CXO Advisory posted a column this morning in which they attempted to refute the effectiveness of the 10-week Nasdaq/S&P 500 lead/lag indicator. In their column they referred to my post of last week and Gerald Appel’s book which I referenced as the origin of the indicator. They then ran several unrelated tests to show that 10-week RSIs are not effective in determining future returns. It is important to understand that RSI was not the indicator described by Gerald Appel nor used in my testing.
Frankly, in the past I’ve found some interesting things on their site, but their lack of attention to detail here is flabbergasting. If they had either 1) opened the book, or 2) downloaded the spreadsheet I provided for FREE with the complete research supplied, they would have understood this. Instead they ran their own tests using a completely unrelated indicator.
Their tests ran from 2005-present. They showed that over the 2005-present time period buying the S&P when the 10-week RSI of the Nasdaq was above the 10-week RSI of the S&P would have lost money. This doesn’t surprise me.
For anyone who has downloaded the free spreadsheet, if you plug a round number (such as $1,000 or $100,000) into the 12/31/04 row you can see the following results from then until 5/22/09. The S&P would have lost about 25% (not including dividends). The model (assuming 0% interest on cash and not including dividends) would have gained about 11.6%.
In the next few days/weeks I will be publishing some more research that pertains to this indicator. I hope readers will find it valuable.
In the future should CXO attempt to refute the work of others I would hope they at least make an effort to look at the work they are refuting.
One final note, the Nasdaq/S&P 500 lead/lag indicator signaled a buy on Friday’s close. Subscribers to Quantfiable Edges are now notified of all model changes via the new Quantifinder technology.
13 comments:
I thought the same thing this morning as I read it. I usually like their research, however they were way off this time. Also, there testing time period was not at all a significant enough time period. Very disappointed in CXO this time.
I've corresponded with him before and found him very responsive. I've sent a message about his misapprehension.
Rob - I too sent a message in asking for a retraction/clarification.
they suck ! you rule hanna !
the nazz/spx indicator has been deadly of late. the spy goes down more than nazz, and the indicator goes up. when and how do we know if the "unreliable period" is over? love this blog, but i'm wondering why the fuss if the stat is in a breakdown period.
Kinda makes you wonder what other articles of theirs CXO)is fouled up?
Piddleposh, tempest in a teapot, compared to how much I'm loving this new Quantifinder!
Besides I'm greatly relieved when people misunderstand Rob and misquote him, the more the better! The less understood, appreciated, and FOLLOWED his methodology is, the better.
(That is what is known as contrarian thinking...)
LOVE that Quantifinder!!!!
Daniel
Rob,
I believe there is a serious error in your spreadsheet. In cell F12:
=IF(F11="Lead",L11*(1+G12),L11)
should read
=IF(F12="Lead",L11*(1+G12),L11)
This reduces the gain to less than the buy and hold return.
The spreadsheet is correct. There is no error.
Please look at cell L12 as an example.
The logic states that if last week (F11) closed with the Nasdaq in the lead, then the market should have been purchased at the close (on last Friday.) Earnings are calculated for the current week if LAST week led. So L12 represents the $100k you started with plus the 1.36% you made that week because the S&P was purchased at the close of the week before (the F11 signal). All trades occur at the close on Friday (assuming Friday is the last day of the week).
Rob
CXO has used a straw man like test before.
I think anything that has merit is publicly flogged (via the straw man test - as was done here), while privately they have absorbed your good/valuable info into their own trading strategy; the operate much like the BORG.
:0
Isn't there a bit of simultaneity
in the algorithm, namely you credit
the gain of the week at the same time
as you sell. Since you don't know whether the week is a lead/lag until the close, and you credit the value of the close to the account.
Perhaps a more realistic model would
be to sell on the Monday open.
I think the spreadsheet is fine. I often make trades in after-hours trading as opposed to the open of the next day.
Recieved email this morning from author of the site. please not correction: http://www.cxoadvisory.com/blog/internal/blog6-01-09/
Post a Comment