performance for a Ranking System Reply to this Post
Reply with Quote

Hello,

When I want to find out the weekly Ranking performance, I guess I have to run Ranks for each consecutive week with a rebalance frequency of also one week. When I did this, I received in some weeks strange results for the return of SP500 (and probably also the buckets).
Example:
For the Period 07/17/2009 to 07/24/2009, the annualized performance was 2675.3179 %. This would have corresponded to a weekly return of 51.3%! In fact this return should have been around 4.2%.

What am I missing?

Peter
[Oct 5, 2009 8:52:57 PM] Show Post Printable Version     [Link] Report threaten post: please login first  Go to top 
marco


UNITED STATES
Joined: Jan 1, 1970
Posts: 4059
Status: Online

Re: Weekly Performance for a Ranking System Reply to this Post
Reply with Quote

Try these dates:

7/11/09 - 7/18/09

You get an annualized return of about 3,500% for the Sp500
If you switch to the Historical Returns option you'll see that the return is 7.11% for that week.

Now, 7.11 compounded weekly for 1 year is about 3,500%


Your examples had dates of 7/17/09-7/24/09 which shows a weekly return of 11.53%. However 11.53% compounded 52 times should actually be about 29,000%, but P123 shows an annualized return of only 2,600%. The problem is with the dates. If you want weekly returns you must start/end on a saturday, because that's when the data is updated and the ranking system performance quietly re-align itself on saturdays. This is not a problem when testing long periods, but in this case it's confusing.

What's actually happening is that the 7/17/09 date is going back to the Monday 7/13/09 (the closest day to the previous saturday) and the actual number of days used is therefore 7/24/09 - 7/13/09 = 12 days.

So 11.53% in 12 days compounded for a year is about 2,500% (there are 30 12-day periods in 1 year)


In conclusion... If you want to see weekly runs make sure that the starting date is on a saturday and use the 'Historical Returns' option (the numbers won't be so crazy, they'll just be the weekly returns)
----------------------------------------
Portfolio123 Staff
----------------------------------------
[Edit 1 times, last edit by marco at Oct 5, 2009 10:12:54 PM]
[Oct 5, 2009 10:11:29 PM] Show Post Printable Version     [Link] Report threaten post: please login first  Go to top 
wolfi123
Member


SWITZERLAND
Joined: Dec 2, 2005
Posts: 7
Status: Offline

Re: Weekly Performance for a Ranking System Reply to this Post
Reply with Quote

Marco,

thanks for the tip and the quick response.
When I run Sims from Fr to Fr with 1 wk rebalance, the way I figured out from the trades, is that it will buy with next Mondays prices but still based on the last rebalance which is before this period. As soon as the rebalance on Sa is reached, it fills in the new stocks with the same prices from coming Monday. These new stocks are then hold until Fr.
So could it be while running the week Fr to Fr as a Sim you'd have 5 days holding the stocks while in the Rank performance it would be 12 days?
But so far I think I could live with historical ranks tests from Sa to Sa.

On my other thread , could you please revisit it, because I still confirmed it as an error.

Peter
[Oct 6, 2009 9:16:39 PM] Show Post Printable Version     [Link] Report threaten post: please login first  Go to top 
[Show Thread Printable Version] [Post new Thread]

Free Trial  /  Log In
Username or Email
Password
Stay logged in
Can't remember username or password?