| Index | Recent Threads | Who's Online | User List | Search |
|
|
 |
Portfolio123 » List all forums » Forum: General Comments » Thread: An interesting observation |
|
Total posts in this thread: 6 |
[Request a Feature]
[Post new Thread] [Add To My Favorites] [Watch this Thread] |
| Author |
|
|
olikea
Advanced Member
|
A while ago I posted a thread about an apparent short covering rally: http://www.portfolio123.com/mvnforum/viewthread?thread=3190 During that week, shares with a high level of short interest did much better than those with little short interest. After a bit of playing around, I discovered something really interesting, if you apply the exact same test to the subsequent week, you get a complete reversal. I have attatched an image of the two rankings over the two subsequent weeks, and you can clearly see the stark contrast between them. I wonder if this means there is a phenonminan such as "factor reversal", whereby if a particular quant factor has been performing very well over, say, a week thne it is time to switch to something new. I am quite excited about this because I suspect very few people are looking at this sort of thing. Unfortunately given how P123 works its quite difficult to test this hypothesis... any thoughts? ---------------------------------------- 
SHORTinterest.png
(20993 bytes)
(Download Count: 101)
----------------------------------------
SHORTinterestPlus1.png
(21558 bytes)
(Download Count: 100)
|
||
|
|
synhawk
Advanced Member
|
Hi Oliver, I do suspect there is mean reversion throughout the markets in a variety of ways. This doesn't mean that nothing works over the long haul, it just means that if say a factor is good for say, 5% a year in the top .01 percent of ranks over a 10 year period, there will probably be good times and bad times with the mean being a slow rising slope. I suspect you will see quite the same in short time frames although probably with a lot more noise to help confuse things. It would be interesting to see if one could come up with a "closed-set" of factors that embody all market conditions, although I think such a finding might be highly dubious in nature. The market is a tricky minx! It may be possible to follow the momentum in factors sort of like how some people do fund timing. It is certainly an interesting thing to discuss/research. |
||
|
|
Stittsville123
Advanced Member
|
"Unfortunately given how P123 works its quite difficult to test this hypothesis... any thoughts?" Oliver - it is not really that difficult to test. To do so I used the following screen: http://www.portfolio123.com/screen_summary.jsp?screenid=15685 I did a weekly backtest for the entire data history for two conditions: (1) FRank("SIRatio",0,0) > 90 & FRank("SIRatio",0,1) <= 100 (2) FRank("SIRatio",0,0) >= 20 & FRank("SIRatio",0,1) < 30 I downloaded both sets of results and consolidated the average weekly gain for each into a spreadsheet. Then I subtracted the average weekly gain of the 20 - 30 rank from the 90 - 100 ranking. This will give an idea of when the SiRatio factor is "working" versus when it is inverted. The raw results did not tell me much. So I plotted a 20 week average. It looks like the graph below. I have also attached the spreadsheet so you can play with it if desired. Steve ----------------------------------------
SiRatio 20 Wk Avg.png
(38986 bytes)
(Download Count: 88)
----------------------------------------
SiRatio Weekly Evaluation.xls
(94720 bytes)
(Download Count: 1)
|
||
|
|
Stittsville123
Advanced Member
|
Please ignore the graph and attachment from my last post. There was an error: the second FRank("SIRatio",0,1) should have been FRank("SIRatio",0,0). The following are (should be) the correct results. Steve P.S. I was wondering why there wasn't much difference between the 90-100 ranking and 20-30 rankings. ----------------------------------------
SiRatio 20 Wk Avg.png
(45622 bytes)
(Download Count: 88)
----------------------------------------
SiRatio Weekly Evaluation.xls
(94720 bytes)
(Download Count: 1)
---------------------------------------- [Edit 1 times, last edit by Stittsville123 at Apr 21, 2008 8:42:40 PM] |
||
|
|
probtrader
Advanced Member ITALY Joined: Oct 11, 2004 Posts: 109 Status: Offline |
My take (to be verified) is that you're going to see outperformance in high short interest stocks only at significant market bottoms, like in 2003, when informed investors rapidly cover their short positions. The whole phenomena probably only last a few weeks (if not less). A good test could be to limit the ranking universe to a maximum SI, for instance 15% SI of float, to make sure extremely high SI values are not the actual determining factor. Then compare April 2003 returns or the March 22 period you have found with other monthly return away from those significant bottoms. |
||
|
|
dwpeters
Advanced Member
|
Hi, My guess is a short covering rally followed by mean reversion. Whether the short covering rally signals a long term bottom or intermediate bottom remains to be seen, but Louis Navellier seemed to think it's a good sign. FYI, Louis Navellier uses factor based systems and commented on the short covering rally here: http://stocks.navellier.com/commentary/marketmail_archive.aspx?MMDate=3/28/2008 and again here: http://stocks.navellier.com/commentary/marketmail_archive.aspx?MMDate=4/4/2008 Don |
||
|
| [Show Thread Printable Version] [Post new Thread] |
