| StockFetcher Forums · Filter Exchange · Help requested for shorting filter | << 1 2 3 4 >>Post Follow-up |
| cegis 235 posts msg #32738 - Ignore cegis |
8/2/2004 9:14:52 AM Rob, I'm glad you found my post helpful. I'm also pleased that you took it in the constructive manner in which it was intended. (That can be an issue with a medium like postings and e-mail...) If you'll induldge me, let me make a few more observations: - True, the test can use any value from 0.01 to 5.00 (not 1.00 <g>) and still filter out *some* stocks. However, since "z" is always going to be an integer, every value between 0.01 and 1.00 will produce the same results, as will every value between 1.01 and 2.00, and so on (adding 1 to both values). - Re your comment "the lower the volume, the better the filter is going to be": Something you may wish to add to the filter is "set{ratio,volume / average volume(90)} add column ratio sort column 5", which will compute *how low* the volume was compared to the average, and sort the list by lowest ratio first. (The column number may change if you have other "add column"s in your filter, but you didn't show any.) Now, your "best" stocks (based on volume) will be listed first! HTH (That's "hope this helps", not my name <G>), C |
| robdavis 69 posts msg #32749 - Ignore robdavis |
8/2/2004 11:27:08 PM C, I've used and implemented many of your good ideas (and a few good ideas from others). After many modifications, well, this is how the shorting filter looks at the moment. I feel it has improved, and is pretty good right now, but still could use some improvement. What do you think? ----- Filter starts ----- Show stocks where /*PRICE AND VOLUME*/ close is between __ and __ and volume is above ______ and average volume(90) is above ______ and volume 30 day low is greater than ______ /*WANTED: STEADILY RISING PRICE*/ and close has been increasing over the last 3 days /*PRICE SHOULD HIT THE UBB*/ and high is above upper bollinger band(20) /*ATTITUDE QD SHOULD BE POSITIVE*/ and count(close below open, 3) is below 2.1 /*WANTED: LOW-LOW VOLUME*/ and set {v, average volume(90)*0.8} and set {x, count(volume above volume 1 day ago, 1) * 2} and set {y, count(volume above v, 3)} and set {z, x + y} and z is below 1 /*SORT BY RELATIVE VOLUME*/ and set{RelativeVolume, volume / average volume(90)} and add column RelativeVolume sort column 5 /*7/30/04: Z USED TO BE BELOW 1.1, NOT BELOW 1*/ /*7/31/04: V IS A NEW VARIABLE, USED TO BE AVGE.VOL(90), NOT 0.8 * AVGE.VOL(90)*/ /*8/1/04: ADDED 30-DAY VOLUME*/ /*8/2/04: ADDED SORTING BY RELATIVE VOLUME*/ ---- End of filter -- And C, sorry about the HTH. I hope this helps. Rob |
| cegis 235 posts msg #32751 - Ignore cegis |
8/3/2004 9:18:49 AM Rob, I can't say I'm any kind of expert at short filters, but here are my thoughts: - Not being familure with short setups, it looks pretty good. Might some additional tests, or at least display of data (add column), help in making picks? (Really guessing here:) Maybe BB width? Maybe high RSI (RumpledOne's favorite)? Maybe different length for BB? Other??? Maybe adding some columns, then using the technique that I posted in the thread "Backtesting" in General Discussion to see returns, you might find some correlation to improve filter quality. - IMHO, the three volume checks is a bit overkill, but I guess it gives readers some ideas as to which they might want to choose. - If you put a filter between " - No biggie on the HTH vs C! I am **terrible** at remembering people's names, so I cut others a **whole lot** of slack when it comes to mine! HTH, C |
| dducey 19 posts msg #32758 - Ignore dducey |
8/3/2004 7:50:11 PM Try this one on for size: |
| robdavis 69 posts msg #32759 - Ignore robdavis |
8/3/2004 8:09:48 PM C, Using your fetcher[] idea -- hey, great idea! -- I'm posting another shorting filter; this one should work just as well... or maybe even better than the previous one (I posted): Click on it... enjoy it. I hope this helps you, as well as helps many other guys who read this post. Any comments? I do welcome your comments. Rob |
| cegis 235 posts msg #32766 - Ignore cegis |
8/4/2004 9:42:15 AM Rob, I haven't looked at the filter results, yet, but here are a few thoughts/observations: - One of the things I don't like about filters is that some very good stocks get filtered out because they're close to, but not quite matching, a restriction. You might want to consider changing "sort column 5" to "sort column 5 descending", and then (possibly) eliminating the "RelativeVolume is above 2.5" from the filter (or lowering the threshhold). That way, you can see stocks that are "close" and decide for yourself if they might be a good short. - Obviously, I was wrong about comments deactivating the Fetcher[] link. Good to know! Thanks... - Why did you change the volume test from being low volume to high volume? I thought (and perhaps incorrectly, so please correct me if I'm wrong) high volume usually indicated follow-through, and low volume indicated reversals. C |
| robdavis 69 posts msg #32770 - Ignore robdavis |
8/4/2004 1:17:12 PM C, You're right about the fact that some very good stocks get filtered out because they're close to, but not quite matching, a restriction. I changed "sort column 5" to "sort column 5 descending". And, I eliminated the "RelativeVolume is above 2.5", and this way, we might be able to see stocks that are "close". And now we can decide for ourselves if the stock might be a good short. Yes, it is good to know that comments do not deactivate the Fetcher[] link. You're welcome. Why did I change the volume test from being low volume to high volume? Good question! I did some back testing, and back testing indicated to me this relationship (i.e. low volume usually indicated reversals) was indeed true. I also believed high volume usually resulted in follow-through, and low volume usually indicated reversals. But, upon more back testing, I found that surprisingly the other extreme (as to the volume) also worked! I.e. high volume also indicated reversals. And, what's more, I was surprised to observe that, over and over, how often high volume produced reversals even more predictably. Now, this has become a challenge. This is a set of condition that needs some more serious back testing. Guys, we need to do some serious back testing. Guys, everyone who reads this, if you have a bit of time, help us out with back testing! Why? Because you, too, might benefit! In the next 12 months the market is likely to decline somewhat, thanks to interest rate hikes by the Fed. Interest rate hikes are very likely. Therefore, in the next 12 months, shorting filters are going to be highly useful. Do some back testing today, and you, too, will benefit! Here's the improved shorting filter with all the above changes. Test it, back test it, and enjoy the experience! I hope this helps. Rob |
| cegis 235 posts msg #32771 - Ignore cegis |
8/4/2004 4:07:07 PM Rob (et al), Perhaps this will help with the backtesting: What I did was add "25 days ago" to all of the indicators used for selection. This is the equivalent of using "offset 25 days", except, with this technique, I have "future" day's data available to me (24 days ago, 23 days...)! Before I get into how this aids in backtesting, I should mention other pertainent changes to the filter: - I decided to use average volume(3) instead of just volume in the calc of relative volume, in the hopes that it indicates a more sustained spike in volume. We may want to remove this change.... - I also changed it to filter based on EITHER very high OR very low volume. For this, I changed the limit to +/-25%, as opposed to the previous filter's 20%. I just figured it would be more restrictive, and therefore, hopefully "better". Again, change to suit your needs... Back to the backtesting. I added calculations of the 3, 5, 10, 15, 20, and 25 day returns (as $, not %, so as not to hit the SF nesting limit; I didn't even try to get the %...). These returns are NEGATIVE if the stock went DOWN for the period, which is what we're looking for. I further segregated this into high volume returns (HVrtn_ variables) and low volume returns (LVrtn_ variables) that are shown in the results table, so it's immediately obvious which selection caused the match. This should make it easier to discern if either volume extreme is better performing than the other. A quick look at some backtest dates shows this works pretty well! HTH, C |
| robdavis 69 posts msg #32780 - Ignore robdavis |
8/5/2004 2:56:38 PM C, Great! With your help, I feel we are getting somewhere! This (combination of) shorting filter (and back testing) is getting better! I do like your ideas of +/-25%, either/or option, and HV and LV returns. But, how about doing some reverse sorting for RelativeVolume? Rob |
| cegis 235 posts msg #32782 - Ignore cegis |
8/6/2004 9:07:13 AM Rob, Glad you liked the changes<g>! Not positive what you mean by "reverse sorting"... Just add "sort column 6 descending" if you want the results sorted by descending Rel Vol. What do you think about using average volume(3) instead of just yeaterday's volume? Another possibility would be to remove any HUGE volume spikes from the average volume(3) before computing the relative volume. I posted a while back on how to do this. The title was something like "Technique: Removing spikes from averages" or some such. The idea is to get rid of (generally news indiced) huge (2x or more, eg) one-day spikes in volume to get a better "underlying picture" of the volume action. Do you think it's worth looking into? C |
| StockFetcher Forums · Filter Exchange · Help requested for shorting filter | << 1 2 3 4 >>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