| Pages:
1
..
5
6
7
8
9 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo
there is also AI wf in gsb. Likely this will do what you ask for. Right click a system and do AI-wf
NT is lagging ts/mc but updates are done regularly on it, and more likely in about 2 weeks.
WF does improve settings, and time filters are in the pipeline.
"So, it is not possible to optimize the parameters of a strategy, neither does GSB?" yes, this is possible. Its one if the main points of doing WF.
wf chooses better more robust parameters, and gives an indication if the system is valid or likely to fall apart in future trading.
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
Quote: Originally posted by admin  | @remo
there is also AI wf in gsb. Likely this will do what you ask for. Right click a system and do AI-wf
NT is lagging ts/mc but updates are done regularly on it, and more likely in about 2 weeks.
WF does improve settings, and time filters are in the pipeline.
"So, it is not possible to optimize the parameters of a strategy, neither does GSB?" yes, this is possible. Its one if the main points of doing WF.
wf chooses better more robust parameters, and gives an indication if the system is valid or likely to fall apart in future trading.
|
Thanks for your help, but I don't need a W.F to tell me if the system is robust or not.
Not all of us think like you.
I don't understand what is done at AI W.F, what is it for? I do not get it.
What I need is to be able to choose the operating parameters, W.F I don't like it, I prefer particular tests out of range, IS-OSS
I get away with it, being able to select the parameters is something basic.
Is it possible to select the parameters? Maximum or minimum?
So how did you select the parameters to operate?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo
ai wf optimizes each paramater indiviually. Gives you a graph of np, ave trade etc.
You are doing something unique that no other users are doing. This also means your disconnected from the methadolgy and the GSB user base. Fine if you
know what your doing.. but the methodology works well and it means your a lone ranger, not a team player. None of that is wrong, but it has weakness's
- but strengths if you know things that work that the rest of us don't.
we select the final parameters by a wf. this chooses min and max by +-50% =100% (input in gsb) from the existing value of the parameter range in gsb
so if a variable-length is say 100, it will use 50 to 150 as a range to optimize.
Afte a WF, there is a drop down under scripts and you select wf code live trading
If your not doing a wf, then you can select parameters in ts/mc/nt under inputs.
In ts you change the word vars: to inputs:
Im not sure how this works in NT
The fact that not everyone thinks like me is the strength of GSB. The community often thinks differently on things, sometimes we argue and often we
have objective robust results that show who was right. The glorious thing about being proved wrong in an argument is... our profitability in trading
improves.
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by REMO755 |
Thanks for your help, but I don't need a W.F to tell me if the system is robust or not.
What I need is to be able to choose the operating parameters, W.F I don't like it, I prefer particular tests out of range, IS-OSS
I get away with it, being able to select the parameters is something basic.
|
Hi REMO755,
Are you willing to share the methods you like to use to build an select systems?
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | @remo
ai wf optimizes each paramater indiviually. Gives you a graph of np, ave trade etc.
You are doing something unique that no other users are doing. This also means your disconnected from the methadolgy and the GSB user base. Fine if you
know what your doing.. but the methodology works well and it means your a lone ranger, not a team player. None of that is wrong, but it has weakness's
- but strengths if you know things that work that the rest of us don't.
we select the final parameters by a wf. this chooses min and max by +-50% =100% (input in gsb) from the existing value of the parameter range in gsb
so if a variable-length is say 100, it will use 50 to 150 as a range to optimize.
Afte a WF, there is a drop down under scripts and you select wf code live trading
If your not doing a wf, then you can select parameters in ts/mc/nt under inputs.
In ts you change the word vars: to inputs:
Im not sure how this works in NT
The fact that not everyone thinks like me is the strength of GSB. The community often thinks differently on things, sometimes we argue and often we
have objective robust results that show who was right. The glorious thing about being proved wrong in an argument is... our profitability in trading
improves.
|
Hi Peter,
I do not question your method, I appreciate your effort in the development of GSB and its methodology, only your system ES, CL, NG, etc, is worth more
than what you paid for the license, thank you!
I am a team player, my criticism is always constructive, never destructive, if something has worked for me, every change is painful,
I would like to be able to optimize the parameters of a strategy with freedom, I don't ask for much Peter.
Editing these fields would be a substantial improvement to be able to test freely.
When setting 0, EWFO serves to optimize the entire desired period, is it possible to set 0?
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by Carl | | Quote: Originally posted by REMO755 |
Thanks for your help, but I don't need a W.F to tell me if the system is robust or not.
What I need is to be able to choose the operating parameters, W.F I don't like it, I prefer particular tests out of range, IS-OSS
I get away with it, being able to select the parameters is something basic.
|
Hi REMO755,
Are you willing to share the methods you like to use to build an select systems?
|
Hi carl
I am always willing to share knowledge, I am a human being and it is the fastest way to progress that we have.
I can share the portfolio management experience in an audited live account, managing a profit of 128% in a year with intraday systems, I never use w.f
to obtain the parameters to trade.
I always did out-of-sample testing and optimization (previous results are not a guarantee of current results, but they take the dance off me)
The best method is portfolio management, good portfolio management with mediocre systems can pay off in the long run.
Portfolio weighting by historical risk of each strategy, all strategies the same risks, rigorous DD management when strategies break, etc.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo755
What you shared is good, and excellent results too.
More detail in what you share would be good.
Paramater choice and optimization method, oos method used. (lots of this is done in gsb methodology.
Books are written on these topics, and some there are no easy answers.
For example, GSBSYS1ES had I think max dd, before going to to clear new highest again, as did zone trader nq.
https://trademaid.info/gsbhelp/GSBfreesystems.html see trade list about trade 492 and drawdown graphs
zt1es and zt2es (based on zt1es) did not recover as of now.
How do you define risk?
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | @remo755
What you shared is good, and excellent results too.
More detail in what you share would be good.
Paramater choice and optimization method, oos method used. (lots of this is done in gsb methodology.
Books are written on these topics, and some there are no easy answers.
For example, GSBSYS1ES had I think max dd, before going to to clear new highest again, as did zone trader nq.
https://trademaid.info/gsbhelp/GSBfreesystems.html see trade list about trade 492 and drawdown graphs
zt1es and zt2es (based on zt1es) did not recover as of now.
How do you define risk?
|
"Method of choice and optimization of parameters, method used. (Much of this is done in gsb methodology"
Your GSB program already gives me the option to confirm that a strategy is no longer over-optimized by testing% training,% testing,% validation, why
do I need to do WF? Through optimization I will look for the best parameters that are now working and I will go to the market with them convinced that
I am going with the best.
How do you define risk? Very good question "eureka"
The risk is defined by the union of the OOS periods in a W.F.
W.F helps me to establish the risk of each series within the portfolio.
Can I get the best parameters with w.f? Yes I can.
Is it easier for me to get better parameters with Optimization? If it is easier.
Would being free to choose parameter ranges and optimize give me more control in operations? If I will have more control and will be able to better
analyze some parameters.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@REMO755
" confirm that a strategy is no longer over-optimized by testing% training,% testing,% validation"
doing this can still be a total curve fit. its not much different to doing 100% training as when the human looks at the total of all results, you have
in effect done the same thing
as your only choosing systems that have good training,test,validation
a rolling wf is very good stress test, as it will only use best parmaters from a short period for the next period. If its not robust and valid system
it falls apart.
Anchored does the same, its less of a stress test but better for final parmaters as you have the best parameters that work over all trading
conditions. Less likely to be a curve fit.
Also if your parameters do not show stability, its a red/ornage flag. Ideally parameters have a bell curve result. If its random with no pattern, its
a red flag
Thanks received (2):
+1 REMO755 at 2021-07-16 16:30:37 +1 SwedenTrader at 2021-06-16 06:03:42
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
20210621.1 released via auto update when you open EWFO. This will be the last update for some time.
This build has a number of improvments, and very important bug fixes. All users should upgrade
If users did not install under c:\ewfo and they are not running as windows administrator, then the update won't occur
successfully unless you re-run ewfo as windows adminstrator.
New fitness types are also included.
Thanks received (4):
+1 REMO755 at 2021-07-16 19:09:40 +1 Daniel UK1 at 2021-06-30 02:16:23 +1 Carl at 2021-06-29 02:58:18 +1 Bruce at 2021-06-29 02:48:09
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | Update to EWFO
Note this WF
Its not valid as the final parameter hit the extreme range used.
value x should have used a value > 0.2
There are now vertical lines on the inputs, as it was too hard to line the parameters up
and current line is now directly below 10
Parameters now in red / orange on both tabs, and parameter range used on both tabs.
Anyone wants to test the update before I release it, just let me know
|
Hello,
How do you increase the parameter range if this happens?
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | Update to EWFO
Note this WF
Its not valid as the final parameter hit the extreme range used.
value x should have used a value > 0.2
|
Hello,
How do you increase the parameter range if this happens?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@REMO755, if its in ts the file are generated, then simply increase the range.
If in GSB double your nearest value.
In this case also increase ga gen / population to 150x150
Less needed if entry type aic is done the increase if gen & population is less needed as there are no weights or entry type to optimze
Thanks received (1):
+1 REMO755 at 2021-07-21 02:33:06
|
|
|
asobi
Junior Member
Posts: 12
Registered: 23-10-2019
Member Is Offline
|
|
| Quote: Originally posted by admin | @REMO755, if its in ts the file are generated, then simply increase the range.
If in GSB double your nearest value.
In this case also increase ga gen / population to 150x150
Less needed if entry type aic is done the increase if gen & population is less needed as there are no weights or entry type to optimze
|
hi Peter, can you please give more details on this topic?
What's your rationales to tweak parameters like weights and ga gens/populations for different entry modes? How would you interpret the role of weights
when doing GSB runs and picking entry modes?
Many thanks!
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@asobi
aic (any indicator cross) has much less parameters.
No weights, entry level. Therefore you can wf with less combinations, and likely you will get greater parameter stability.
same applies for NCC (no conflict cross - though I dont use this entry type)
single and dual cross have even more parameters so may need more iterations, but I like to keep things simpler
and rarely use these. (use cross instead)
if you have doubled the nearest value, it means there is now a great deal more parameters, so we need more combinations (wf ga gen/pop)
if wf gen / pop is too low then stability figures lower (bad), identical wf will vary each time its done to a greater degree (bad) (due to random seed
and so many possible combinations)
If gen / pop is too high, its slower but no other problems
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
Hello,
Wouldn't it be possible to do something easier here?
Wouldn't it be a good idea to be able to copy the scripts in EWFO, as is done in GSB?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo, Its easy, see this field here at the bottom
your 30 parmater needs to be tested lower as the final parameter is 30, and thats the lowest in your range
same with the 3rd parameter of 1.25
Thanks received (1):
+1 REMO755 at 2021-07-22 17:26:41
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
Hello,
I want to w.f to ewfo GSB ultra fast, 5 generations.
If I do this, am I conditioning all tests in EWFO to 5 generations?
Can I handle generations in EWFO?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by REMO755 | Hello,
I want to w.f to ewfo GSB ultra fast, 5 generations.
If I do this, am I conditioning all tests in EWFO to 5 generations?
Can I handle generations in EWFO? |
5 generations will be useless as there is not enough combinations for it to be valid.
If we used brute force, my wild guess is there is 100,000 or so combinations
(could be 1 million, depend on entry type, parameter range etc
EWFO can only optimize the data that its given.
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by REMO755 | Hello,
I want to w.f to ewfo GSB ultra fast, 5 generations.
If I do this, am I conditioning all tests in EWFO to 5 generations?
Can I handle generations in EWFO? |
5 generations will be useless as there is not enough combinations for it to be valid.
If we used brute force, my wild guess is there is 100,000 or so combinations
(could be 1 million, depend on entry type, parameter range etc
EWFO can only optimize the data that its given. |
I wanted to use 5 generations to transfer the data to EWFO quickly and then test in EWFO.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo. If you do that then the wf is invalid. Not enough combinations. you need at least 100 x 100 gen / pop
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
Hello Pedro
I like EWFO, very powerful and fast.
Once the data and files have been downloaded to EWFO, why is it not possible to configure the genetic options there? Generations, population, etc.
If I use GSB Generations 80, does this affect EWFO? Will I always work at EWFO with 80 Generations?
Why this dependency?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@Remo
EWFO can only work with the data given to it, it cannot make data. All these things are outside its control as it doesnt have the code for trading
systems required to make it.
i use generations / population of 100 / 100 to 150/150 depeding on how many combinations there are.
ie aic entry has less combinations, while cross2 has the most.
If search space area is increased from 100 to 200, there are many more combinations, so this also makes me go towards 150/150
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | @Remo
EWFO can only work with the data given to it, it cannot make data. All these things are outside its control as it doesnt have the code for trading
systems required to make it.
i use generations / population of 100 / 100 to 150/150 depeding on how many combinations there are.
ie aic entry has less combinations, while cross2 has the most.
If search space area is increased from 100 to 200, there are many more combinations, so this also makes me go towards 150/150 |
Hello Pedro,
Why don't you integrate EWFO into GSB?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@remo, a great deal of work to do with nearly zero upside.
It would also means ewfo and gsbewfo need to be maintained. More work again.
I only do it some times, but I still like to go from a wf into gsb -> ewfo then test each input individually in ts then ts to ewfo
Programmer time is very limited, and has to be used to maximize the entire customer base as much as possible.
|
|
|
| Pages:
1
..
5
6
7
8
9 |
![[*]](images/davis/default_icon.gif){kind=icon}
