nibor100 1,046 posts msg #140234 - Ignore nibor100 |
12/19/2017 3:19:16 PM
@graftonian,
I've been working on Kevin filter #10, which I will post when finished with comparative testing of SF results vs. SS data.
Thanks,
Ed S.
|
graftonian 1,089 posts msg #140236 - Ignore graftonian |
12/19/2017 3:31:39 PM
roger that, Nibor.
#11 was to easy, I must be missing something.
|
Kevin_in_GA 4,599 posts msg #140247 - Ignore Kevin_in_GA |
12/20/2017 8:05:47 AM
My signal email from last night - 2 new BUY signals:
Side | Action | PF | OrderType | Product | Symbol | Name | BuyDate | BuyPrice | ClosePrice | GainLoss | System Name |
---|
Long | Buy | 10 | Market | Stock | XIV | XIV | | | 134.81 | | XIV MULTISYSTEM #20 | Long | Buy | 10 | Market | Stock | XIV | XIV | | | 134.81 | | XIV MULTISYSTEM #35 | Long | Hold | 10 | | Stock | XIV | XIV | 12/19/2017 | 135.5 | 134.81 | -0.51% | XIV MULTISYSTEM #33 | Long | Hold | 10 | | Stock | XIV | XIV | 12/18/2017 | 135.17 | 134.81 | -0.27% | XIV MULTISYSTEM #40 | Long | Hold | 10 | | Stock | XIV | XIV | 12/5/2017 | 116.7 | 134.81 | 15.52% | XIV MULTISYSTEM #23 | Long | Hold | 10 | | Stock | XIV | XIV | 12/18/2017 | 135.17 | 134.81 | -0.27% | XIV MULTISYSTEM #6 |
|
Kevin_in_GA 4,599 posts msg #140313 - Ignore Kevin_in_GA |
12/22/2017 9:06:33 AM
New signals for today - I noticed that for some reason the signal for 12/19 (#33) is not showing up. Not sure why.
Long | Buy | 10 | Market | Stock | XIV | XIV | | | 136.09 | | XIV MULTISYSTEM #32 | Long | Hold | 10 | | Stock | XIV | XIV | 12/5/2017 | 116.7 | 136.09 | 16.62% | XIV MULTISYSTEM #23 | Long | Hold | 10 | | Stock | XIV | XIV | 12/18/2017 | 135.17 | 136.09 | 0.68% | XIV MULTISYSTEM #40 | Long | Hold | 10 | | Stock | XIV | XIV | 12/20/2017 | 137.04 | 136.09 | -0.69% | XIV MULTISYSTEM #35 | Long | Hold | 10 | | Stock | XIV | XIV | 12/20/2017 | 137.04 | 136.09 | -0.69% | XIV MULTISYSTEM #20 | Long | Hold | 10 | | Stock | XIV | XIV | 12/18/2017 | 135.17 | 136.09 | 0.68% | XIV MULTISYSTEM #6 |
|
mahkoh 1,065 posts msg #140341 - Ignore mahkoh modified |
12/23/2017 6:05:11 AM
I have noticed that it is not uncommon for Yahoo to have a way off price (usually the open) and to correct that in the next couple of days.
Looking forward to following your new thread. You state that it is a problem to use SS for weekly data. I agree that transposing all formulas into weekly would indeed be a pain. But how about going the oppsite way, importing weekly prices as an ASCI file and run the regular formulas against that data?
|
Kevin_in_GA 4,599 posts msg #140343 - Ignore Kevin_in_GA |
12/23/2017 7:53:50 AM
That would work. I haven’t tried it yet but I have found with combined long and short systems SS tends to give you win rates that are in the 70/80% range over longer time frames. Certainly good but not at the level I’m seeing with the NN systems.
Maybe a contest to see who can produce the best XIV long and short system from 2011-2017 using either daily or weekly data?
|
mahkoh 1,065 posts msg #140364 - Ignore mahkoh |
12/24/2017 6:56:24 AM
As you already have a complete system for XIV, shall we do GDX? If the neural networks are about spotting a pattern and expecting this pattern to continue; well, in hindsight even I could have spotted the pattern in XIV over the last 2 years. The GDX chart however is a different animal and might just prove to be a challenge for Neuromaster. I know it is a challenge for SS as I have tried several but have yet to find strategies that are consistent going forward.
I have purchased the software as well by the way, after all the PS4 I'm getting my son for X-mas is more expensive and of that I'm sure it will not pay for itself. As I understand support is pretty much nonexistent so you'll probably be hearing from me when I encounter issues..
GDX data starts in May 2006, that's 11 years or 570 bars. Roughly the same as a 2 year backtest when based on daily data. Currently, after 8000 runs I'm looking at 86 % profitability and 6% annual return on a portfolio size of 10.
|
Kevin_in_GA 4,599 posts msg #140369 - Ignore Kevin_in_GA |
12/24/2017 4:44:55 PM
HOW ABOUT THIS?
Weekly data used (I did three separate OneClick runs and combined them). Portfolio of 10 returns 21.78% per year, annualized Sharpe ratio is a ridiculously high 4.22. Win percentage is not astounding (73%). Total number of trades is 296 (about 50/50 long and short).
|
mahkoh 1,065 posts msg #140373 - Ignore mahkoh modified |
12/24/2017 6:02:50 PM
Impressive.. Assuming you created the weekly GDX sector today or yesterday you must either have a very fast computer or give the one-click few runs before moving to the next. In this "multi-one-click" approach I let the one-click go for some 700K to 800K before starting the second, at what point do you move on to the next?
I'm running the system from the end of 2006 and have an initial capital of 50000, so we're a bit off here. As I'm still in my first run I assume a 5 position portfolio. 82% profitable, 116 trades also 50/50 long/short, annual return 12 % but Sharpe ratio only 0.40.
How do you resize the picture and get the white background?
|
Kevin_in_GA 4,599 posts msg #140375 - Ignore Kevin_in_GA |
12/24/2017 7:30:32 PM
I am using the High Sharpe Ratio OneClick, and letting it run about 25-40000 runs. By then I usually have a pretty good system. I then save it into a Multisystem, clear the results in the OneClick, and start over.
Make a signal for each MS, Merge the signals into one single signal system, then Create MS from that signal to combine all of the MS into one.
Just resize the Detailed results box to get the picture smaller. I right-click on the graph and choose Settings - from there you can modify the colors as wanted.
|