GSB Forums - Powered by XMB 1.9.11

Questions from Beginners

saycem - 21-7-2018 at 01:35 AM

Hi - Wanted to create a thread for some of the basic questions of new users...

I've just started playing around with the software and built around 2000 systems ES.30-SPX.30 (30%) training Nth day (2) NoTrd.

My dates are 2000/1/1 - 2017/12/31 however I can only see equity curves from 2001/6/19 - 2017/02/09 so not sure why that is?

What is the best order to do things in? WFO first and then switch Nth day to Trd? or switch Nth day first? If I do WFO first when I switch Nth day to Trd I get a different result - the full equity curve that's much flatter.

If I switch Nth day to Trd before WFO I see just the Nth day equity curve. Which looks acceptable.

I chose a system with a reasonable WFO curve (although stability was only 56) I switched Nth to All so that the code would not skip Nth days in TS. Is this correct? or does it not matter? I copied the regular code not the WFO one - is this correct? I am using custom sess in TS 830-1500.

One thing that is very noticeable is the recent DD of the system once I include 2018 data in TS. I tried several strategies and this occured for every strategy. Does this just indicate that I have an over fitted strategy and there is something wrong in my process or is something else going on?

saycem - 21-7-2018 at 02:07 AM

Could I please confirm my settings are correct for exporting Daily ES bars from TS. (see attached)

Do I need to change my time zone to local and change my clock on my computer to Central US time? or can I just leave it on exchange. (I have created the custom sess time)

filename would be: ES.390.Minute.830_TO_1500_centralUSAtime_20180721

admin - 23-7-2018 at 01:04 AM

Upload a ts code, and I will see why its not trading earlier. Check es and spx have data from 1/1/2000
Your asking some good and basic questions. I dont see a big need for nth on es with 29,30,31 min bars, or even just es30.
GSB has changed to much, it means the answer to your basic and foundational questions are changing.
I would build on 29,30,31 with no nth. (or maybe nth 1 or 2 - no trade. Pick systems with better metrics
You could verify on all or some bars 25 to 35 excluding the 29,30,31)
(Done by choose verification data, then right click and veridy system(s)
You can optional verify on emd er emd nq. (dont do dow for now) Dont place to much weight on these markets till we
have changed close-closeD filter to be $ based, not point based.

Then verify on es 30 only
then wf
then choose systems with good metrics and stability scores.
Right now we cant WF 29,30,31 min bars, but that should be soon.

Not sure on your 2018. Ive seen this before but if its common in most systems, something is not ideal in your settings.
You can do a support upload so i can see the settings.
The tanking in 2018 means youve made system(s) on the wrong side of volatile markets. Ive seen this before, but for me its less common.
You es 390 looks ok, but you should be on local time. Reason is if you add any data2 on different time zone, TS wont work.
From memory though ES is on central so you should be ok.

saycem - 23-7-2018 at 02:13 AM

Thx for your response Peter.

Trying to understand what your saying here...
"(Dont by choose verification data, then right click and veridy system(s)" Did you mean "do this by choose verfication data, then right click..." I don't know of any other way.

If I understand you correctly:
Build on multiple timeframes of the same symbol,
then verify on other timeframes that you didn't build on.
Also verify on other markets same timeframe that you intend to trade on ie 30 min EMD NQ etc - just look for positive curve doesnt need to be perfect
Then change verfication data back to (30 min only) before you do wfo? (because we can't wfo on mulitiple timeframes yet)

I'm struggling to understand how changing the chart timeframe by only 1 or 2 min can help add robustness. To really test for robustness should we not use like 15 min 20 min 30 min 40 min 60 min? Also I thought the idea of using Nth for OOS was a good one - has this proved to not add anything?

re getting trades for 2018 - I changed TS code: dateYmd <= 20180630 to 29990630, which worked but would you have to do this everytime?

admin - 23-7-2018 at 02:31 AM

Quote: Originally posted by saycem

Thx for your response Peter.

Trying to understand what your saying here...
"(Dont by choose verification data, then right click and veridy system(s)" Did you mean "do this by choose verfication data, then right click..." I don't know of any other way.

If I understand you correctly:
Build on multiple timeframes of the same symbol,
then verify on other timeframes that you didn't build on.
Also verify on other markets same timeframe that you intend to trade on ie 30 min EMD NQ etc - just look for positive curve doesnt need to be perfect
Then change verfication data back to (30 min only) before you do wfo? (because we can't wfo on mulitiple timeframes yet)

I'm struggling to understand how changing the chart timeframe by only 1 or 2 min can help add robustness. To really test for robustness should we not use like 15 min 20 min 30 min 40 min 60 min? Also I thought the idea of using Nth for OOS was a good one - has this proved to not add anything?

re getting trades for 2018 - I changed TS code: dateYmd <= 20180630 to 29990630, which worked but would you have to do this everytime?

Sorry for bad typo, I fixed the post. Done by, not dont
29 30 31 is a big stress test.
reason is it changes the bars from 12,13,14 per day when 830 to 1500 used.
Bottom line is the market validation / nth clearly shows this
es30 with spx gave market degradation -20.6
es 29,30,31 around -9 %
I feel 15 20 30 40 60 is too extreme. But what I feel doesnt matter. The nth / market validation can prove me right or wrong. I havnt done that combination yet.
Why Im less fussed on not using nth of out of sample is ES is a very easy market to make systems for, and if it works on 29,30,31 and does fine in walk forward, there is less need for nth. Using nth reduces the sample size you have to play with, but the sample size is big enough regardless.
You also have extra validation with the other markets and time frames.
I suspect you have a typo
"Also verify on other markets same timeframe that you intend to trade on ie 30"
should be "Also verify on other markets same timeframe that you NOT intend to trade on ie 30"

saycem - 23-7-2018 at 04:16 AM

ok thx
I will re-watch degredation video again.
re typo - yes that is what I meant