Kevin_in_GA
4,599 posts
msg #103543
- Ignore Kevin_in_GA |
11/30/2011 2:23:24 PM
End of month, and therefore a good time to look at these systems. I have posted an update for each (will finalize at the end of trading today, but the results are pretty clear). Also updated the pairs trading section with a compiled list of pairs as well as the criteria for selection and trading (see post above). I'll be tracking these mechanically as we go forward.
For some reason I'm unable to update the blog portion of the site, or add new blogs. Hopefully this will get fixed soon.
|
mahkoh
1,065 posts
msg #103599
- Ignore mahkoh |
12/3/2011 6:15:51 PM
Excellent website, Kevin.
I notice that you pair stocks from different sectors, doesn't this make the system vulnerable to sector risk?
Also curious on how you calculate the cointegration value and whether you make use of Stockfetcher in selecting trade opportunities.
|
Kevin_in_GA
4,599 posts
msg #103600
- Ignore Kevin_in_GA |
12/3/2011 7:10:12 PM
mahkoh
107 posts
msg #103599
- Ignore mahkoh 12/3/2011 6:15:51 PM
Excellent website, Kevin.
I notice that you pair stocks from different sectors, doesn't this make the system vulnerable to sector risk?
Also curious on how you calculate the cointegration value and whether you make use of Stockfetcher in selecting trade opportunities.
++++
I do pair from different sectors here, and that is a concern for exactly the reason you suggest. However, these stocks are highly cointegrated, highly correlated, and when traded this way return an 80+% win rate over the past two years. I figure with that type of historical behavior the cross-sector risk is mitigated. Besides, most of the sectors are now highly correlated in behavior anyway.
Cointegration was calculated using a new program called Pair Metrics (www.pairmetrics.com). NIce software but still needs some further refinement. Free trial for 30 days - feel free to check it out.
Stockfetcher is used for the nightly values - the TAA scores and rankings, and the XLU-XLI performance. I still wish SF would give us monthly data ... it would be a real help.
|
TheRumpledOne
6,411 posts
msg #103608
- Ignore TheRumpledOne |
12/4/2011 10:29:50 AM
Yes, monthly data and correct weekly open price.
|
Kevin_in_GA
4,599 posts
msg #103791
- Ignore Kevin_in_GA
modified |
12/14/2011 10:02:28 PM
Website updated.
Most recent data on all systems in place, including the pairs trades.
----------------------------------------------------------------
|
ceebee
20 posts
msg #103806
- Ignore ceebee |
12/15/2011 11:54:06 AM
Fantastic work.....thanks for sharing
|
MarkS
10 posts
msg #103901
- Ignore MarkS |
12/20/2011 6:35:06 PM
Kevin,
If I understand correctly, your TAA system uses basic rank or specifically for SF ROC(). I have done the same thing rather extensively in the past, especially with mutual funds (prior to them becoming so restrictive in trading rules). My results were usually pretty solid, but had a tendency to not do well when the market is highly volatile like it has been this year. That is most likely due to the lag inherent in this system not trying to catch the absolute bottoms and tops. Basically, it is a momentum indicator.
Have you tested ROC() against some of the other momentum indicators? Maybe against Slope including non-linear and quadratic.
Have you done any contrarian testing with the system? Instead of taking the top ranked, use the bottom ranked. The assumption being that whatever is at the bottom will eventually move to the top. That may actually work well in a choppy market.
I have a variation on ROC that I think there are some good arguments for why it is better. I am also looking at a variation on the variation which I believe could be even better, if the arguments are valid for the first variation.
OK, I know that is a little vague, but I don't want to share it just yet. If you are interested, I may present it here. If you are happy with ROC, I would rather not share it though. At least not until I can do more extensive testing myself. That will mean either upgrading my SF account or I am currently looking at another testing system that I can get a 30-day demo on. I just need to make sure I have enough time to take full advantage of that demo period.
Mark
|
Kevin_in_GA
4,599 posts
msg #103904
- Ignore Kevin_in_GA |
12/20/2011 10:46:53 PM
Mark:
I looked at a series of momentum-based indicators using StrataSearch (www.stratasearch.com). I still have not found anything as simple and profitable as the ROC. Like everything in life, the key is in the timing.
Check out StrataSearch - fully functional and free for the first 30 days. Great at finding and optimizing systems, and a very good backtesting capability as well.
Happy to discuss this in more detail - just send me a note through the web page and I'll give you some more details on what I have looked at.
|
MarkS
10 posts
msg #103925
- Ignore MarkS |
12/21/2011 7:57:32 PM
Kevin,
I sent a rather lengthy reply through your website. Let me know if it did not go through.
|
Kevin_in_GA
4,599 posts
msg #103953
- Ignore Kevin_in_GA |
12/22/2011 3:47:58 PM
Mark: Got your note. This approach is testable in Stratasearch, but I'm not sure it will beat what I am currently doing for both ease of use and profitability.
|