| Pages:
1
..
80
81
82
83
84
..
98 |
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Quote: Originally posted by ChuckNZ  | For the record, I followed the directions in excruciating detail in the "recent nq methodology" document. I found the directions and the concept to
be very worthwhile. The fact that I like to see how random data may or may not affect the action shouldn't really offend anyone.
Chuck, in case you are reading the replies to your messages.
Fwiw,a good tip for you, would be to use mote than one type of validation, I use Random data, Other timeframes but most importantly other 3-4 similar
markets, to give me hints of what i have is valid or not..
I dont touch systems with anything below 70 in stability in wf and stay away from systems that shows negative numbers in wf oos.. and demand all
systems in family to have high wf scores...
This is of course my own personal opinion, and I am sure other does things differently.
| Quote: Originally posted by admin | @chuck
this means the wf stability of parameter is 60% for rolling, and 52% for coarse (reasonable)
the faster a system gets stable wf parameters, the better it is. If its a curve fit more likely to jump all over the place and have a low score
I have zero faith in random noise as validation.
You are asking questions which imply your doing your own learning (great if your an expert in trading and GSB), and not going down the path where
years of collective research is clearly laid out in the recent nq methodology. It means you waste lots of your own time with poor outcomes.
the bolleneger bands means the equity curves are close (good)
In all fairness, awesome your able to try new things at a mature age |
|
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
@Peter, when do you think we can start working on CL again? will each GSB indicator be tested by your dev team, so it matches GSb vs TS MC again?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
My understanding is the most recent build has fix for the 80 or so indicators in the trial mode of GSB. All the best indicators are put into the trial
mode.
I will push out 65.21 by monday.
THe big killer for CL was that GSB wrongly assumed the 0 was the lowest price possible. That would cause missmatch on all CL systems from the point of
cl being below zero.
There were other fixes like gsb_highd <> highd (ts function) This is a ts bug that was never spotted for years. We now do match checks on 50000 bars,
instead of 5000 and thats what picked that up
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
How to see parameter ranges being tested?
Somewhere over the last couple of weeks I came across (or saw in a video) where it was possible to see the parameter ranges being tested for
indicators. For instance, the range of parameters for a particular indicator might be 10 to 50 incrementing by 5. In particular, I would like to
see (and be able to change) the range of parameters and increments for the ATR Trailing Range.
If you can point me in the right direction, I would really appreciate it.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@Chuck,
its here. You might need to be in beta mode to access these features, which gives you even more to play with and go wrong.
Beta features are more buggy and may be removed / changed.
If you change these settings, you will loose compatibility with previously saved settings in gsb
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
Days with shortened hours?
Going through some trade listings of GSB-generated trades, I see trades entered in the morning on shortened trading days (like the day before
Thanksgiving and Black Friday) that aren't closed using the MOC statement. I can easily understand how/why this happens. One or two trades in ten
years or so isn't going to upset the stats, but can we overcome this minor inconvenience? I guess this is really a Tradestation problem?
Perhaps the answer is to simply be in the driver's seat on those days?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@Chuck, yes we are doing this in the ts end,
you are correct it wont affect gsb stats in any significant way, but for live trading this is important to get right.
Holidays is complex. I put a lot of work into it.
different markets and countries have different holidays.
metals have some days they close early, but I still get the full session out of the 'holiday'
and i have made some good money on holidays
However I never will trade a day on a specific market when the theoretical close is after the holiday close.
There also are a new holiday last year (forgot the name)
this topic was discussed a few weeks ago in the forum somewhere,
also ts code used was too cpu intensive for my liking.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I have updated the GSB diagnostic documentation features to show how to test any and every indicator.
There were quite a number of bugs, some which took years to show. Became apparent on CL when price was below zero for example
This is only for fairly technically competent users.
Many bugs were fixed in the last build, but in the beta tester indicators, more bugs are known to exist.
https://trademaid.info/gsbhelp/GSBDiagnostics.html
|
|
|
Piet
Junior Member
Posts: 18
Registered: 27-9-2020
Member Is Offline
|
|
| Quote: Originally posted by admin | I have updated the GSB diagnostic documentation features to show how to test any and every indicator.
There were quite a number of bugs, some which took years to show. Became apparent on CL when price was below zero for example
This is only for fairly technically competent users.
Many bugs were fixed in the last build, but in the beta tester indicators, more bugs are known to exist.
https://trademaid.info/gsbhelp/GSBDiagnostics.html |
Hi Peter,
what if we found something? I found 2 out of 5000 of the GSB_S3R31 results within the SM futures have significant differences. (in MC) Not worth it I
think, but could you provide some more guidance?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@PIET
We are putting a great deal of work into this at the moment. WIll know more next week.
Will keep you posted.
|
|
|
TradingPrice
Junior Member
Posts: 21
Registered: 4-8-2022
Member Is Offline
|
|
Hi Peter,
I'm missing GSB_CloseLessOpenDFilter
I've installed latest script GSB_Scripts_2022_12_23_2
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@Tradingprice
Opps 
its here
Attachment: Login to view the details
Thanks received (1):
+1 TradingPrice at 2023-01-16 04:54:44
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
Pearson's Metrics
I'd like to play a bit of Devil's Advocate regarding the use of the Pearson's Indicator. There seems to be a lot of emphasis on this site and the
documentation/videos on using Pearson's, looking for equity curves with a fairly straight line. The straighter the line, the higher the Pearson
number.
When looking at the equity curves for @NQ and @ES, the curves do have a fairly straight line from 2010 to 2019. Then, the equity curve almost goes
vertical for 2020 through the end of 2022. The result is that a lot of excellent strategies are being rejected due to having a Pearson's less than
(say) 94.
So, I'm questioning the wisdom of using Pearson's as a filter. I'm not rubbishing the idea. I'm just interested in other users' thoughts on the
subject.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Excellent comment. I have very deliberately excluded data from 2020.2.28 from the the build process.
Its really problematic to have this in the build process. The reason is most of the money to be made is in this period and you will have a system
thats curve fitted for the extremely volatile and profitable time
I made this mistake after the extremely profitable 2007 - 2008 period, and didnt make significant money the following year when volatility dropped.
A great deal of thought and effort has gone into the default nq /es settings, and so most details are extremely well thought out
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
@Peter, was just about to create a new macro for my new GC run, however i can see that i cant create stats for anything further than H, could you
kindly allow more stats letters to be used please, that would be very kind of you
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@daniel,
Im using the nq macro here.
Let me know your f stats. Reason I ask is I have one user who is getting really low f stats, but dont see any issues with his setup.
I have not had much time to do any system building for some, so never got to finished revised NG and GC
Attachment: Login to view the details
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
| Quote: Originally posted by admin | @daniel,
Im using the nq macro here.
Let me know your f stats. Reason I ask is I have one user who is getting really low f stats, but dont see any issues with his setup.
I have not had much time to do any system building for some, so never got to finished revised NG and GC |
Hi Peter,
Not sure what dates you use for your F stats in your macro that seems to be Nct derived form the name of it. But in my new GC Macro my F stats period
for GC is 2019 03 - 2022 02..
What period does F represent for the user you had in mind ? I can see what that period delivers if i have a matching period in my macro, since i am
running GC now... just me let know and i can check
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I think f is the entire out of sample period.
From memory I deliberately avoid any stats after 2.28.2020 as the profit in the years after this is so large, you will end up finding systems that
only work in extreme volatility
I made this mistake during global financial colapse in 2007 to 2008 and didnt make money the next year. (no big deal in light of the massive profits
of 2007-2008)
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
@Peter, got it, but I thought you wanted my stats since you wanted to compare for the user you mentioned, so in order to find a period in my GC macro
comparable to the users F stat period, i just need to know the dates. I dont think my whole OOS period, represent same dates as this user.
And also does "OOS" includes "trd" period (or no trd depending on your setup) in the build.
I do agree with you in regards to high vol environment and the challenges it provides in dev.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@daniel
try this file
date file is in central usa time, not exchange
Attachment: Login to view the details
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
@Peter i cant run this file, to many differences to make it possible.. but i can share some stats for GC, i am not finished with my GC run, but should
look like this at least or much better in the end...
I see in the shared opt setting, that build date ends 2021 02, (must be wrong?) I can suggest also to test much earlier session start time than in opt
setting, and also other Nth than 80, since that is not what i found is "best" and stop less than the opted for 2k.
Just my 2cents
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I havnt looked at the file for some time. I though end should have been 2020.2.28. I need to revisit this but are too busy to do system building for
some time.
Point taken on the stop. I would optimize that last in your system in ts/mc just to get an idea of the range that works best.
In times of extreme volaitlity , tight stop works badly - but im havent even tested if GC is much more volatile in the last 2 years.
I thought 3am or so was best session time. Not even sure whats in the settings as gsb_automation would be over riding whats in them anyway.
Your comments appreciated
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by ChuckNZ | I'd like to play a bit of Devil's Advocate regarding the use of the Pearson's Indicator. There seems to be a lot of emphasis on this site and the
documentation/videos on using Pearson's, looking for equity curves with a fairly straight line. The straighter the line, the higher the Pearson
number.
When looking at the equity curves for @NQ and @ES, the curves do have a fairly straight line from 2010 to 2019. Then, the equity curve almost goes
vertical for 2020 through the end of 2022. The result is that a lot of excellent strategies are being rejected due to having a Pearson's less than
(say) 94.
So, I'm questioning the wisdom of using Pearson's as a filter. I'm not rubbishing the idea. I'm just interested in other users' thoughts on the
subject. |
Hi Chuck,
Yes, I agree. I stopped using Pearson as my fitness function. I mainly sue Peters NPxAT and NP/DD.
Recently I was re-reading Pardo book and came across his PP (= perfect profit) approach. And comparing the results of a trading strategy to the PP
number. This is the "model efficiency".
I exported ES 390 minute 0830-1500 data from 1997-now.
And choose as my "perfect profit": the absolute value between open and close.
Then did a summation starting in 1997 up to today. Result with one ES contract (without costs and slippage) 3.8 million USD.
Next I took one of my better systems, aggregated the trade results per day and put both equity lines in one chart.
Pardo states: getting 5% or more of PP is a very good strategy. Especially when the "model efficiency" is stable across time.
Maybe this "perfect profit" approach is a good addition to GSBs fitness function?
Edit Feb 2: I meant adding the correlation between a equity line of a strategy and the "perfect profit" equity line as a fitness function.
What are your thoughts about this, Peter?
|
|
|
timemeantnothing
Junior Member
Posts: 22
Registered: 7-7-2022
Member Is Offline
|
|
Hi Peter, is it possible to make an ES or NQ system that does well during low volatility periods? For example maybe the system automatically doubles
the contract size when the volatility is low or another approach.
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
Max.Dup.Indics. ???
When I run GSB, I get thousands of systems that are using more than one instance of the same indicator. I have even seen instances where a strategy
is using three copies of the BollingerBand indicator with different parameters. I have also seen a situation where BollingerBand indicator is used
twice with the same settings.
I tried setting Max.Dup.Indics. to zero, one or even two. Then I end up with NO systems being generated. I think it might be counting occurences of
GSB_Norm4.
I would really like to generate systems without duplicate indicators if possible. I may find that performance degrades, but at least I will know.
Thanks for any assistance in this area.
|
|
|
| Pages:
1
..
80
81
82
83
84
..
98 |
![[*]](images/davis/default_icon.gif){kind=icon}
