GSB Forums

General support questions.

 Pages:  1  ..  18    20    22  ..  25

B.Wooster - 5-9-2022 at 01:25 AM

> What market is this?
> if you fiddle the data in GSB, it wont match whats in ts
> dax has its issues twice a year too due to daylights savings issues, and bug in ts

The Bund on Eurex (FGBL). Probably applies to any Eurex, LIFFE, or ICE-EU contract with a London opening time.

I was considering opening a TradeStation account, but it sounds like without IBLINK its a pain to manage a portfolio.
I'd rather get NinjaTrader working if possible; the commissions and margin requirements are much better than IB and TS, the programmer has a ton of control, and their support team for programmers are the best in the business. I set the target platform to NinjaTrader and now it is generating Ninja scripts, so thanks for that hint.

admin - 5-9-2022 at 01:34 AM

IB and ts commissions are very low. If you have concerns over margin, probability is you will blow up your account. Until the next major release of GSB NT, there are significant limitations in GSB NT. Im frustrated progress is so slow but cant do anything about that. NT also doesnt have up / down volume which is important but not essential for NQ systems. We are working on using historical data for up down, then live data for short term up down. Thats still a pain.

B.Wooster - 5-9-2022 at 02:08 AM

> Until the next major release of GSB NT, there are significant limitations in GSB NT. Im frustrated progress is so slow but cant do anything about that.

Hmm. Well, maybe I can help. What's the blocker?

> NT also doesnt have up / down volume which is important but not essential for NQ systems. We are working on using historical data for up down, then live data for short term up down. Thats still a pain.

I'm not sure what y'all are doing, but the Up/Down data is certainly available via NT. Heck, I can even get historical and live level2 resting order book data from NT if I want. I don't, I mean I'm not doing HFT on it, so I don't really care about level2 resting orders, but its there. There are tons of footprint and cumulative delta indicators for NT, so saying that the bid/ask aggressor labeling is missing is mystifying. I'm happy to help sort that if you'd like.




admin - 5-9-2022 at 02:24 AM

thanks for the offer, I dont think the data up down volume goes back far in NT.

B.Wooster - 5-9-2022 at 04:31 AM

> thanks for the offer, I don't think the data up down volume goes back far in NT.

Yes. They, NinjaTrader, themselves provide no history. The amount of history depends on the data vendor you choose; that's a separate subscription to CQG, IQFeed, Rithmic, etc.

> We are working on using historical data for up down, then live data for short term up down. Thats still a pain.

I think I'm so used to this, that I don't understand the alternative. It sounds like TradeStation is a bit unusual in that they provide alot of historical data along with realtime, if I'm catching your drift. Please confirm, because I'm probably missing something.

At the moment I'm using data from SierraChart which is very high quality data and very inexpensive for tick data that goes back 9+ years, to generate historical back adjusted bars based on volume-determined rollover. They have a date-rule rollover option too; probably appropriate for some but not all contracts. I don't see a problem with training a GSB model on that data, and then trading that model in NinjaTrader (albeit with CQG as feed; not as good as Denali but still respectable). The only mildly pita thing is that I have to reorder the columns, and delete a couple of them. The data looks like this:


$ head FGBLU22-EUREX.30m.volumebackadj.txt
Date, Time, Open, High, Low, Last, Volume, NumberOfTrades, BidVolume, AskVolume
2022/9/4, 21:00:00, 149.14, 149.19, 149.11, 149.18, 192, 139, 88, 104
2022/9/4, 21:30:00, 149.19, 149.22, 149.17, 149.19, 369, 261, 142, 227
2022/9/4, 22:00:00, 149.20, 149.21, 149.12, 149.12, 331, 235, 188, 143


Is there any easy way to tell GSB the Last == Close, AskVolume == Up, BidVolume == Down, and just to ignore the Volume and NumberOfTrades columns? Right now I run the data through a little script to fix up each file, but it would be nice to be able to feed it to GSB directly.

admin - 5-9-2022 at 06:42 PM

yes, ts provide full historical up down volume on minute data above. On tick data, data (o,h,l,c,v) only goes back 6 months or so.
Im not totally happy about providing systems that need volume to NT users. Even if you have live volume you still need some historial volume for the indicator to work.
I also dont want people chasing me to provide data.
GSB has auto detect data format, but likely it wont work for your unusual situation
Go to tools appsettings data.
Then program the format in

What is " Denali " a typo?
can you send me nq 30 minute data continuous back adjusted from say 5am to 1500 central?
that way I can test my nq systems on sierra data

B.Wooster - 5-9-2022 at 07:26 PM

Thanks for the hint on where to adjust the format. I'll try it.

Thanks for the information about TS's data. I may have to open a TradeStation account now :)

Denali is SierraChart's name for the datafeed that they created to collect data themselves directly from the CME/Eurex, because they were tired of dealing with the crap and poor service from CQG, Rithmic, etc. They run their own Aurora, IL servers colocated with the CME. It is high quality tick data. I suspect they are backfilled with the BarChart data they used to provide before a couple of years ago when they stood up Denali though.

Attaching the SierraChart NQ data you requested. The Up/Down volume started in 2011, so that's when the file starts. Timestamps are in Chicago timezone.

Attachment: Login to view the details


B.Wooster - 5-9-2022 at 07:38 PM

Addendum: For that NQ data, I used volume-based rollover points, that SierraChart determined. I've never really trusted the rollovers used by vendors. IQFeed for instance used to be super bad at this, using the actual expiration of the contract date as the rollover point. FuturesTrader71 worked with them to fix this some time back, but I still would be wary of their continuous backadjust '@NQ#C' contracts, for example, and double check them oneself before relying on them. I have no idea if TradeStation does a reasonable job of rollover points or not. I am curious though.

Anyway, this is just to say, if you see data discrepancies between the SierraChart data and the TradeStation data near the rollpoints, I'd be very interested to know about that. Of course you'll have to compare to data that was back-adjusted starting from today (2022 Sept 5 in the September 2022 NQ contract) to be able to compare price at all. Older data whose most recent point was the June NQ contract will of course have zero chance of matching price-wise, because of the back-adjusting.

admin - 5-9-2022 at 07:57 PM

Im passing this info to my NT programmer. A big thank you

B.Wooster - 5-9-2022 at 08:06 PM

Anyone looking at that NQ data, here's my hacky cygwin bash shell command to fix it up for consumption. There doesn't seem to be any way to tell GSB to ignore or rename columns in app settings->Data.

$ cat NQ.30.minute.volumebackadj.2011_2022.0500_1500.from.sierrachart.denali.txt | sed 's/Last/Close/' | sed 's/BidVolume/Down/' | sed 's/AskVolume/Up/' | tr -d '\r' | awk -F, '{print $1 "," $2 "," $3 "," $4 "," $5 "," $6 "," $10 "," $9}' > NQ.30.minute.volumebackadj.2011_2022.0500_1500.from.sierrachart.denali.stadard.format.txt


for instance:
$ cat NQ.30.minute.volumebackadj.2011_2022.0500_1500.from.sierrachart.denali.txt | sed 's/Last/Close/' | sed 's/BidVolume/Down/' | sed 's/AskVolume/Up/' | tr -d '\r' | awk -F, '{print $1 "," $2 "," $3 "," $4 "," $5 "," $6 "," $10 "," $9}'|head
Date, Time, Open, High, Low, Close, Up, Down
2010/12/28, 05:00:00, 2306.75, 2308.25, 2306.75, 2308.00, 71, 90
2010/12/28, 05:30:00, 2308.00, 2308.75, 2307.75, 2308.50, 53, 40
2010/12/28, 06:00:00, 2308.25, 2309.00, 2308.00, 2309.00, 53, 46
2010/12/28, 06:30:00, 2309.00, 2311.00, 2308.50, 2309.50, 313, 162
2010/12/28, 07:00:00, 2309.50, 2310.00, 2308.50, 2310.00, 193, 165
2010/12/28, 07:30:00, 2310.00, 2310.00, 2307.75, 2307.75, 128, 504
2010/12/28, 08:00:00, 2307.50, 2308.50, 2307.50, 2308.25, 152, 160
2010/12/28, 08:30:00, 2308.25, 2308.75, 2308.00, 2308.25, 137, 196
2010/12/28, 09:00:00, 2308.00, 2311.00, 2307.75, 2310.25, 959, 1125

B.Wooster - 5-9-2022 at 10:06 PM

Quote: Originally posted by admin  
Im passing this info to my NT programmer. A big thank you


You're welcome. I added a few lines to the GSBDataExporter indicator NinjaScript, and now it exports Up/Down volume too. See attached.

Attachment: Login to view the details


admin - 5-9-2022 at 10:51 PM

NT program said "I did not want to use volumetric bars because they are part of Ninjatraders built in Order Flow.
That is only supplied to NinjaTrader Lifetime Licenses which is $1099. So that solution would not work for a lot of users."

B.Wooster - 5-9-2022 at 11:04 PM

I compared the NQ 30m bars from Sierra with the TradeStation default NQ file that ships with GSB. Plot attached. Here is a summary of the Close price differences from N=56930 bars.


Min. 1st Qu. Median Mean 3rd Qu. Max.
-417.00 -39.50 -35.25 -34.05 -29.50 420.25


They look pretty similar. The blue line overlaps the black line almost exactly.

compare_tradestation_sierra_backadjustment.PNG - 120kB

B.Wooster - 5-9-2022 at 11:13 PM

> NT program said "I did not want to use volumetric bars because they are part of Ninjatraders built in Order Flow.
That is only supplied to NinjaTrader Lifetime Licenses which is $1099. So that solution would not work for a lot of users."

I'm speechless. To quote the cartoon Archer, "this is why we can't have nice things."

Obviously this is a choice that you've got to make.

If it was me, I wouldn't hesitate to require it. GSB is more expensive than a NT Lifetime license. Neither are for dabblers in any case.

admin - 6-9-2022 at 05:12 PM

Hi Bwooster
I don't use NT, so this is not relevant to me. I have little hesitation spending $ where needed, but only some NT users will do this. Likely we will support a few options.
Can show you the performance reports on the two nq data sources. Make the start date the same on both bits of code so we can compare apples with apples.
The graphs certainly look great

portfolioquanttrader2020 - 7-9-2022 at 09:47 AM

Hi Peter,
How is the secondary filter of the built in indicators different from the SF Indicator and the Secondary Filter Mode on the left side of Gsb?

portfolioquanttrader2020 - 7-9-2022 at 10:34 AM

When I build the nasdaq it doesn't take me 75 indicators, it only takes me 52. What is the reason?

portfolioquanttrader2020 - 7-9-2022 at 10:36 AM

I have 150 indicators and I don't have 180, why?

portfolioquanttrader2020 - 7-9-2022 at 11:25 AM

when I start workers this error appears in the photo and it does not build the manager with all the workers

GSB.JPG - 23kB

admin - 8-9-2022 at 12:48 AM

@portfolioquanttrader2020, you must have at least one (and only recommend one at a time ) secondary filter enabled in gsb

portfolioquanttrader2020 - 8-9-2022 at 04:40 PM

hi peter
Why in the new methodology of the NQ, no data is entered from the year 2002?

admin - 8-9-2022 at 05:01 PM

That is very deliberate. System building results were not better if 2001 was used compared to 2007.1.1. But 2001 takes much longer as more data to process.
ES however you can use 1997, 2000 or 2007 as start date

No data in 2022 gives further confirmation if a system has gone well, or not. Also the profits in 2022 are so high, it will force the system to curve fit for conditions of extreme volatility like we have now.
When those conditions dont exist, system will do poorly.
I had the same issue in 2008 after the end of global financial crisis.

portfolioquanttrader2020 - 12-9-2022 at 10:01 AM

Hello
When I build a system, which system do I take, the code with wfo, or the code without wfo?
I put images

imagen_2022-09-12_110137077.png - 119kB

admin - 12-9-2022 at 04:50 PM

'portfolioquanttrader2020, I always use the wfo code.

Mark2022 - 15-9-2022 at 04:57 AM


Hi all, I am new on the forum.

Before I start evaluating GSB and building strategies, I would like to ask fellow traders how to best use different strategies on the same ticker symbol? I am a Tradestation user.

One workspace, 1 @ES chart and 5 strategies on the same chart?

Or one workspace, 5 separate @ES charts and 1 strategy per chart?
Do I need to use unique names for the different entry signals, so the trades of different strategies don't "overlap", so are independent of each other?
What about the stoploss used?

Thanks in advance and have a great day!
Mark



admin - 15-9-2022 at 05:16 AM

Hi Mark
Welcome.
I recommend one trading system per chart.
Multiple strategies is fraught with issues, esp exits and limiting a chart to say 3 contracts when you have 5 systems is a idea that sounds good, but is really bad when you look into it.
Its been discussed in the last few weeks somewhere in the forum.
also some markets can have multiple valid session times, which wont work with different systems on the chart.

admin - 19-9-2022 at 11:28 PM

Does anyone have a i9 cpu (prefer fast one) that I can do a benchmark test on. Need it for < 30 minutes via anydesk.com

portfolioquanttrader2020 - 25-9-2022 at 02:39 AM

Hello
1. Better to group strategies by market and timeframe in a workspace.
2. I think that using different names can be done through code as long as you identify the input signal with an ID, and also put it to the output signal in the code of each system. But if you use Automation it is simpler.
3. The really important thing is that the Tradestation Trademanager puts in all the True positions, even if the stop loss is executed.
4. Using Alertmon you can also see that everything is true.
5. My advice is to use Automation + Alertmon, to have control over contingencies.
6. I could help you with operational management, if you need.

admin - 29-9-2022 at 08:39 PM

The most major thing thats stopped errors in execution for me is my monitoring on VM service that checks every chart + health of the machine.
https://trademaid.info/gsbhelp/AlertmonMonitoring.html

For the first time in my trading career, I got woken up due to lack of ram on the server.
I have 32 GB of ram, ts 9.5 and about 4 copies of trader workspace. (5 gb of ram in total)
Most user should get away with 4GB of ram.

Today I got a email from another VM user that his ram is low. 8GB but less charts.
please watch out for this issue. Nothing has changed on both trading computers.
I know there is a rare bug thats impossible to reproduce that ram blows out while optimizing ts. When ram hits the 3gb (rough 32 bit limit) ts crashes. I havnt seen this bug for about 3 years.

ts-ram.jpg - 235kB

admin - 30-9-2022 at 03:32 AM

if crashing due to ram is an issue, best post here too.
https://community.tradestation.com/Discussions/Topic.aspx?To...

admin - 30-9-2022 at 03:44 AM

ALERTMON HAS BEEN UPDATED, SMALL TWEAKS STILL NEEDED BUT IT WORKS

NEWALERTMON.jpg - 159kB

admin - 30-9-2022 at 03:58 AM

user from ts forum said
I'm on Windows Server 2012 R2 on AWS EC2 t3.medium instance.
And after the windows update last sunday I started to see the problems.
Giuseppe

admin - 30-9-2022 at 04:01 AM

These are my suspect cause


updates2.jpg - 82kBupdates1.jpg - 61kB

admin - 30-9-2022 at 05:49 AM

further updates to alertmon
sms field added
no more alarms for 30 / 60 minutes as default after the first alarm

NEWALERTMON2.jpg - 146kB

admin - 30-9-2022 at 07:04 AM

i removed most recent windows updates on one machine. took 30 minutes roughly to finish starting up

admin - 30-9-2022 at 05:20 PM

This is my ram useage of 48 gb machine. Clearly an issue

mem-afterclose-graph-bt.jpg - 114kB

admin - 6-10-2022 at 01:10 AM

my machines are now stable as far as memory usage goes, which i put down to removal of windows service pack.
Every version of windows server has its own patches, but the problem seems common on all builds of windows server.
Will advise when I have more solid info.
SOme nice updates to alertmon.
Graph for max orchart.exe and we save the memory useage of orchart.exe in csv, so we could track down the day the error started

Daniel UK1 - 18-10-2022 at 02:21 AM

@Peter,

Hope you are well.
I just wanted to ask about if you have an ETA for VWAP and the other one i sent you code for in Feb/March?

I assume not 100% of dev resources goes into NT anymore?




admin - 21-10-2022 at 05:15 AM

Hi Daniel
As mentioned previously, this will not happen until GSB NT is done. While I know your in a hurry for this, asking will not help at all. Lead programmer is stalled in hours due to family situation. Programmer number 2 is still very active, but thats on alertmon, ewfo, pa etc and just the odd GSB functions.
Im completely tied up in the last week on TS memory leak issue, and got nothing else done apart from getting most of the systems on the current set of brokers.

Daniel UK1 - 21-10-2022 at 08:31 AM

Got it Peter. I have no knowledge of how far ahead you are on the GSB NT roadmap, or the issues you are facing with TS memory leaks, so just wanted to check in on the matter. Sounds you have lots of things on your plate, I hope it get resolved for you soon, issues like that is a pain.. :thumbup:

admin - 21-10-2022 at 05:25 PM

Hi Daniel
here is an update.
This is how bad the memory leak is.
This is new win10 install, no windows updates since the os was installed. (disabled them was the first thing I did)
Ive run out of options for now.
This is ts 9.5 which has had no updates for a year.
Bottom line is I can live with the memory leak, but I had to upgrade ram of all my vm users to 8GB or more, and reboot every 1 or 2 days

The nt work is really close to running, but programmer has done very little work for some time, apart from a few tech support / bug fixes. While this frustrates me, if GSB never got better - i could live with it. My time is 100% tied up with trading and driving programmer #2, and the market conditions are roughly equal best to the two other most glorious periods in my trading career. (2007/2008 03-04/2020)
We are still in a very good place that GSB as is works great, and some users are using GSB in slightly different ways with really good results.
GC and or CL settings will come when I have time to validate them.
I did a quick review on GC recently with great stats, but didn't get time to look how the systems went.

memleak.png - 127kB

portfolioquanttrader2020 - 23-10-2022 at 02:10 PM

Hello, I've been looking for robust raw systems for months and I can't find good systems.
Could someone experienced give me a clue about the configurations of CL, EC and Soybean?

Daniel UK1 - 23-10-2022 at 03:50 PM

I believe you will struggle with EC and s.. try NG, GC and CL, .. CL not best time this year, but systems seems fine, so does GC.. You will see that Short side on most systems performs much better...

admin - 23-10-2022 at 04:06 PM

I think EC is now possible, but I dont yet have the finer details. NQ settings on @ec session where the volume is.
The short profitability bias may also be related to the secondary filter used. CloseToHighLow3v4 tends to have a short bias on NQ. Not sure why

portfolioquanttrader2020 - 23-10-2022 at 10:52 PM

Hello Daniel
Could you tell me CL configuration?

Daniel UK1 - 24-10-2022 at 04:43 AM

Quote: Originally posted by portfolioquanttrader2020  
Hello Daniel
Could you tell me CL configuration?


Not sure its that easy, I spent the last 9 months doing research on CL alone, you have multiple possible configs to run.

But first, since you ask the question, what have you yourself arrived to in your own research ? and i will be happy to point you in "the right" direction and tell you what to change or not.

But regardless, you will find 2 SESSIONS that work well, but they will use same trading time, 10-1430.

You will also find that these two sessions prefers different data, i.e different bar sizes and.

So test using 2-3 data, same market.

CL is the most complex market to do research on due to the fact that it offers the most amount of good combinations.

One thing to note on CL, build training filter values will affect OOS greatly. )PF and Pearsson



(If i was you I would start with Peters shared excellent guides for CL)

portfolioquanttrader2020 - 24-10-2022 at 08:23 AM

Hello I have achieved good results with schedule 630 1330 and secondary filter equal to NQ.
But I would like to improve the profit factor.
Out-of-sample results are good, with no out-of-sample losing periods.
Any more clues?

Daniel UK1 - 25-10-2022 at 06:47 AM

Quote: Originally posted by portfolioquanttrader2020  
Hello I have achieved good results with schedule 630 1330 and secondary filter equal to NQ.
But I would like to improve the profit factor.
Out-of-sample results are good, with no out-of-sample losing periods.
Any more clues?


Thats good,
In my research i have not found 630 session start, any good, see my screenshot of my test of session times, please note its session times and not trading times.

In regards to data, see included screenshot of comparison between using data3 compared to not using data 3 on this "late" session, 9-1430
Top is using data 1, 2 and 3, bottom is data 1 and 2 only.

Perhaps you get some clues from this and can test against your own data.

My macro for stats is not the same as yours for the periods. Bug G period stands for OOS up until dec 2021.

Also note, your training values used for building will affect your OOS.
My tests is done on a large set of fixed indicators, and not opt 10, but for me this does not matter since its only relative to itself during the research, and I dont care about max performance actual numbers, until final setup for build live systems.



dataScreenshot 2022-10-25 084011.png - 251kB Screenshot 2022-10-25 083544.png - 123kB

portfolioquanttrader2020 - 25-10-2022 at 07:47 AM

Hello Daniel
Do you build that table or is it exported in that GSB format?

Daniel UK1 - 25-10-2022 at 07:57 AM

Quote: Originally posted by portfolioquanttrader2020  
Hello Daniel
Do you build that table or is it exported in that GSB format?


Its done by a python app, shared to me by a kind soul, for me to export from GSB json, and then option to veiw in app, or output it to excel like this for manual review.


portfolioquanttrader2020 - 25-10-2022 at 11:19 PM

Questions
I understand that GSB does not have the option to configure breaks or breakouts, correct?
I would like to know what test I have to do to find the best secondary filters in a market?

admin - 26-10-2022 at 12:13 AM

We did brute force to test secondary filters but the closelessClosedbpv, highlow3, RSI and DeCyclerOscillator are the firsts ones I would try
GSB doesnt have stop entries, and its very over due to get them.
However you can add stop entry in any gsb system after its built. It made systems like gsbsys1es worse using stop entries, but they are a good idea on some markets

Daniel UK1 - 26-10-2022 at 03:32 AM

Quote: Originally posted by portfolioquanttrader2020  
Questions
I understand that GSB does not have the option to configure breaks or breakouts, correct?
I would like to know what test I have to do to find the best secondary filters in a market?


Pick top 30 ish indicators for a market, (based on a std sf) then run those on ALL sf, get stats of SF, then pick top 10, then run individual tests on each of those SF using same top 30 ish main ind, after this you have a pretty good idea.. actual results not interesting, only interesting to view relative to each other.. you could run top 10 main ind on each to further evaluate if results are very close and hard to pick

portfolioquanttrader2020 - 26-10-2022 at 07:32 AM

Hello Daniel
Sorry but I don't quite understand your explanation
Could you post a photo with settings?

portfolioquanttrader2020 - 26-10-2022 at 08:23 AM

Hi
Several important things about builder: the statistics are from the 300 systems???

On the other hand, the 300 are Favorites A, the best in sample? and Favorites B,C,D, are out of the sample???

imagen_2022-10-26_092428182.png - 34kB

Daniel UK1 - 26-10-2022 at 12:15 PM

Quote: Originally posted by portfolioquanttrader2020  
Hello Daniel
Sorry but I don't quite understand your explanation
Could you post a photo with settings?


Take a manager, pick your market, take an SF that is commonly used... for example for CL you can use close less prev low, build 50k, put top 300 in A, then do stats and look what the top 30 ish indicators, save those main indicators... then take a new manager using these top 30 indicators, and turn on ALL SF, then run same again, and this time save stats for SF, look and note down top 10 SF.. and then run stats on each of these SF using same top 30 main indicators.. this should give you a good idea what to test further on in regards to SF

Daniel UK1 - 26-10-2022 at 12:22 PM

Quote: Originally posted by portfolioquanttrader2020  
Hi
Several important things about builder: the statistics are from the 300 systems???

On the other hand, the 300 are Favorites A, the best in sample? and Favorites B,C,D, are out of the sample???


That would depend on the macro you are using, if you look at the macro it will tell you the answer to your question, or share it and we can have a look.

Out of curiosity, what dates is your ABCDE etc ?



portfolioquanttrader2020 - 26-10-2022 at 11:36 PM

are the same as the Nasdaq with new methodology

Daniel UK1 - 27-10-2022 at 01:41 AM

Quote: Originally posted by portfolioquanttrader2020  
are the same as the Nasdaq with new methodology


I dont know what your macro you are running does, please open and look in your macro you used and share the dates for ABCDE etc please

Your numbers in your table you shared looked quite interesting for CL, but not possible to fully know unless you share the dates, and build date end.


portfolioquanttrader2020 - 27-10-2022 at 08:08 AM

A 20170301------------20180228
B 20180301------------20190228
C 2019301-------------20200229
D 20200301-----------20210228

Daniel UK1 - 27-10-2022 at 09:20 AM

Quote: Originally posted by portfolioquanttrader2020  
A 20170301------------20180228
B 20180301------------20190228
C 2019301-------------20200229
D 20200301-----------20210228


Thanks for sharing, somehow I am less confident that your stats shared reflects that ? Perhaps I am wrong, but i would be very interested in your setup, if for example between 2018 03 and 2019 02 you delivered avr 90k in profit in CL

I think that perhaps there is a mistake somewhere in those stats, if these dates really is what you intend to populate stats from ?

Or perhaps i am wrong.. thanks for sharing regardless

portfolioquanttrader2020 - 27-10-2022 at 10:42 AM

I am seeing that the highest volume session in crude oil is 730-1330

imagen_2022-10-27_114312236.png - 71kB

Carl - 28-10-2022 at 02:10 AM

Please be aware to de-activate the TRIX indicator when building on CL or S, because TRIX is log based and does not like negative prices.

portfolioquanttrader2020 - 1-11-2022 at 03:24 AM

Hello
I need to know where the red, orange or green indicators are explained in a very detailed way.

admin - 3-11-2022 at 01:43 AM

@portfolioquanttrader2020, its been years since I looked at this, so im less familiar with it now
It works on the top 30% by ration vs the bottom 30% by ratio
and how often its used.
I dont not recommend to look into this more. I have and the results were not consistent enough to justify any changes to what works already

gui-parms.png - 61kB

TradingPrice - 3-11-2022 at 04:19 PM




1- why test Long & short VS long only or short only?

2- why not test from 2010 instead of 2007 (to avoid 2008)?

3- across different markets you've tested do you get better results out of adding secondary data (eg. ES with SPX)? as NQ systems work well with only primary data

4- have you tested basing session time on how favorable market going up or down during certain hours of the day (bias systems)? eg. ES rising overnight vs during cash session

Thanks

admin - 3-11-2022 at 04:26 PM

Excellent questions
1) depends on the market if its valid. long or short only reduce trades by 50% so its less robust. Some markets like ES swing systems like long only, but the systems would get hammerd right now in bear market
2) In 2008 to 2009 this was a critical mistake I made. Systems curve fitted to 2008 with bad results
Now with much more data after the date, it doesnt matter + we are now in 2008 conditions.
SPent a lot of time on dates.
so for es 1997 , 2001 and 2007 all valid start dates.
nq 2001 or 2007 give almost identical system results
gc 2007 start
3) Ive tested this in generally nothing helps, but it might depend on the market and other users would be welcome to test
GSB automation spreadsheet has the option to count how often data1 vs data2 is used.
If you dont look at how often data2 is used, so can come to wrong conclusions. ie one test might be better, but data2 was rarely ever used
4) A bit. ES had strong overnight up bias, but the concept getting killed right now.
default nq es setting work great

SwedenTrader - 4-11-2022 at 05:30 AM

Quote: Originally posted by TradingPrice  



1- why test Long & short VS long only or short only?

2- why not test from 2010 instead of 2007 (to avoid 2008)?

3- across different markets you've tested do you get better results out of adding secondary data (eg. ES with SPX)? as NQ systems work well with only primary data

4- have you tested basing session time on how favorable market going up or down during certain hours of the day (bias systems)? eg. ES rising overnight vs during cash session

Thanks


Regarding the last point. These systems for sale in the bottom called "swing daily" always Enter at intraday session-end and exit at intraday session-open next day, have a tendency to catch the "Long bias" overnight even in a bear market. Great complement to daytrading systems. ES at new highs right now, probably the others too.

Overnight-edge.png - 48kB

Daniel UK1 - 4-11-2022 at 01:17 PM

Quote: Originally posted by TradingPrice  



1- why test Long & short VS long only or short only?

2- why not test from 2010 instead of 2007 (to avoid 2008)?

3- across different markets you've tested do you get better results out of adding secondary data (eg. ES with SPX)? as NQ systems work well with only primary data

4- have you tested basing session time on how favorable market going up or down during certain hours of the day (bias systems)? eg. ES rising overnight vs during cash session

Thanks



Imho,
about 3, many markets could certainly be helped by using a secondary other market, BUT, and its a bit BUT, you add something to your mix, that can break. You add the dependency of correlation.

Also, you cant test this properly, until GSB allows you to force the usage of data.

admin - 4-11-2022 at 03:34 PM

@daniel
Forcing is problematic that you may need a combination of data streams, not an alternative data stream
I still like just looking out how often data2 was used in the top 300 systems

REMO755 - 18-11-2022 at 04:39 PM

Hello, why doesn't TS work?

Value1 = GSB_Scripts_2021_11_11;

GSB_SCRIPTS_2021_11_11+WITHUPDATEDGSBSYS1ES_V1.21




REMO755 - 18-11-2022 at 04:39 PM

Hello, why doesn't TS work?

Value1 = GSB_Scripts_2021_11_11;

GSB_SCRIPTS_2021_11_11+WITHUPDATEDGSBSYS1ES_V1.21


???


error.JPG - 21kB error.JPG - 21kB

RandyT - 19-11-2022 at 02:47 PM

Peter,

I have to respond to this. I do not understand the continued resistance to allowing GSB to specify whether to use all data streams provided or accept current behavior of allowing GSB to use GA to choose.

It absolutely does make it impossible to accurately test the use of secondary data. If I configure a run in GSB to use secondary data, I expect it to use it. The behavior of GSB not using the configured data is very much unexpected and violates the principle of least astonishment.

https://en.wikipedia.org/wiki/Principle_of_least_astonishmen...

Quote: Originally posted by admin  
@daniel
Forcing is problematic that you may need a combination of data streams, not an alternative data stream
I still like just looking out how often data2 was used in the top 300 systems

Bryan - 20-11-2022 at 04:17 AM

Hi . Is there any guide for NG systems like which Secondary filter and entry type?

admin - 21-11-2022 at 02:47 AM

Quote: Originally posted by Bryan  
Hi . Is there any guide for NG systems like which Secondary filter and entry type?


Good question
vSF = GSB_CloseToHighLow2v2(iSFLength) of Data(iSFData);

// entry type = Compare2

Chituan - 22-11-2022 at 04:35 PM

Hi. I just downloaded the trial version. I went through the tradestation (EasyLanguage) code and Ninja trader (C sharp) code for the free GSBSys1ES system. They seem to be different and not 100% the same. The main signal is still Stochastic and AccumDist but the weights and normalization seem to be a bit different if I am not mistaken. Can you please confirm and explain why we have different logic for same system ?

Thanks

admin - 22-11-2022 at 04:36 PM

@Chituan, I will look into this and get back to you. Need to ask the NT programer

Bryan - 25-11-2022 at 03:13 AM

I dont have GSB_CloseToHighLow2v2 in my gsb (trial) platform ? However I have it under tradestation functions .


Quote: Originally posted by admin  
Quote: Originally posted by Bryan  
Hi . Is there any guide for NG systems like which Secondary filter and entry type?


Good question
vSF = GSB_CloseToHighLow2v2(iSFLength) of Data(iSFData);

// entry type = Compare2



Untitled.png - 114kB

admin - 25-11-2022 at 03:18 AM

Hi Bryan
GSB_CloseToHighLow2v2 is not used a lot as a secondary filter
Are you looking at NG or es ym nq etc? These markets use GSB_CloseToHighLow3V4
The trial version has a reduced indicator set.
IM doing some tests on NG now and will have much better idea of what secondary filters are options next week. There are a number that will work well

Bryan - 25-11-2022 at 05:12 AM

Yes im looking at natural gas, should be a very volatile market this winter !

Chituan - 25-11-2022 at 11:40 AM

Hi Peter, Unfortunately TradeStation is not working in my country so I spent last couple of days trying to replicate the free ES Sys1 that comes free with GSB trial. I used the data provided inside GSB: es.1.minute.1997_20210331.Central.VolisTicks.txt which I have processed into 30-min bar data as you can see in my attached picture.

I am not able to replicate the Equity Curve or Trade List that I see here and the general result is also vastly different: https://trademaid.info/systems.html

The last few rows of my calculation is attached with the normalized fastK and normalized accumDist included . Is it possible I can get your normalized fastK and normalized accumDist for these rows so I can check ?
I really appreciate this since I want to do some sanity check before going on to purchase other systems.

Thanks for your time

Chituan



Screenshot 2022-11-25 173348.png - 92kB

admin - 25-11-2022 at 04:47 PM

@Chituan are you on muliti charts or Ninja?

admin - 25-11-2022 at 05:11 PM

Quote: Originally posted by admin  
Hi Bryan
GSB_CloseToHighLow2v2 is not used a lot as a secondary filter
Are you looking at NG or es ym nq etc? These markets use GSB_CloseToHighLow3V4
The trial version has a reduced indicator set.
IM doing some tests on NG now and will have much better idea of what secondary filters are options next week. There are a number that will work well


use CloseLessPrevCloseDBpv with entry type compare2

Chituan - 26-11-2022 at 02:32 PM

Quote: Originally posted by admin  
@Chituan are you on muliti charts or Ninja?


I think I managed to replicate it now using. Thanks Peter.:)

dhillo - 26-11-2022 at 03:59 PM

I am trying to upload stock price data to GSB from Tradestation following the steps on the site. I added the symbol to the contracts list using the settings used for SQQQ (I am uploading TQQQ)

This is from the file:
"Date","Time","Open","High","Low","Close","Up","Down"
11/26/2012,09:31,1.05,1.05,1.05,1.05,4027776,1204032
11/26/2012,09:32,1.05,1.05,1.05,1.05,160560,533280
11/26/2012,09:33,1.05,1.05,1.05,1.05,56160,55200
11/26/2012,09:34,1.05,1.05,1.05,1.05,206400,329040

I made a session time setting from 930 to 1600 and I try to import from the Opt Price Data field in the left and I set the session and click OK but it doesn't put the info in the field as it does with the default selections. I don't know what I am doing wrong.

Chituan - 27-11-2022 at 03:28 AM

Hi Peter, I am following your video here https://www.youtube.com/watch?v=QYg5eH2Q1_s . After step 1, running 50,000 systems, My statistics table is empty. I attached the screenshot below of GSB Manager and Worker. Can you please let me know why ?



Thanks



Daniel UK1 - 27-11-2022 at 03:50 AM

Quote: Originally posted by Chituan  
Hi Peter, I am following your video here https://www.youtube.com/watch?v=QYg5eH2Q1_s . After step 1, running 50,000 systems, My statistics table is empty. I attached the screenshot below of GSB Manager and Worker. Can you please let me know why ?


I am unsure if that specific macro saves any statistics by looking at your screenshot.
If you manually go to systems tab, and right click and chose save to statistics, you will then see the stats.

You need to look in the macro you are showing and see if there is a line where it saves anything to stats, does it do that in the macro you are showing ?

Imho, get to know and understand Peters methodology in the video if that is what you want, and actually learn and understand what each, each step want to accomplish and why. After that, understand and learn to do it manually, and when you understand and know how to do that, first then, and not before, use the macro to automate what YOU want to do.

Thanks

portfolioquanttrader2020 - 27-11-2022 at 07:43 AM

Hello
I have been building oil only Friday and Wednesday, and by the time you build systems two or three days a week, the systems are not replicated to TradeStation
I think there might be a bug.

imagen_2022-11-27_074338185.png - 164kB

portfolioquanttrader2020 - 27-11-2022 at 09:28 AM

Hi, I have an important question regarding optimization.
If TradeStation only limits the tests to 8,000, if we do an optimization of 200,000 tests, 292,000 would be eliminated.
So, if we do wfo it would not be valid?
It is right?
Multicharts does not limit the tests.

image (3).png - 34kB

portfolioquanttrader2020 - 27-11-2022 at 03:09 PM

In the video that Chituán comments, in that case you have to save statistics manually.
In oil the macros are for configuration.

admin - 27-11-2022 at 04:11 PM

@portfolioquanttrader2020
what you show there has nothing to do with ts writing out files for ewfo/wfo.
Its only the amount of results you see when looking at optimization results displayed. It doesn't limit how many optimizations are run. Ts will in this list show the highest (up to) 8000 results. you dont need more.
With a Walk forward, TS doesn't limit tests to 8000. If there is only say 2000 tests by brute force, I would often use brute force.
Otherwise aim for the GA amount to be 10% of the brute force amount. Often this is UN-realistic and so we have to use less %

I dont put generations over 100, but increase population if needed

JozefSusko - 21-12-2022 at 01:45 AM

Hello Peter,

I want to ask you if you could help me.
I have a problem when my strategies are traded on a holiday, don't exit (half a day, SetExitOnClose).
Can you please help me fix this?
Strategies created by GSB.

Thank you
jozef

admin - 21-12-2022 at 01:50 AM

a number of the new systems for sale have this code added. I will give an example of this tommrow. Just stopping work for the day. I intend over time to add this into all day trade systems

goodoboy - 21-12-2022 at 07:50 AM

Hello Everyone,

Glad to be here.

May I have help some help please? I am studying the GSB documents before buying GSB. https://trademaid.info/gsbhelp/ImportingastrategyfromGSBtoNi...

I am a NinjaTrader8 user as my trading platform and would like to build systems in GSB and run the systems in NinjaTrader8.

Questions:

1. Can GSB build systems and the systems run reliable in NinjaTrader8? I just want to confirm for my understanding before purchase. I am not experienced with Tradestation.

2. Can I import tick data in GSB to test systems using Tick Bars and Range Bars? Or does GSB only work for minute bars?

Thank you so much.

Daniel UK1 - 21-12-2022 at 02:21 PM

Quote: Originally posted by JozefSusko  
Hello Peter,

I want to ask you if you could help me.
I have a problem when my strategies are traded on a holiday, don't exit (half a day, SetExitOnClose).
Can you please help me fix this?
Strategies created by GSB.

Thank you
jozef


My 2 cents, Sometimes the easiest solution is the "best"
In my own trading for the past 7 years, on half days, i just pause all systems, restart servers, and see it as a perfect time to relax and enjoy some red wine.

No code needed :)


engtraderfx - 21-12-2022 at 05:19 PM

i like that solution too! (especially the vino;) )..

If interested have seen a similar thread which uses a file to log holiday dates, can either prevent any trade on that day or add a line to exit a pre-determined earlier time. Haven't implemented yet, i use above :lol:

https://capstonetradingsystems.com/2020/01/02/tradestation-f...

admin - 21-12-2022 at 10:51 PM

Thanks for both of your inputs on the topic. Always good to discuss things.
There are lots of issues here.
What I do agree on is rest with some red wine can be a good thing.
the code given is 251 lines of code. That means best case every system is doing 251 extra calculations on the first bar of the day (or every bar)
For live trading it needs to be truncated. ie remove old dates.
Different markets have different holidays, and the metals/energy have less holidays, and when the holidays occur, we often still have the entire trading day for energy / metals.
Metals / energy have from memory been very profitable in this last year on 'holidays'
After Nov 30th my most profitable day in 22 years (interest rate news), im not keen to stop trading any day the markets are open on the entire session of what I trade.
If the marker closes earlier than my normal exit time, I wont trade it.
This is my code.
Its simple and written for execution speed.

Its 2 calculations only on the instant a trade occurs. No over head on any other bars.

Personally, I will not shutdown trading servers, unless I have alarms auto set to tell me when the markets open. Human error is a big factor in trading, and ive heard too many stories of, I forgot to put ts online and .....:(
Hence why i have alertmon

,Holiday1(20221124{DontTrade1/2tradingday}),Holiday2(20221125);


hol1=holiday1-19000000;
hol2=holiday2-19000000;
and date <> hol1 and date <>hol2// this skips all the entry code if there are no entries

and date <> hol1 and date <>hol2// this skips all the entry code if there are no entries


holidaypart1.png - 228kB holidaypart2.png - 184kB

JozefSusko - 22-12-2022 at 02:24 AM

Quote: Originally posted by Daniel UK1  
Quote: Originally posted by JozefSusko  
Hello Peter,

I want to ask you if you could help me.
I have a problem when my strategies are traded on a holiday, don't exit (half a day, SetExitOnClose).
Can you please help me fix this?
Strategies created by GSB.

Thank you
jozef


My 2 cents, Sometimes the easiest solution is the "best"
In my own trading for the past 7 years, on half days, i just pause all systems, restart servers, and see it as a perfect time to relax and enjoy some red wine.

No code needed :)



Yes, that's how I did it until now, turned off TS, but with TMA it's a little different, because the last time I did it, the next day when I turned on TS, the store opened automatically.

goodoboy - 23-12-2022 at 12:15 AM

Quote: Originally posted by goodoboy  
Hello Everyone,

Glad to be here.

May I have help some help please? I am studying the GSB documents before buying GSB. https://trademaid.info/gsbhelp/ImportingastrategyfromGSBtoNi...

I am a NinjaTrader8 user as my trading platform and would like to build systems in GSB and run the systems in NinjaTrader8.

Questions:

1. Can GSB build systems and the systems run reliable in NinjaTrader8? I just want to confirm for my understanding before purchase. I am not experienced with Tradestation.

2. Can I import tick data in GSB to test systems using Tick Bars and Range Bars? Or does GSB only work for minute bars?

Thank you so much.


Hello Everyone,

Happy Holidays,

Just checking in if anyone can help me with these 2 questions please.

Thank you kindly.

admin - 23-12-2022 at 02:25 AM

@goodoboy Sorry i missed your last post.
1) The trial version of GSB is working with NT, but missing some functionality.
The beta version of NT has much more functionally but not ready yet. Its the number one focus of lead programmer once some critical bugs are fixed for some other users.
2) GSB works with tick date etc, and I suspect in theory tick data is much better.
In practice tick data is really problematic. Many data providers give only short tick data history. Thats often not usable as the data is too short and or the sample size too small to make working systems

Trading on a different data provider to your historical data also is very problematic.
The GSB users who I consider experts in this field, most have given up. Not a GSB issue but data feed issues.
Please read this
https://trademaid.info/gsbhelp/RenkoPointKasebars.html


 Pages:  1  ..  18    20    22  ..  25