| StockFetcher Forums · Filter Exchange · TRO CROCK POT - SLOW COOKERS FOR SWING TRADERS | << 1 ... 16 17 18 19 20 ... 26 >>Post Follow-up |
| DMUNCASTER 34 posts msg #115815 - Ignore DMUNCASTER |
10/10/2013 2:41:47 PM Kevin, I guess the reason you're seeing 90% winners over the past few weeks and I'm seeing closer 50%, is that I'm looking at all the stocks selected on Fridays, while you're picking only the first 5. Are you comfortable that the first 5 will always do better than the total scan? As I posted earlier, selecting only the stocks that trade above Friday's high, usually 4 or 5, I get 100% winners, but that too could just be a lucky short-term run. That's why I hope you're able to somehow code this in SS for backtesting. Also, the system desperately needs a set exit strategy for losing trades. The reason I keep nit-picking at this, is that I know there's something here, but I just don't feel that it's right yet. There's a reason this thread shut down for so long. Either it didn't work out so great in the real world, or everybody involved back then is living in big homes at The Villages that they bought for cash. |
| Frank460 4 posts msg #115823 - Ignore Frank460 |
10/10/2013 7:01:29 PM Is there a collection of TRO postings anywhere on this website?? |
| Marco Sicily 20 posts msg #115825 - Ignore Marco Sicily |
10/10/2013 10:44:50 PM If you click on" The Rumpled One" on the first page of this thread it will take you to over 600 pages of TRO's posts. |
| Kevin_in_GA 4,599 posts msg #115836 - Ignore Kevin_in_GA modified |
10/11/2013 1:03:19 PM Kevin, I guess the reason you're seeing 90% winners over the past few weeks and I'm seeing closer 50%, is that I'm looking at all the stocks selected on Fridays, while you're picking only the first 5. Are you comfortable that the first 5 will always do better than the total scan? You need a cutoff somewhere, and for practical purposes you probably would only invest in 5-10 each week. Taking the top ten performers seems to be the right approach - this is all driven by statistics, as you well know. Sports analogy - I'm more comfortable Kobe at the foul line than I am putting Shaq there. Why? Historical performance. As I posted earlier, selecting only the stocks that trade above Friday's high, usually 4 or 5, I get 100% winners, but that too could just be a lucky short-term run. Are you testing this manually? Is this that the weekly high, regardless of what day it falls on, is used or is it only the Monday high? For backtesting one could try the same thing using the prior week high, but that is probably something outside of the topic of this thread. Also, the system desperately needs a set exit strategy for losing trades. It's there - the Friday close. I wouldn't try to tinker with this too much, but if you feel that it is critical to have a loss mitigation component try analyzing each trade on the daily charts to make sure you aren't turning too many winners into losers in the hopes of limiting a few big losses. However, a single loss can wipe out one or more weeks of gains, as some are 15-20% drops. Even one of them can set you back a month, and it's not like they are all that uncommon. The only way to offest the increased losses one will take from adding a stop loss is to simultaneously raise the target gain percentage. Might be worth exploring. |
| mahkoh 1,065 posts msg #115844 - Ignore mahkoh |
10/11/2013 4:33:09 PM TRO has an interesting suggestion somewhere in the beginning of this thread: Close out all positions once the total portfolio profit is 2%. |
| Frank460 4 posts msg #115847 - Ignore Frank460 |
10/11/2013 8:44:21 PM Marco, I think I found the TRO information at http://www.stockfetcher.com/sfforums/?q=byuser&author=TheRumpledOne?? If this is wrong please let me know. Thanks for you help. Frank Hopkins Houston,Tx |
| bjfeen 24 posts msg #115848 - Ignore bjfeen modified |
10/11/2013 9:23:55 PM Here is a slight modification of Kevin's streamlined filter of TRO's original. Its not a filter per se, but rather a means of backtesting Kevin's streamlined filter by pulling in all the matches from LAST week Friday, and testing to see which ones were triggered by THIS week Friday... Fetcher[ /*calculate percent difference between weekly high and weekly open (1 WEEK AGO)*/ set{weekhigh, weekly high 1 week ago} set{weekopen, weekly open 1 week ago} set{highminusopen, weekhigh - weekopen} set{pctdiff, highminusopen / weekopen} /*calculate how many times in the last 52 weeks the weekly high has been 1%, 2%, 3%, 4%, and 5% above the weekly open (1 WEEK AGO)*/ set{1pct, count(pctdiff > .01, 52)} set{2pct, count(pctdiff > .02, 52)} set{3pct, count(pctdiff > .03, 52)} set{4pct, count(pctdiff > .04, 52)} set{5pct, count(pctdiff > .05, 52)} /*which selections from LAST WEEK FRIDAY have hit their target THIS week?*/ set{hiopen, weekly high - weekly open} set{pctdiffthiswk, hiopen / weekly open} set{1pcttarget, count(pctdiffthiswk > .01, 1)} set{2pcttarget, count(pctdiffthiswk > .02, 1)} set{3pcttarget, count(pctdiffthiswk > .03, 1)} set{4pcttarget, count(pctdiffthiswk > .04, 1)} set{5pcttarget, count(pctdiffthiswk > .05, 1)} /*criteria*/ 2pct above 44 close above 5 average volume(90) above 500000 add column hiopen {high - open} add column separator add column weekhigh {weekly high last week} add column weekopen {weekly open last week} add column separator add column weekly high {weekly high this week} add column weekly open {weekly open this week} add column separator add column 1pct {1% 52 wks} add column 2pct {2% 52 wks} add column 3pct {3% 52 wks} add column 4pct {4% 52 wks} add column 5pct {5% 52 wks} add column separator add column 1pcttarget {1% target hit} add column 2pcttarget {2% target hit} add column 3pcttarget {3% target hit} add column 4pcttarget {4% target hit} add column 5pcttarget {5% target hit} chart-display is weekly ] Why can I not make this clickable?? |
| Kevin_in_GA 4,599 posts msg #115849 - Ignore Kevin_in_GA |
10/11/2013 10:23:18 PM This sometimes happens when you have a lot of comment lines - I just took most of them out and it now works. |
| DMUNCASTER 34 posts msg #115851 - Ignore DMUNCASTER |
10/12/2013 2:49:31 PM Are you testing this manually? Is this that the weekly high, regardless of what day it falls on, is used or is it only the Monday high? For backtesting one could try the same thing using the prior week high, but that is probably something outside of the topic of this thread. I am using only the high of the Friday filter, not consecutive days. I'm not sure the historical statistical argument applies here. If you go in at the open on Monday, you enter at a fixed point from which the price can go up or down. If you enter only when the high of the previous Friday is breached, you're entering during an upward swing. The chances are that you haven't entered on the high for the day. I'm backtesting manually and it works for me, so far. I am using the following Friday close to exit any losers, and cancelling any GTC buy stops that haven't hit by Thursday close. Is it possible in SF to scan some weeks in 2008 to early 2009? I'd love to see how this works in serious down markets. And thank you for responding to my posts. I'm amazed at all the things you manage to get to. Is there a Mrs. Kevin that puts up with all of this? |
| Kevin_in_GA 4,599 posts msg #115855 - Ignore Kevin_in_GA |
10/13/2013 11:58:32 AM Just thinking about this approach, and it came to me that these stats are most likely wrong. As an example, the current filter calculates the number of times a given stock hits the 2% target from the weekly open, and requires that the weekly open be above X and the weekly volume be above Y. Seems reasonable. But in actuality you are comparing apples and oranges - while it is true that a given stock hit the 2% target 48 times over the last 52 weeks, it likely was not also meeting the screening criteria when it did so. Example - CSIQ seems to have hit the 2% target 49 times over the last 52 weeks, but you also required that the close be above 5 and the average volume(90) be above 500000 to enter the trade. If you apply those criteria to each week, you'll quickly see that it only would have qualified as a potential trade 24 times. EVC is another example - the current filter shows 47 times the target was hit, but in reality it would have only done it 15 times since it did not meet the screening criteria for price and volume. Look at HSOL - how many times over the last 52 weeks did it meet the price and volume criteria? Not 48 times as this filter would lead you to believe, but only 4 times since it only broke above $5 a few weeks ago. I still think that this concept is valid and can be very profitable if executed properly, but it is statistically-driven and therefore you must always meet the entry criteria before you can count win frequencies. Now you can lower the entry price requirement, but you soon get to a point where the bid/ask spread will take 1 percent back, and then after commissions you make next to nothing while still assuming the same downside risk. I rewrote TRO's original code to correct for this, and will post it in a new thread later today. |
| StockFetcher Forums · Filter Exchange · TRO CROCK POT - SLOW COOKERS FOR SWING TRADERS | << 1 ... 16 17 18 19 20 ... 26 >>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