| StockFetcher Forums · Filter Exchange · Help please? What does this mean? | << 1 2 >>Post Follow-up |
| karennma 8,057 posts msg #54773 - Ignore karennma |
9/13/2007 5:31:46 PM 3-day linear regression line below -.9 standard deviations OR 3-day linear regression line above +.9 standard deviations |
| conyeuchua 83 posts msg #54777 - Ignore conyeuchua modified |
9/14/2007 2:11:39 AM Karenma, Linear regression line can be thought of as a best-fit line drawn through the data points. For example, the following filter shows linear regression lines for etf such as SPY, with the top linear regression line at one standard deviation above the center line. The bottom linear regression line at one standard deviation below the center line. On average, in this example, you can expect the 14 data points to be within the lower limit of "average - one standard deviation" and the upper limit of "average + one standard deviation" 67% of the time draw linear regression line(14,1) etf On average, in this example, you can also expect the 14 data points to be within the lower limit of "average - 2 standard deviations" and the upper limit of "average + 2 standard deviations" 96% to 98% of the times. Data points that fall outside these limits are considered to be outliers and are expected to fall back to the average, hence "regressing to the mean". Run the above filter and take a look at the regression lines for SPY (quite interesting.) Essentially, you are looking for a high probability for the next data point to behave in a certain way. I did not use 3 and 0.9 because sF does not seem to like it. Also, statistically, linear regression lines(3,0.9) are not meaningful. A final note: I believe at least one of the resident gurus of this forum advocate using linear regression lines(14,1) or linear regression lines(20,1) in conjunction with their favorite indicators, be them RSI(2) or Williams %R(10) or Stochastics %K(14). Hope this helps. Happy trading. |
| karennma 8,057 posts msg #54780 - Ignore karennma modified |
9/14/2007 8:31:25 AM Cony: Thanks so much for the explanation. Here's the dilemma .... I signed up for a "trial" of DataView's "Hotscans" because it said they had a feature that was pre-programmed to automatically pick out stocks at certain points in a pre-set linear regression channel which were highly reliable. I signed up yesterday. However, when I ran the scan last nite, I got no picks. What I found out was, the LRC program doesn't give you picks every day. Some days it has picks, some days it doesn't. Since I knew the parameters, I tried to use them on SF, but to no avail. Tell me what you think ... or anyone else who's reading this. Click 6/27/07 (Hotscans) http://www.traders.com/Reprints/ArticleList.html$DV_HOTSCAN OR http://scans.marketgauge.com/help/LinRegStrategies.asp |
| maxreturn 745 posts msg #54783 - Ignore maxreturn modified |
9/14/2007 9:58:44 AM Karen, I looked at the scans on the scans.marketgauge.com website. I don't know what you were using on SF to try to code this. I don't believe SF has a linear regression channel which draws the uppper and lower lines based on standard deviations. Following is the description of how the upper and lower lines of the SF linear regression channel are calculated: "The linear regression is built by computing a least-squares line over the specified last number of days. This line represents and expectation of where the price should be. To create the linear regression channels, we use the popular method of finding the point furthest from the least squares line and create two channels of equal width from that point. StockFetcher allows you to customize the width of the channel through a scaling factor. By entering a scaling factor of 0.50, a channel of half the normal width is created." OK...the hotscan uses upper and lower lines that are drawn .9 std deviations above and below the center line. A 1.0 standard deviation line drawn above and below the center line should contain about 67% of the data. Although SF does not allow you to draw upper and lower lines I think you can get a pretty close fit by using the center regression line function. Following is an example of SF code which tries to mimick the hotscan code for buy signals in an uptrending channel. NOTE: I used .67 to draw the channel at 67% of the normal width to approximate the 1.0 STDev lines used by hotscan: Please let me know if you have any questions or if anyone sees any flaw in the logic, let me know. Best Regards. Max |
| conyeuchua 83 posts msg #54784 - Ignore conyeuchua |
9/14/2007 10:28:26 AM Karenma and Maxreturn, I think it's better to learn and develop your own than to pay for a service (a black box) that we do not understand. It is also better to focus on the strategy and concept rather than to get too hung up on the technical details. I believe a generic desription of this "genre" of strategy was described by some site as "the strategy is to wait for prices to approach the upper trendline or lower trendline to find a setup. There are two ways the prices can go: continue in the channel bouncing up and down the channel like a ping-pong ball or break above or below the channels; either prices accelerate the dominant trend or break and reverse in the opposite direction." Best of all, TheRumpledOne has given us this great filter. Check it out: http://forums.stockfetcher.com/sfforums/?q=view&fid=1002&tid=47847&qrid= Good luck and Enjoy. cyc |
| maxreturn 745 posts msg #54785 - Ignore maxreturn |
9/14/2007 10:39:50 AM conyeuchua, you said.... "Karenma and Maxreturn, I think it's better to learn and develop your own than to pay for a service (a black box) that we do not understand. It is also better to focus on the strategy and concept rather than to get too hung up on the technical details." AGREED. No need to preach at me brother. The way I read her post she was asking for SF code to mimick the hotscans. That is exactly what I attempted to do. As good as TRO's code is, it needs further refinement to get at what Karen was requesting. |
| karennma 8,057 posts msg #54786 - Ignore karennma |
9/14/2007 10:44:32 AM Yes, that's all I was trying to do .. As I said, hotscans gave me NO picks .... So, I figured .. no problem. I'd just duplicate the parameters on SF. SF always gives me picks because it's so much more flexible. THANKS for your help guys. :>) |
| karennma 8,057 posts msg #54787 - Ignore karennma |
9/14/2007 10:47:53 AM Max: good point about the tweaking .... altho' I've been using the same filters for 5 years ... OPFs (Other People's Filters) LOL! I've "tweaked" them to my specifications. :>) |
| TheRumpledOne 6,411 posts msg #54802 - Ignore TheRumpledOne |
9/14/2007 8:32:57 PM "AGREED. No need to preach at me brother. The way I read her post she was asking for SF code to mimick the hotscans. That is exactly what I attempted to do. As good as TRO's code is, it needs further refinement to get at what Karen was requesting." What code? |
| conyeuchua 83 posts msg #54807 - Ignore conyeuchua |
9/14/2007 11:10:07 PM Maxreturn, There was no intention of preaching. Over the years, I've learned from many people in this forum like yepher, TRO, Alf44, niko, noahedwinbeach, etc too many to list all. So, I just wanted to help. cyc |
| StockFetcher Forums · Filter Exchange · Help please? What does this mean? | << 1 2 >>Post Follow-up |
Copyright 2022 - Vestyl Software L.L.C.•Terms of Service | License | Questions or comments? Contact Us
EOD Data sources: DDFPlus & CSI Data
Quotes delayed during active market hours. Delay times are at least 15 mins for NASDAQ, 20 mins for NYSE and Amex. Delayed intraday data provided by DDFPlus