Basic Search




We speak

 Contact me for Custom work
or One on One Platform Instruction 

follow me on: 

« 3/5/2013 | Main | 2/24 to 3/2 2013 »


03-03-2013 11:55:12
Good morning
03-03-2013 11:55:39
Good Morning KB
03-03-2013 11:56:02
Hows it going?
03-03-2013 11:56:09
Must be close to afternoon there.
03-03-2013 11:56:22
nope, Im on west coast time now
03-03-2013 11:56:42
It is still going.......slowly
Ha, moved or just visiting?
03-03-2013 11:57:22
Moved, last June, I'm in Seattle now
First blue sky ive seen in months this morning
03-03-2013 11:59:38
Yeah, I spent a couple of rainy days up there a few years ago. I like 360 +/- days of sunshine in Phoenix. ;)
03-03-2013 11:59:58
lol, yeah too hot for me
but work and my girlfriend dragged me out here
cant complain these days
Testing a new algorithm, do you have a stock list of stocks that tend to follow technical trading better than others?
I guess I'm not sure what I'm looking for in the stocks I run against
I usually just pick a random list, and I think it might be affecting my results lol
03-03-2013 12:02:56
Most of the lists I make are for volatility, high Beta. But at other times I have made lists that correlate to an index.
03-03-2013 12:04:15
I see, high beta, thats a good idea, Ill give that a shot
03-04-2013 09:36:15
Morning fellas.
03-04-2013 09:36:49
Morning PM :)
03-04-2013 10:22:08
Im still having issues with signals in trade optics not automatically trading in rts.. anyone else have this issue?
03-04-2013 10:31:39
Sorry, I'm still mostly running Pattern rules in TO to get examples for the manual.
Are all of the signals in TO valid. Maybe RTS is right not to take.
03-04-2013 11:32:52
not having that issue, but I am still having issues with TO showing signals late, and also showing when the rules are not actually true
03-04-2013 11:37:59
I was just talking to Dan, and he says that they have most of these issues fixed for the new version for Live trading and that they will be porting the fixes back to this version soon. but no date yet for either Live or update here.
03-04-2013 11:39:21
Thanks Bruce
03-04-2013 11:39:53
sounds promising, I hope
03-04-2013 11:42:00
Your welcome :)
03-04-2013 11:42:09
anyone else having issues of TO sending signals when the rules weren't actually met? i.e., my main rule has multiple parts, but one subrule is set to RS<30, but I'm getting hits on my main rule, and when I look at it, RSI was actually 35
03-04-2013 11:44:53
I've seen some stuff like that in the Patterns, I've got mostly some of the Pattern templates running at the moment in TO to find examples to put in the manuals but for the moment I have kind of given up looking at them. I have also sent in reports to what I am seeing to Dan.
03-04-2013 11:49:56
Good morning all
03-04-2013 11:51:25
Good Morning Drewski :)
03-04-2013 11:52:47
I was fooling with Prodigio this weekend and could not find a divergence script. Is there one?
03-04-2013 11:53:49
Correlation? maybe.
03-04-2013 11:56:15
I have no clue...
But I think they are two different functions
Divergence is a function that determines whether two different values have moved in opposite directions over a period of time.
03-04-2013 12:02:09
Mobius did a article on divergence awhile back.
03-04-2013 12:02:11
Correlation A statistical function that measures the degree to which two variables are related. The resulting value is called the correlation coefficient.
03-04-2013 12:02:38
to anyone who is interested, probably no one :) I think I see the issue with the late signals in TO. I'm using 2 minute bars, and if my rule is true at the close of a bar, then I get the arrow on the next bar, but I'm not getting a signal until the bar with the arrow closes, which is 2 minutes late
03-04-2013 12:03:21
I can't access that address
03-04-2013 12:04:09
It is a free article but you need to be signed into the site. (Free Membership)
That is how Mobius wanted his stuff setup.
03-04-2013 12:05:50
I am signed in
03-04-2013 12:06:45
Oops I'm sorry he put that in the strategy lab, that is in the subscription area. :(
03-04-2013 12:07:54
It doesn't look to me like ProD supports the divergence function
03-04-2013 12:08:08
Hmmm, interesting. I don't remember him putting much in the subscription side.
According to Mobius it is the the measure of two indicators telling you the same thing or if they are diverging, that telling you things are changing.
03-04-2013 12:12:05
The correlation coefficient ranges between -1.0 and +1.0. For perfectly correlated variables (coefficients equal -1.0 or +1.0), changes in the dependent value can be completely explained by changes in the independent value. The correlation coefficient is zero when the explained variation is zero.
03-04-2013 12:12:06
You would basically need to define what rules, over what lengths. In his article he used RSI and Linear Regression Curve.
03-04-2013 12:13:09
Divergence is a function that determines whether two different values have moved in opposite directions over a period of time.
This function returns the following values:
0: The defined values did not diverge (i.e. move in opposite directions).
+1: The first value moved in a positive direction; the second value moved in a negative direction.
-1: The first value moved in a negative direction; the second value moved in a positive direction.
So, clearly they are quite different
03-04-2013 12:29:09
I did a search on divergence on the site and came up with 8 posts in the basic search. Some of the technicals are a measure of divergence. But I can see where that could be a useful function. It should have two inputs for values and maybe a period box. Output would be your +1, 0, -1. Send in your request :)
I will second.
03-04-2013 12:32:26
OK..will do
03-04-2013 12:36:38
This is also something we could probably do the math for. Maybe a bit involved but doable.
03-04-2013 12:46:09
I also was trying to find a place on the site where a calculation could be verified. For instance, if I create a signal that returns the RSI(14), the actual calculation could be run and return a numerical value. Is there such a thing?
03-04-2013 12:48:20
I have been wishing for that for a long time. I have a testing idea that works but is clumsy. I am hoping that we will fairly soon get a Plot node function that would plot a value on a chart.
Or a column in the quotes tab.
03-04-2013 12:54:06
To me that is a serious can one tell if their calculation is being written or handled properly?
03-04-2013 12:55:20
Yeah, Please also send in that request. That is my most wanted item at the moment.
Here is an example of how I have done it in the past.
I forced the answers to be in a stock price range and then let the calc find stocks that answer my question. Like I said clumsy but it does work in a pinch. Mostly I also write code for TOS and compare their charts to my results. Not Ideal either.
03-04-2013 01:06:01
Any robotic system trading real money must have that functionality. Otherwise, bad scripts lead to disaster.
03-04-2013 01:09:22
I'm the choir here :)
I bring this up almost every time I talk to Mike, He says it is coming. They are looking to see if it can be shoehorned into the current charting package or if it has to wait until they can get back to work on the new charting package that they had started but got diverted from.
03-04-2013 01:16:22
Well I would never trade real time if the scripts couldnt be verified independently
ANd it also compromises the bactesting...which I think is amazingly slow...
03-04-2013 01:17:23
I think TOS put a halt on the new charting because it would have significantly increased Prodigios capabilities and I think that they were very worried about that. I think OH is waiting for until they get it running live to give a green light to a lot more projects.
Hmmm, I find backtesting to be fast for the most part. It depends a lot on how many symbols you have in the list and how long a period you are testing.
And I do independently verify the results before I trade live. It is just more work then I would like it to be.
But I did trade live for more then a year.
and did quite well with it.
03-04-2013 01:23:59
an s&p 500 backtest took over an hour
tested for one year
03-04-2013 01:27:07
Right but what else can test that many stocks for a whole year. I generally prove a strat over a short period of time to start. Then a few symbols over a longer period of time. Once I have the strat the way I want I document with a long many symbol test.
03-04-2013 01:27:47
There are other sites that are much faster...Stratasearch for one
03-04-2013 01:37:09
Interesting but it is not doing realtime processing. And I didn't see any mention of what timeframes they are processing. 1 min is a lot of data.
And in real time it is doing 5 sec timeframe resolution. That is a huge increase on that.
And you are having to download the data and process it on your own machine. So that means you need a pretty good machine and a good connection. Even then you will have more latency then Prodigio when you are trading live.
Make a Strategy that only looks a Daily OHLCs and run that, You will see a big difference in the time it takes to process.
03-04-2013 01:53:58
They dont trade live
Thats their major downfall
03-04-2013 02:07:34
stratasearch looks interesting. so am I correct in assuming it's only for backtesting purposes?
similar to Prodigio's BT feature, where you can put in your rules, pick your names to scan with those rules, and it will spit out results where your rules were true for a past period of time?
03-04-2013 02:11:34
Yeah, that is what I see. But I did not see at what resolution it can backtest either. If it is only Daily data then.....
03-04-2013 02:11:45
ahh yes, that's a good point
wouldn't be of much help then
03-04-2013 02:16:00
It is not a real time system but it is much more robust than what I see in Prodigio
03-04-2013 02:16:19
yeah, it's end of day only. no good.
03-04-2013 02:16:46
Put Stratasearch together with a real time plaform and then I think you have a powerful combination
Essentially develop strategy on SS and implement on something like Prodigio
03-04-2013 02:19:17
if it supported intraday data, then it could be awesome. anyone know of any other back testing options that do support intraday data?
03-04-2013 02:20:26
StrategyDesk did/does I think it is still around.
03-04-2013 02:21:15
But when Prodigio doesnt have a calculation function to check code results, or functions like Diverence filters, or WFA, or Monte Carlo analysis, it pales by comparison as a backtest tool.
The hassle is translating SS code into Prodigio.
Does Prodigio permit variable substitution in code?
03-04-2013 02:22:33
it does have some shortcomings. but I have yet to see another BT system that allows for intraday data that's as easy to use
03-04-2013 02:22:58
Thats why Im looking at it
As i said, the two together might be a good solution.
03-04-2013 02:28:44
any of you guys tried backtesting with tradestation?
03-04-2013 02:29:09
What does the code look like, The price isn't bad, it might be a interim solution since there will be more people coming that won't have TOS as a charting alternative.
03-04-2013 02:29:17
it supports intraday data for 15+ years, but I have heard it gives some dodgey results
which I guess isn't that big of a deal, bc Prodigio does too sometimes
03-04-2013 02:30:00
I have not tried TradeStation.
03-04-2013 02:30:54
Bruce, You mean the code from SS?
03-04-2013 02:31:01
I've looked at it before and decided it was lacking something for me, but now I can't remember why
03-04-2013 02:31:03
03-04-2013 02:31:51
crossabove(mov(close, @shortdays, simple), mov(close, @longdays, simple))
the @ symbols indicate variables that the program will sift through to find the best combination
Does Prodigio have that capability?
03-04-2013 02:35:38
Sift through as in a integral type calculation?
Or as in symbols?
Looks like it is a period here.
03-04-2013 02:36:39
In the exaple above, it will sift through an integer range, user defined, to find the best cobination of the long and short moving averages
so for instance, sift through a short ma from 5 days to 30 days against a long moving average of 20 days to 50 days...
Variables are user defined
03-04-2013 02:38:51
No, it does not do that yet. It was requested (long ago) but I have not heard any reports on when/if it might be implemented.
03-04-2013 02:39:50
I was in touch with Dan at Prodigio and he accepted my suggestion re divergence filters
I wonder how long that might take...
03-04-2013 02:40:38
I do that kind of testing by hand, I just do a high/low game to come to a best fit. I guess it could be better though.
I can generally come to a result in 4 tries.
03-04-2013 02:41:55
To me, that capability is awesome....makes strategy construction more powerful
03-04-2013 02:43:02
It also promotes curve fitting, not always a good thing.
03-04-2013 02:43:05
I believe the correct term is optimaztion :)
no doubt
There is also extensive curve fitting protection...
I havnt seen anything as powerful at the price
BUT..there are some support, which is thin
They do however have an extensive reference manual....unlike Prodigio
03-04-2013 02:48:12
You did see how much stuff they added to ver. 5.0 from 4.0
and ver 3.0 to 4.0 was a similar jump in nodes and other goodies.
Mike has assured me that once they get a stable version running Live again that the updates will come at a faster rate again. A lot of this version was a rebuild to get a base to build from that didn't depend on any TOS links.
03-04-2013 10:22:44
Had to work today after 11am but let the system run for an hour this morning. It seems to be working ok. I have noticed however that one version of my main strat has never had a TO or RTS hit even though WL shows many times it should have traded. RR, do you have any strats that act like that?

Reader Comments

There are no comments for this journal entry. To create a new comment, use the form below.

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>