| Pages:
1
..
15
16
17 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@karish
there are no gui performance issues in GSB due to coding. Likely you are low on ram or free CPU.
also make a very big windows swap file. say 50 gb as min and 100 gb as the max
also reboot your pc
More next week...
Thanks received (1):
+1 Karish at 2024-03-01 17:13:59
|
|
|
Karish
Junior Member
Posts: 10
Registered: 26-9-2020
Member Is Offline
|
|
Quote: Originally posted by admin  | @karish
there are no gui performance issues in GSB due to coding. Likely you are low on ram or free CPU.
also make a very big windows swap file. say 50 gb as min and 100 gb as the max
also reboot your pc
More next week... |
Tried that too, didnt help
i got 64gb of ram,
i even turned off all my workers and developed 8,000 systems via cloud workers,
when:
Selecting all rows.
Sorting for families, using the +/- too
the GUI hangs.
the hanging GUI gets worst when i got more than 50K rows..
|
|
|
TradingPrice
Junior Member
Posts: 21
Registered: 4-8-2022
Member Is Offline
|
|
@Karish
not sure what you're trying to do when selecting that many systems, but if let's say you're trying to get stats or save indicator perf., I would do
that via macro
hope that helps
Thanks received (1):
+1 Karish at 2024-03-02 17:46:09
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@karish
a few things.
My opinion is you should NEVER build 50,000 systems in one go. Others may disagree.
It takes a long time, and if you take say the top 300 systems of 50,000- what i suspect will happen is the top 300 are a very narrow range to each
other (too similar)
also manipulating 50,000 systems is going to take a massive amount of time. Hence I would never do this. If you have under app settings set all your
cores to do this,
likely there is not windows resources to make the gui responsive.
There has so so much thought put into the methodology and its also done to be very fact in human and cpu time. So manipulating 300 systems is fast and
efficient.
An observation. People who are new to GSB, and vary from the well trodden path, you end up being less productive and can easily cause yourself lots of
problems.
Its better to vary from the default methodology after you have followed, and understood it. GSB is a very complex product.
Tradingprices comment is good, but note how the macro is doing the top 300 systems put in favourites A
Thanks received (1):
+1 Karish at 2024-03-03 16:11:52
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@KARISH
The number that shows next to the Unique-Systems (123) and Walk-Forward (123) tabs is not getting updated when deleting a strategy from the tab.
I TESTED THIS, WORKED FINE FOR ME. PERHAPS TRY NOT ALLOCATING ALL CPU CORES TO GSB TASKS UNDER APP SETTINGS
Would be cool to have the "Entry Modes", "Weights Mode", "Entry Level Mode", "EntryLevelValue" etc as an optional list of Use/Do Not use with
min/max/steps just like the "Built-In Indicators form we got already, Let the software decide what's good based on genetics etc, rather than have them
fixed.
you the user can fix them at zero, one , 0 to 2 step 0.5. telling gsb to use ga to find whats best is like giving a 2 year old a choice between
lollies or green peas.
only via massive sample of testing in automation, can you answer whats best, and you need to understand what your doing.
ie aic (any indicator cross) weight we use zero as the cross point. the actual range of the indicator normalized is -100 to 100 so it would be be
valid to use -50 to +50 step 5 BUT MY DEFAULTS ARE ZERO.
if your multiplying indicators (entry type compare2) then 0 to 2 step 0.5 would be better.
If you dont understand this (as its advanced) use defaults
"# of Filters" when set to 1 for example, it will be forced up on all systems.., would be cool if it wasn't, some systems will have it and some
wont, again.. more freedom, let the genetics spit its results, would it be 0,1,2 etc..
CAPS TO MAKE IT EASIER TO SEE MY COMMENTS. YOU CAN CHOOSE BOTH ALL TRADES AND SOME OTHER FILTER. GA WILL THEN CHOOSE WHAT IT THINKS IS BEST.
I THINK
UNDER STATS, YOU CAN SEE WHAT BEST. BUT THIS IS NOT NECESSARILY A GOOD IDEA (SEE 2 YEAR OLD COMMENT ABOVE) BEST TO LET GSB AUTOMATION TO SORT OUT
WHATS BEST HERE.
"Stops" same thing, it is currently forced on all systems..
THERE IS VARIABLE STOP. WHEN ITS BIG THIS IN EFFECT IS NO STOP
I AM NOT A GREAT FAN OF PUTTING A LOT OF GA WORK INTO THE STOP AND PT. MAKE THE STOP FAIRLY BIG SO IT RARELY HITS, AND NO PT FOR DAY TRADING SYSTEMS -
BUT A SAY 2500 PT (TOO SMALL FOR LIVE TRADING, WOULD ALSO HAVE MERITS)
WHATS GOOD ABOUT 2500 PT IN 23 HOUR NQ SYSTEM IS YOU GET A LOT MORE TRADES THAN BIG PT. MORE TRADES = LARGER SAMPLE = MORE ROBUST
FOR 23 HOUR A DAY SYSTEMS THE STOP IS FAIRLY TIGHT AND PT TIGHT (2500)
THIS IS DELIBERATE. IF ONE TINY VARIABLE TWEAK CHANGES A TRADE FROM 2500 TO 5000 PT, AND 5000 IS A BIG PART OF THE RETURNED EQUITY, YOU HAVE TRAINED
GA TO CURVE FIT LOOKING FOR BIG TRADES AND YOU WILL HAVE SYSTEMS THAT ONLY MADE MONEY IN EXTREMELY VOLATILE YEARS. THIS WAS MY MISTAKE IN LATE 2008
MADE A KILLING IN GLOBAL FINANCIAL CRISIS 2007 MID 2008 AND THEN MADE VERY LITTLE THE NEXT YEAR AS SYSTEMS WERE OPTIMIZED TO MAKE NEARLY ALL THE
PROFIT IN TIME OF HIGH VOLATILITY. AFTER EXTREME VOLATILITY CAME VERY LITTLE VOLATILITY
"Targets" also the same, it is currently forced on all systems..
SAME AS ABOVE
When using "Nth" it seems that the rules we set in side "Performance Filters" wont hold any merit, as if the NoTrd period of the "Nth" wont take any
of its results into account.
THIS IS VERY IMPORTANT. THE RULES APPLY TO THE INSAMPLE OF WHATS SEEN IN NTH, AND SO THE METRICS OOS WILL BE MUCH WORSE THAN IN SAMPLE RULES.
IF THIS WASNT THE CASE, GSB WOULD DECEIVE YOU INTO THINKING WHAT YOUR DOING IS BRILLIANT, WHEN ITS NOT.
SEE HOW BAD THE OOS IS CRITICAL TO DO.
Thanks received (1):
+1 Karish at 2024-03-04 01:12:29
|
|
|
| Pages:
1
..
15
16
17 |
|
![[*]](images/davis/default_icon.gif){kind=icon}
