| Pages:
1
..
7
8
9
10
11
..
19 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Update,
there is a bug in .pa files which shows some large losses in the system report that did not occur in TS. Working on a fix now.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Goldminer report now correct
Thanks received (1):
+1 cotila1 at 2023-04-13 01:50:38
|
|
|
SwedenTrader
Junior Member
Posts: 63
Registered: 3-8-2020
Location: Stockholm, Sweden
Member Is Offline
|
|
The PA-file is not here yet ? Is it uploaded somewhere so we can test it in PA ?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@sweden trader
all pa files are here.
Only gc are up to date
https://u.pcloud.link/publink/show?code=kZXfhjVZFVuT40tgMfBC...
Thanks received (1):
+1 SwedenTrader at 2023-04-13 02:28:12
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
important update
gsb_GC9 on my machine and 2 others has no trades, even if data goes back to 2007
I have no idea why, and will look into this tommrow
possible other systems / markets effected
never seen anything like this
gsbgc4 seems the same, not gsbgc1 and gc6
comments from others welcome
|
|
|
SwedenTrader
Junior Member
Posts: 63
Registered: 3-8-2020
Location: Stockholm, Sweden
Member Is Offline
|
|
GC_sys1, GC_sys3, GC_sys5, GC_sys9 looks fine on TS10. I lost GC_sys4 A and B they are somewhere in the backups.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Thanks for the comments.
gc9 and 4 is not fine on my ts 10. In 22 years have not seen an issue like this
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I posted on ts forum
https://community.tradestation.com/Discussions/Topic.aspx?To...
as of today, this code is busted
vars:v3(0);
v3=AccumSwingIndex;
print (date,time,v3)
results always = 0
ts 9.5 and 10 affected, numerous users but not all
symbol @gc30, @nq fails
|
|
|
portfolioquanttrader2020
Member
Posts: 269
Registered: 30-4-2020
Member Is Offline
|
|
I have tried this symbol and it works
@GC=102NC+FGHJKMNQUXZ
gc9 230-1230 exchange
Thanks received (1):
+1 admin at 2023-04-16 20:57:31
|
|
|
jeffreed44
Junior Member
Posts: 2
Registered: 27-2-2023
Member Is Offline
|
|
GSBsysFLT5.NQ
Quote: Originally posted by admin  | @cheesman,
well spotted. I have not released all systems I trade, and some of that is not at all deliberate, so Im happy to sell it.
Currently it doesnt have all the extra code like auto detect emini / micros, time zones etc, so I can release it later next week when I do all of
that.
|
Hi Peter, are you still planning to release this system for sale? Would very much be interested in this updated system once the auto-detect for
micros, etc is updated. Thanks!
|
|
|
jeffreed44
Junior Member
Posts: 2
Registered: 27-2-2023
Member Is Offline
|
|
GSBsysFLT5.NQ
| Quote: Originally posted by admin | @cheesman,
well spotted. I have not released all systems I trade, and some of that is not at all deliberate, so Im happy to sell it.
Currently it doesnt have all the extra code like auto detect emini / micros, time zones etc, so I can release it later next week when I do all of
that.
|
Hi Peter, are you still planning to release this system for sale? Would very much be interested in this updated system once the auto-detect for
micros, etc is updated. Thanks!
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Jeffreed44
Yes I am, just drowning in things to do, but its getting better.
Some of the time hungry research and tax tasks are nearly complete
apologies over lateness of getting this done.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Updates to GSBsys1.3 to 1.5 series
This is free update to all users who bought the systems.
Sadly these systems are in high drawdown, which i put down to market conditions.
Your existing pcloud urls that you got when you bought the system should work
also update to GSBSYSv2-15minES
I still trade all these systems but a greatly reduced leverage.
The eld names have not changed, but the contents of the eld files have change
All systems should have an input under notes which has a comment on the new build date.
  
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
here is GSBSYSv2-15minES
new version is at new highs, but regardless its risky conditions for day trading ES
I have not forgotten about new nq system (not new but just new for me to sell)
im reviewing systems, trying to improve them and not sure what workspace its on.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
correction, results on 1.3es were not correct. one filter not applied
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Happy to say, im on a bit of a roll!
I added the pattern filter from the ES systems to ZoneTraderNQ
ZTnq results were not to bad regardless, but significant improvement and new highs
Even better the upgrade is free for any users who bought ZT2 from me. (not all users bought it from me)
  
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Update to Zonetrader 2.04
Small bug fix today (fix that was in older version that accidently got ommited)
and tiny improvement.
System for sale now too
https://trademaid.info/gsbhelp/ZoneTraderNQ.html
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
Can you share the test and validation process that you are using when you add additional filters to existing strategies?
Thanks
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@CARL,
Good question.
The OpenD filter worked on all the ES systems I tried, and gave a nice bell curve as far as parameters go.
With the trend filter, I tried many things and one was clearly the best.
Again it had a nice bell curve.
Other filters might have made tiny improvements, but they might reduce number of trades by say only 1%, and no bell curve, which means they were not
valid.
What you want is say 10% or more trades reduced, but much better metrics, and ideally (often not the case) more profit and less drawdown, profit
factor and average trade.
Do this on any system that has secondary filters as ClosedBPV and you will see this very clearly
Is interesting that the trend filter seem to work on all es systems ive tried so far, and I think its due to ES tends to revert to the mean.
NQ which I think has poorer mean reversion it did not work on, except for zone trader NQ
Even more interesting is zonetrader counter trend likes the long filter on short entries and short filter on long entries, but for trend following
likes the long filter on long entries
(ZT2 has trend and counter trend module)
by the way, zt2 is equity looking good.$2860 win per today (live fill) per contract
Thanks received (2):
+1 Carl at 2023-05-06 02:54:58 +1 Ketil at 2023-05-06 01:53:45
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Thanks, Peter.
The shape of the bell curve you mentioned in your previous post is over the entire dataset?
So you don't use an in-sample segment to test a filter and validate/verify this on out-of-sample data?
|
|
|
portfolioquanttrader2020
Member
Posts: 269
Registered: 30-4-2020
Member Is Offline
|
|
Filters must be tested independently, not stacked because then we would be changing the baseline, or heart of the strategy.
There must be a period in sample and out of sample and an order in the development and implementation of filters. The bell curve is the test of
parametric sensitivity, but I understand that information is missing from the information that Peter has given.
|
|
|
portfolioquanttrader2020
Member
Posts: 269
Registered: 30-4-2020
Member Is Offline
|
|
First, the heart of the strategy is defined. Later, each new rule or filter is a brick and must be tested independently of the rest. For this reason,
the filter that Peter has put in I understand that it should be tested only on the heart of the strategy, deactivating the rest of the filters. If it
is stacked, we run the risk of over-optimization, having changed the heart of the strategy.
I would like a more extensive explanation by Peter
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Sofar I have been testing filters this way:
Test the filter on in-sample (50-80% of all the data).
Then look for a good bell curve shaped graph of the net profit and ROA.
I like the bell curve on the out-of-sample data to be similar to the bell curve of the in-sample data.
The preferred parameter value on the in-sample data must be almost the same like the preferred parameter value on the out-of-sample data.
So if I select a parameter value based on in-sample data and this differs a lot from the "best" parameter value on the out-of-sample data, I will
reject this specific filter.
With that said, I still don't have a Ferrari, so maybe there are much better methods then what I have been doing.
|
|
|
portfolioquanttrader2020
Member
Posts: 269
Registered: 30-4-2020
Member Is Offline
|
|
Hi Carl,
Yes, but the filters must be deactivated, to try one by one, as you are saying, and then choose them, and combine them. But sometimes, when you do not
do independent tests, it can happen that you optimize and see the bell of a filter that if you test it without deactivating the rest, the results are
excellent, but if you test it isolated from the rest, the results are bad. This independence test is used by Jeff Swanson, to avoid overfit.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
"Thanks, Peter.
The shape of the bell curve you mentioned in your previous post is over the entire dataset?
So you don't use an in-sample segment to test a filter and validate/verify this on out-of-sample data?
"
you can test this at any stage by doing a WF of the new filter, but there are many other ways.
For example the new filter worked on all the ES systems I tried. You can argue they are significantly correlated.
There is a very logical, explainable reason why this filter should work.
The filter also worked just as well on zonetrader2nq. This system is complex compared to GSB systems, has trend following
and counter trend (GSB are not normally counter trend, but there are exceptions) and the filter worked inverse on counter trend in zt2
There is no way im going to test a filter that I believe is solid on a complex system like zt2 by removing existing zt2 filters. Not saying its wrong
to do so, but im not going to do it
There is danger that with too few trades, invalid statistical results occur. zt2nq had about the 750 trades after the filter was applied. so thats a
decent sample, esp as gsbsys..es also worked well with it and both long and short trades improved too.
If we were to build systems with no filters, then add them later, we should be building them with no SF, and doing that last.
Your welcome to try that approach, and it could be done with gsb as is, as sf can be turned off.
However what we know to be true is the sf concept has worked well years out of sample.
There are other practical issues. Sometimes you get such good market conditions, valid filters for a time are best left out.
|
|
|
| Pages:
1
..
7
8
9
10
11
..
19 |
![[*]](images/davis/default_icon.gif){kind=icon}
