fotchstecker
304 posts
msg #122281
- Ignore fotchstecker |
12/16/2014 11:35:47 PM
I've been working on an idea but have no idea how to turn it into a filter. If anyone has pointers or critique, please let me know.
QQQ approximates 1/40th the value of the Nasdaq 100.
The basic idea is that at tops shorts begin covering while QQQ still has a bit of lagging overvalue. So, for a brief period of time (a day or so), QQQ can briefly take on an atypical ratio to the NDX. The change in proportion provides the hint when market makers perceive a top and begin staking profits on the/a forthcoming downmove. So, at tops QQQ tilts its hand by taking up slightly less than its normal proportion to the NDX, and we might see 1/41+.
I cracked open a google spreadsheet to test the idea. I divided the NDX close by the QQQ close for 400 days and did a few other calculations. I opened a chart and mapped out the results. On the days where NDX was closer to 1/41+ of the NDX, my impression is that there's a somewhat reliable connection to most meaningful downmoves. And for the times where there are groups of 41s, there are general downtrends.
The most recent downtrend from 12/5 to 12/15 is one exception. QQQ hasn't signaled itself as it did in hundreds of previous days. This leads me to think that one problem might be the relative changes to the prior period and not the 1/40 baseline itself. I didn't use any other confirmation via indicator except that the prior 41s seemed fairly well mapped to RSI below 20.
Any thoughts on how a filter might be used to capture statistics and confirm against useful indicators? Some other structure?
Kevin's "A Simple Market Timing Filter" is a neat format, along with maybe grouped day-counts (1, 5, 10, 15, 20, 30, 50, 75, 100)?
|
fotchstecker
304 posts
msg #122283
- Ignore fotchstecker |
12/17/2014 9:10:07 AM
Looks like I might have been pulling the wrong composite symbol into google spreadsheets. cleaning this up and re-posting.
|
fotchstecker
304 posts
msg #122293
- Ignore fotchstecker |
12/17/2014 10:41:41 PM
Oh, got things sorted.
Anyone have ideas similar to this one?
|
fotchstecker
304 posts
msg #122342
- Ignore fotchstecker |
12/22/2014 9:24:42 PM
FWIW, today was the first "41" signal since 11/4.
|
guspenskiy13
976 posts
msg #122348
- Ignore guspenskiy13 |
12/22/2014 11:21:15 PM
Not sure if I understood the concept right, but you are saying that etf QQQ gets overvalued in comparison to NDX?
If so, they are definitely arbitraging that w/ HFT
|
fotchstecker
304 posts
msg #122350
- Ignore fotchstecker |
12/23/2014 9:13:20 AM
Hey, Gus. Yeah, I didn't write it too clearly. I've been thinking about ways to look for turn on the Nasdaq 100. I've been looking at fluctuations in the ratio of NDX to QQQ, since QQQ normally trades around 1/40 or smaller of the NDX. At tops, the proportion looks to possibly change, and I was using 41 as a baseline for spotting potential downturn. I say possibly because I have not done any kind of statistical analysis; I've only done a 10-year history looking at the gap between the two and mapping those gaps to charts as a way of spot-checking.
In general, I posted the idea because I realized I might be looking into something someone had already thought through.
I agree with you, too, that the HFTs are definitely finding this gap with the expectation of gaining on the move down.
In general, I'm going to watch to see if yesterday's "41" is followed by a few more, and I'll be watching the NDX with interest.
Thanks, Gus.
|
mahkoh
1,065 posts
msg #122359
- Ignore mahkoh |
12/24/2014 5:54:26 AM
Have you taken QQQ's dividend into consideration?
|
fotchstecker
304 posts
msg #122360
- Ignore fotchstecker |
12/24/2014 9:55:54 AM
mahkoh, great question.
I used Yahoo data for the closes. This data set includes "adjusted close" (alongside non-adjusted close figures) for for factoring-in splits and dividends. I have been using the unadjusted-close data to spot-check against charts, but I have the adjusted close ratio also calculated.
Using adjusted closes, you can see that the "we're at a top signal" (not really quite there as a signal yet) is a "41" two days earlier.
non adjusted adjusted
41.06419729 41.06419729 12/23
41.05632052 41.05632052
41.04467025 41.04467025
40.93784173 41.09156557
40.91453831 41.06783672--
40.90418084 40.90418084
40.91546108 40.91546108
40.90084738 40.90084738 12/12
Looking through data for all periods including dividends, it looks like the adjusted data leads the non-adjusted data. In some cases, there's just a large series of 41+ data for the adjusted close where the non adjusted is in the high 40s.
Your question brings up a few new thoughts:
1. If there's value in using this relationship, it seems that a more fine-tuned number (than 41) is needed, or;
2. General change in frequency provides more of the indication, or;
3. Maybe a combo of looking at days count and better baseline will provide a sharper indication.
Anyway, thanks mahkoh.
|