I downloaded SPY weekly data from 1/2000. I calculated the open to close change in percentage terms for each week. Next I used the excel function (=WEEKNUM) to find the week number. 1/3/2012 was week 1, 1/30/12 was week 5 which makes this week 7.

I looked back to 1/2000 to see what happens in various weeks (forgive my simplistic counting Rocky, fwiw fundamentally I am wary of being long here). My N was 12 for any period not yet covered in 2012. Only 4/12 week 8s were up (66% down) with an average expectation of -0.80%.

Further only 2/12 week 9s were up (83% down) with an average expectation of -1%.

Market certainly doesn't seem interested in going down and the N is pretty small but probably worth considering.

Steve Ellison writes:

To avoid the problem of multiple comparisons, I would suggest running a simulation in which you randomly reshuffle the weekly changes and sort them into 52 groups. Repeat the random reshuffle 500 or 1000 times . Then tabulate the extremes (highest and lowest number of up weeks in ANY group) for each repetition. You can then sort the extreme results from each repetition and check what the top 2.5% or top 5% of the highest of 52 groups was in the simulation. If you find actual results that exceed these cutoffs, you may be onto something.

There are probably computer programs that will run such a simulation very elegantly; I just use the Excel RAND() function and copy and paste until I have 1000 repetitions



WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '7211' AND comment_approved = '1' ORDER BY comment_date




Speak your mind


Resources & Links