GSB Forums - Powered by XMB 1.9.11

GSB & the Dax, + day of week filters

admin - 28-4-2021 at 11:45 PM

This thread has two purposes.
1) To show day of week filters AFTER a system has been built.
2) to discuss how to make trading systems on the Dax with GSB.
I have put no time into this, and wont for some time. I do however have some ideas of where to start.

There often is a reasonable strong and consistent bias per day of week, that is separate long and short.

In most but not all cases, week days are still profitable, but less so than other days. The trend is even stronger if slippage and commission is added.
I do not like to build systems with DOW filters. My logic behind this is I want as many trades as possible to avoid curve fit.

Shown here is a DAX system with zero slippage and commission.
I think the Dax market is not well explored with GSB yet, especially with the closedbpv filter combined with secondary filter Closetohighlow3 filter. (NOT normalized)
The Dax is possible the market I have made the most profit on, with RC2dax. The first system that the pre GSB system builder made. RC2dax has a significantly different architecture to GSB systems and it had a number of human designed filters to reduce the amount of trades.
The following system uses the GSB_CloseLessPrevCloseDBpv2 secondary filter which used to be the default secondary filter. Now many markets work better with Closetohighlow3
The Dax system has data 900 to 1500 , 30 minute bars, with @es.d 900 to 1500 30 minutes as data2

If you are trying the methodology on the Dax, you should also try 830 and 930 start times. See the gold videos for how this was done.
I did not write this Dax system, and don't think its quite good enough to trade live. The example is shown for the teaching of concepts only

admin - 28-4-2021 at 11:48 PM

Day of week code added for longs. Note we allow a short trade to reverse on the long day of week filter. (if current contracts=1)

Skipping Tue (2) and Thur (4) gives mild reduction in Net profit, but better metrics

admin - 28-4-2021 at 11:53 PM

Day of week code added for shorts.
Note the long code is set back to no DOW for now

admin - 28-4-2021 at 11:55 PM

final system

edgetrader - 6-5-2021 at 11:08 AM

Did you compare filtering results of the example Dax system with RC2dax? With that I mean, are the same weekdays long/short the weakest in both systems?

If it's the same it could be trusted, if it bounces around I'd be skeptical.

Also as opposed to skipping a day, in testing I'd look at results just for that day.

admin - 6-5-2021 at 04:37 PM

Quote: Originally posted by edgetrader

Did you compare filtering results of the example Dax system with RC2dax? With that I mean, are the same weekdays long/short the weakest in both systems?

If it's the same it could be trusted, if it bounces around I'd be skeptical.

Also as opposed to skipping a day, in testing I'd look at results just for that day.

No I didn't, but that's a good idea. I will try to get that done but busy on other things now.
I agree that I don't want to skip a day testing while building systems. A day might be weak but still gives us valid sample of trades.
day of week has been used by very competent system developers, including Andrea Unger for decades. I think its safe to use, but do at the end of the build process.
If I manually re-optimize a strategy, I do it with all days of the week. DOW is the very last thing I would do.

admin - 6-5-2021 at 06:58 PM

Ove done the work on rc2dax.
wahts good about rc2dax is it has a out of sample since 2017
So I did dow with all data form 2003 till today, and applied dow on pre 2017, and looked at the OOS
of post 2017 with and without DOW filters.

results between the two methods varied significantly, but DOW was worth doing.

Attachment: Login to view the details

SwedenTrader - 9-11-2022 at 11:31 AM

Have anyone worked with Dax or minidax since the last post here ? Just the fact that the market opens on a different session is an reason to belive the correlation on US index intraday systems should be low even if the markets themselves are correlated.

Daniel UK1 - 9-11-2022 at 11:50 AM

I spent a bit of time earlier this year, very short though like 3-4 weeks, i thought i had something good, but in the end lots of systems failed, and this has not happened before like this... So dax to me, seems like a tough animal to tame.. I stay away for time being... perhaps others have had better success ?

admin - 9-11-2022 at 04:08 PM

@SwedenTrader
you can use @es.d as data2.
Beware, from memory when we get to daylight savings changes, the ts has a bug where it is looking forwards one bar and this gives spectacular false results
it happens end of march, and i think right now
the possible times to use are (830 900 930) to 1500 (central usa time)
and with only dax the work hours of germany. 830> to 5pm (where the volume is)
some users have had success on dax, but i havnt seriously tried

Piet - 11-11-2022 at 01:04 AM

Tried a lot with dax and data2 but came up with quite simple one. Dax is one of the hardest market. All with data2 from a different time zone failed at the end. Could be the daylight saving bug, never get into it.
My life system isn't one of my best, has quite few trades but is very profitible on the paper. Session is 230-1130 (CT), MOC on losing swing trade so it goes for long shots.
Tried some DOW tweaks but didn't use it. Felt like overfitting to me.
Some years with small losses alternate with big gains.