GSB Forums - Powered by XMB 1.9.11

Likely Monday, I will do a small GSB article called Meatpies, if you worked in a meat pie factory would you eat one.
Reminds me of a story.
I went into a meat pie factory to do some server repairs 20 + years ago. I weighed about 63 kg at the time, 5'11 feet high.
security guard is supposed to stop all exiting vechicals to search for stolen food.
He said to me, your too thin, no need for a search. ....

So...what exactly are you trying to say?

admin - 21-12-2020 at 06:00 PM

this is continuation of the article on previous page "

posted on 16-12-2020 at 11:38 PM"
Earlier I told my meat pie story.
If you work in a meat pie factory, will you eat a meat pie? There has to be a certain % of meat
but meat could be snouts, ears, tongue roots, tendons and blood vessels.
So the moral is avoid cheap meet pies.
How does that relate to GSB
A lot of work has gone into GSB so it doesn't produce systems that are flawed.
IE
IF LOW <= HIGH & CONDITION2 THEN BUY
Well low will always be less than high, so you have redundant logic.
Now entry types
each market normally has a strong bias to certain entry type
GC & ES to cross, CL to compare 2
It is critical to get the correct entry type for each market, and do not run multiple entry types at once. One commend is the 3 cross entry types are so similar, there is little difference between them
Some users use any indicator cross or no conflict cross
I tried it of later with success on gold.

v1=inidcator1
v2=inidcator1
v3=inidcator1
So if any of these indicators cross say 0 it trades.
This in practice allows 2 indicators that are totally crap to be in the system

But when I dig down into the system..
Here was results of a system using AIC

v1 = inidcator1; //13 loosing trades only.
v2 = inidcator2; //9k8 pf 2.76 by itself. It was about 20 trades so statistically non significant
v3 = indicator3;//60k pf 2.39 by itself. Nearly all the profit made by this one indicator. About 300 trades
Often i got moonsine as one of the crap aic inidicators

So if you used AIC, and im less keen on it still - you need to comment out all but 1 indicator, and see where the profit is.
Next build of gsb has a indicator called zero. It returns only zero

v1 = inidcator1;
v2 = gsb_zero; so if aic gives this you know its not used.
v3 = indicator3;

should you use zero with compare or cross entry types and multiple operand, zero will NEVER be used

the reason is
inidcator1*zero*indicator3 will always = 0
Thats the beauty of multiply operand in gsb architecture, and why its slightly better than add operand

I suspect the same flaw is in No Conflictcross entry type, but I havnt looked into that.

admin - 22-12-2020 at 07:32 PM

can all users look for workers that say pending stop
Icon will be 107 (pink) or icon about 96.
If in doubt kill all workers and restart them.
Im not sure why this has occurred, but I cant kill them on my end.
Update on the gold project. Im working on indicator selection. The green pre filter on indicators works well, but I feel there is big room for improvement.
I dont yet have the solution though

admin - 23-12-2020 at 04:47 AM

problem above is resolved. ignore

admin - 30-12-2020 at 11:56 PM

Just an update on Gold
Ive found one indicator that works very well, and making other indicators that are improved on this.
Need updates to GSB to test some new methods - user definable parameters. Its not that simple to do as we need to preserve compatibility with older indicators.
Also making changes (new versions) of other indicators.
My limited testing shows highest high/lowest low works better on daily bars than intra day - and there is a lot less possible combinations of lengths needed
Here are my current gold systems, im trading all or most of these on eminis

meldinman - 4-1-2021 at 11:25 AM

Hi peter,
Any thoughts on adding indicators that use 2 data sources?
Some of my mean reversion systems use relative value between two data streams. I think there would be a lot of potential systems if one could add relative value/correlation indicators.

idea of improvements

bartek - 4-1-2021 at 06:54 PM

Peter, Have you ever tried to implement market filtering approach, where you recognize if market is in trend or trading range and depends on the recognition, apply different set of indicators? From my discretionary trading experience, type of the market is the first thing to recognize to be successful in day trading.

admin - 4-1-2021 at 08:17 PM

Peter, Have you ever tried to implement market filtering approach, where you recognize if market is in trend or trading range and depends on the recognition, apply different set of indicators? From my discretionary trading experience, type of the market is the first thing to recognize to be successful in day trading.

I thought about it in the past (just in ts) tried it, and never succeeded
Though I think other traders have succeeded.
What I found it the markets that were not trending how I like often were profitable, just less profitable. ie if vix This whole concept is easy to curve fit, but if we applied the logic to 50,000 or 300 systems if would be much more solid.
Im open to ideas on all of this.

admin - 4-1-2021 at 08:22 PM

Hi peter,
Any thoughts on adding indicators that use 2 data sources?
Some of my mean reversion systems use relative value between two data streams. I think there would be a lot of potential systems if one could add relative value/correlation indicators.

I have thought about it, and I think the idea has merit.
Have not talked to programmers about it yet, but will do so.
There are tweaks that are simple in the pipeline that have got immediate significant update. Whats been worked on right now is user customizable parameters, and a number of new indicators

bartek - 4-1-2021 at 08:49 PM

So idea is to divide market to 3 or 2 types: e.g. Trending, Trading Range, Reversal. Try to recognize these types using some indicators like ADX or similiar. In next step run optimization engine for specific type with set of indicators. In my opinion this is how current institutional computers run. If we have strong trend, they start trend robots adding to the market. If Trading Range they run robots based on such indicators like stochastics.

idea of improvements

bartek - 4-1-2021 at 09:06 PM

My another idea of improvement is how to manage computational power in GSB community. I think we can apply P2P logic where we measure upload/download power. With such logic the more power we share, more we can use. I think this significantly increase amount of servers in the GSB cloud. The GSB cloud is the main reason I've purchased license.

admin - 4-1-2021 at 09:07 PM

Hi Batek, can you start by finding some things that are know to work on some existing TS systems?

bartek - 4-1-2021 at 09:34 PM

Hi Batek, can you start by finding some things that are know to work on some existing TS systems?

Of course I can try to find out my own developed system having some promising results. I have TS account. But I think trying to optimize for specific systems will take decades vs GSB with implemented methodology. This is worth considering.

admin - 4-1-2021 at 09:42 PM

Hi Batek, can you start by finding some things that are know to work on some existing TS systems?

Ive not studied market regimes at all, so need to see some specific examples of what would work before this is put into GSB. Im open to doing it, but likely this is a long term goal, not short term. There is many more years development work to go onto GSB - as im sure you appreciate.

RandyT - 5-1-2021 at 10:52 AM

Hi Batek, can you start by finding some things that are know to work on some existing TS systems?

I believe that SF and TF mechanisms provide the way to apply filters to certain system logic to determine what type of market the system is applied to. It may just be a matter of having indicators that are reliable for identifying these market regimes, which is the bigger challenge IMO.

Some of the recent discoveries with length applied to SF indicators may also apply.

admin - 5-1-2021 at 03:32 PM

Randy, I think it would be good to do some reading on the topic, and try some known concepts on existing systems. I think BTA group do this sort of thing, but im not sure if its under NDA. My problem is things i have tried over the years might help a little, but come at the cost of still reducing trades that are still profitable to trade. Its just not a subject I have persued greatly.

admin - 10-1-2021 at 07:22 PM

just an update on gold. New stuff I am trying is going really slow as its taking lots of human time (mine) and gsb still needs updates to custom parameter.
While I feel I can do better in gold, I can say existing systems are performing well. I put that down to ideal market conditions

account1 14.2 profit
account2 12.8k profit
doing 8 emini contracts all up, but some started more recently
shown also feb month of gc results

admin - 11-1-2021 at 06:17 PM

I would recommend you all look at the last post on pa pro, as I think its quite significant for all users.
https://trademaid.info/forum/viewthread.php?tid=127&page=3#p...

erlendsolberg - 13-1-2021 at 02:53 AM

When selecting systems to walk forward out of the B-systems with the new Gold methodology, do you recommend sorting (creating families) by IS, OOS or the whole period (IS + OOS)?

admin - 13-1-2021 at 03:09 AM

Quote: Originally posted by erlendsolberg

When selecting systems to walk forward out of the B-systems with the new Gold methodology, do you recommend sorting (creating families) by IS, OOS or the whole period (IS + OOS)?

a good question.
when the macro has finished running stats, Its on the last oos period (2017.6 onwards)
This definitely should not be used.

Ive used the In sample period a year ago, but later are using all dates.
I dont think its critical.
The top families are ranked first but how common they are, then by the fitness with the dates that are used.
So what dates are used will not affect the top families chosen, but will affect the parameters.
A wf should be trying a wide range of parameters regardless.

If you need me to explain in more detail I can do so in this theory.
But the bottom line is use all dates, else In sample

meldinman - 13-1-2021 at 07:26 AM

interesting discussion. after some time using GSB I have seen some systems I created fail out of sample and some work well. SO far GC methodology has worked quite well for building GC systems but on other markets it struggles. Seems like each market has its own wrinkles that one must account for to get it right. Do you find that to be the case with the current methodology? Would you look at an equity curve of a potential CL system the same way as a GC system? is it just an understanding of each market and idea of what a suitable system should look like?

admin - 13-1-2021 at 06:04 PM

interesting discussion. after some time using GSB I have seen some systems I created fail out of sample and some work well. SO far GC methodology has worked quite well for building GC systems but on other markets it struggles. Seems like each market has its own wrinkles that one must account for to get it right. Do you find that to be the case with the current methodology? Would you look at an equity curve of a potential CL system the same way as a GC system? is it just an understanding of each market and idea of what a suitable system should look like?

GC is perhaps the easiest market, but thats also because we figured out whats unique to it. There still is significant upside to go in GC. Some not released indicators show promise too.
CL has also been harder since end of april, while as gold has been as good as it gets the entire year.

Im not sure what you mean exactly by "Would you look at an equity curve of a potential CL system the same way as a GC system?" the the likely answer is yes.

What we learnt from GC, will likely have to be re-applied to all markets.
But you are very correct. Each market has something unique to it, and once thats found much better success awaits.

meldinman - 14-1-2021 at 08:14 AM

Well using GC methodology on other markets has given mixed results. Which is what makes me curious. Seems like each market you have developed had tweaks/updates to the methodology ES - CL/NG - GC etc. My testing using GC methodology on KC has produced some strong systems in theory although forward observation has been lackluster. I have some hunches as to why (similar to things you observed on CL systems) but I have yet to prove that.

One may think that this is due to certain markets being "easy" for GSB to find systems, but it is more likely based on the nature of how each market moves and how to use GSB to capitalize on that. Genetic algorithms are like blood hounds, they will do a good job tracking down the scent you are after but if you give them the wrong scent they wont find what you are looking for.

Do you believe, after developing systems on various markets, that the current methodology will continue to hold up across various markets or that one must understand the unique characteristics of a market and the methodology will evolve to something different for each market? How much weight would you put to developing a robust method across all markets vs finding a strong method that works for a particular market?

admin - 14-1-2021 at 03:59 PM

Well using GC methodology on other markets has given mixed results. Which is what makes me curious. Seems like each market you have developed had tweaks/updates to the methodology ES - CL/NG - GC etc. My testing using GC methodology on KC has produced some strong systems in theory although forward observation has been lackluster. I have some hunches as to why (similar to things you observed on CL systems) but I have yet to prove that.

One may think that this is due to certain markets being "easy" for GSB to find systems, but it is more likely based on the nature of how each market moves and how to use GSB to capitalize on that. Genetic algorithms are like blood hounds, they will do a good job tracking down the scent you are after but if you give them the wrong scent they wont find what you are looking for.

Do you believe, after developing systems on various markets, that the current methodology will continue to hold up across various markets or that one must understand the unique characteristics of a market and the methodology will evolve to something different for each market? How much weight would you put to developing a robust method across all markets vs finding a strong method that works for a particular market?

The current methodology likely will work on many markets unchanged. But what takes time on a new market is the session time(s) used, secondary filter(s) entry type,
and then perhaps something that's unique to a market. ie CL dont enter on the first hour. (that will be done by genetic algorithm later)
Some markets require pattern tests, day of week / month patterns and until we have that, some markets wont be GSB friendly. Thats ok as if GSB never got better, it could still crack many markets. Ive done nothing on KC at all so no comments.
One user put a lot of work with success into fx markets, but is not sharing this until they make an improved video. You also need to figure out if lack of forward performance is market conditions. I looked at a very old non GSB ES system yesterday. Very linear curve for many years, but now at max dd last few months. Thats my take on most day trading ES systems, regardless of them being human or machine designed. To me this implies they likely are not busted. Nearly everything in metals is new highs. New indicators may also help things. There are a few new ones in the next build that im sending to 3 early adopters today

admin - 1-2-2021 at 03:45 AM

This comment might not belong here, but this is one of the most read threads, and so I will put it here.
I have not digested the full implications of this... but recommend you all read
We are possibly in for significant stock market correction short term.
I am not an expert in this area, and have not fully digested the information.

https://www.zerohedge.com/markets/its-not-just-robinhood-red...
https://www.zerohedge.com/news/2021-01-30/robinhood-brink-co...
https://www.cnbc.com/2021/01/31/silver-futures-jump-7percent...
Comments are welcome

meldinman - 1-2-2021 at 07:31 AM

As systematic traders our goal is strive to be agnostic to market direction and refrain from predicting future market behavior. That being said, I think it quite apparent to most observers that the markets have entered into quite overheated territories. Unfortunately, markets by nature tend to move against the consensus and it seems that most believe that a bursting of this bubble is imminent and are salivating for another march 2020 event. If this is the case, we probably see some sort of rolling correction in time that will frustrate both sides. I'm not so sure how that will affect ES systems but after a wild 2020 I will be happy to get through the next few months in one piece. In the meanwhile I continue to look at other markets to develop as I see larger trends beginning to take form across various commodities.

OUrocketman - 1-2-2021 at 05:51 PM

I think recent market events have certainly reinforced the importance of having as large a number of possible of noncorrelated systems across as many markets as possible. Though, getting to that point is certainly a lot of work. I imagine diversification among secondary filters is also important to achieve this.

Also, I think it's important to have some sort of 'index' that system developers can compare their performance to. This helps to better differentiate between a bad system and a bad market. Say for example, you develop 30 gold systems, but trade the "best 5" from some starting time. How does say the "best 5" initially selected compare to the average performance of the other systems not selected over time? If they all do terribly, and say, CTA index for that market is down, then you can probably take comfort in knowing that particular market is just bad presently. If the systems not selected are good, and the "best 5" are now the worst 5, that could be a fairly good indicator that some / all of those systems may need to be replaced.

At least those are the kind of thoughts I have bouncing around in my head lately. It can be hard enough to develop one robust system though, much less a robust library from which to draw systems.

admin - 1-2-2021 at 06:09 PM

I think recent market events have certainly reinforced the importance of having as large a number of possible of noncorrelated systems across as many markets as possible. Though, getting to that point is certainly a lot of work. I imagine diversification among secondary filters is also important to achieve this.

Also, I think it's important to have some sort of 'index' that system developers can compare their performance to. This helps to better differentiate between a bad system and a bad market. Say for example, you develop 30 gold systems, but trade the "best 5" from some starting time. How does say the "best 5" initially selected compare to the average performance of the other systems not selected over time? If they all do terribly, and say, CTA index for that market is down, then you can probably take comfort in knowing that particular market is just bad presently. If the systems not selected are good, and the "best 5" are now the worst 5, that could be a fairly good indicator that some / all of those systems may need to be replaced.

At least those are the kind of thoughts I have bouncing around in my head lately. It can be hard enough to develop one robust system though, much less a robust library from which to draw systems.

I think there is merit to what you say, and I have enough gsb and non gsb systems often to get an idea about how this things are going. Other clues are market range
2005 2006 was very stressful for me, as nearly all my systems were day trading on ES or min russell. You just couldn't make $ doing that in that period. With the wisdom of hindsite that is now very clear, but it wasn't at the time.
I don't have the resources to do what you ask, but the idea is good.
Im also reviewing some old system setups that were used on various markets in gsb (not that have been published) and its very clear that setup is not correct. IE I have a number of systems that are on the wrong session time. They happen to be going well but I expect them to fail earlier than those on the correct session time.

admin - 10-2-2021 at 12:13 AM

Bruce with additions from me made good progress on emini dow (@YM)
no data pre 2007 was used in the build process.
GSB needs some new features to make this as well as I would like. They were planned after customizable indicator lengths
Whats really good sign is the same system went well on ES with zero changes in parameters too.
Emini russell - RTY also, but not as good.

Whats also great is this trades very different to anything else I have seen

OUrocketman - 10-2-2021 at 10:05 AM

Bruce and Peter,

Those are very nice results. Noticing the Long and Short profit gap is smaller than what I’ve observed on the handful of systems I currently trade as well!

Was this done with essentially the same methodology we are following on gold or are there some market specific nuances here? Also, do you view performance on similar markets as a must-have in a candidate system or do you consider it icing on the cake when it does happen?

Also, another general question I have on walk forward testing— how many months do you recommend we leave out of sample for the walk forward parameter optimization? For example, Gold currently we have our IS period end on or around 10/09/2020. Should we optimize until, say, 12/31/2019, and then verify the optimized parameters result in improved performance over the period from 1/1/2020 until today? Or, am I really putting too much thought into this and optimizing over the entire range is generally fine, provided parameters are stable and WF OOS looks like current parameters performance?

Thanks!

admin - 10-2-2021 at 05:35 PM

Bruce and Peter,

Those are very nice results. Noticing the Long and Short profit gap is smaller than what I’ve observed on the handful of systems I currently trade as well!

Was this done with essentially the same methodology we are following on gold or are there some market specific nuances here? Also, do you view performance on similar markets as a must-have in a candidate system or do you consider it icing on the cake when it does happen?

Also, another general question I have on walk forward testing— how many months do you recommend we leave out of sample for the walk forward parameter optimization? For example, Gold currently we have our IS period end on or around 10/09/2020. Should we optimize until, say, 12/31/2019, and then verify the optimized parameters result in improved performance over the period from 1/1/2020 until today? Or, am I really putting too much thought into this and optimizing over the entire range is generally fine, provided parameters are stable and WF OOS looks like current parameters performance?

YM was done with entrytype crossandClosed sf CloseToHighLow3
This is all very new and I don't normally even use crossandClosed
(A feature that was experimental and should not have been released)
I think the better way would be tertirary filter CloseToHighLow3
and sf Not normalized CLoseD
I have not used tertiary filters at all yet. Im more keen for filters which is in GSB but not populated with filters that are useful yet

I optimize over all data in the date file. (2007 to 2020)
This gives a few more months OOS, but I always use wf parmaters, not non wf.
Note also wf optimizes over global dates, not dates. This is a long standing bug that needs to be fixed.
Hope this helps. good questions and Im playing with YM right now.

Thanks!

admin - 10-2-2021 at 06:49 PM

Im just starting on YM today, not sure how much I will pursue it.
Big picture is I have not build ES systems on GSB for perhaps 2 years, ... and GSB is so much better at building systems since then.
My assumption on YM & other had been, we need aritectural changes. IE pattern filters, tertiary filters etc
Good chance I'm wrong. Just did a test with 139 systems in Favourite B. Thats good for end result of lots of testing on a market, so considering IM just starting its excellent.
Obervation is its the newer inidcators that work well, not the original 47 or so

admin - 10-2-2021 at 11:26 PM

An update on YM.
YM is working exactly like I would have expected GSB to work years ago, with no changes to what we always did- except have the beta indicators
quick results look like secondary filter of closeLessClosedbpv (VIP NOT GA)
My test setup is
@YM (NOT @YM.D) 830 TO 1500 Central USA time 30 min bars
data 1.1.2007 to 9.30.2020
fitness np/dd
using all the same macros as we do for gold
one pass green max 10 indicators

I am not using tertiary filters. Not sure why but they did not work for me

YM is working great, esp as this is first runs on it, and there is always lots of room to improve. I would expect ES to be even better

meldinman - 10-2-2021 at 11:32 PM

I'm getting similar things as you, though I figured this may be a fun time to give tertiary filters a try. Only thing seems to be a heavy bias towards 2008 and 2020 where the bulk of profits are. WF doesnt seem to like that...

admin - 11-2-2021 at 03:13 AM

update again.
Ive think I made a critical mistake in dates of indicator selection, so need to re do everything.
I used post 2017 dates instead of pre 2017 dates. This breaks out of sample seperation from in sample.
Will keep you all informed.
Bottom line is I like Bruces YM system in that OOS pre 2007.1.1 is great which was not the case on the last few of my YM systems I tested
and it worked well on other markets.

admin - 11-2-2021 at 06:02 PM

I made another mistake
tried 5 different secondary filters 4 times, but has secondary filter on not normalized closeminusCloseDbpv by mistake (not ga and the SF I chose)
This means sf was closeminusCloseDbpv
here is the variation in fav B for what is 10 identical tests.
This shows the variation possible on identical tests, and why I run 4 tests to get a more accurate picture.

Ave. Fav. B 49.85

Fav. B: 25
Fav. B: 25
Fav. B: 25
Fav. B: 27
Fav. B: 33
Fav. B: 34
Fav. B: 34
Fav. B: 36
Fav. B: 39
Fav. B: 41
Fav. B: 42
Fav. B: 44
Fav. B: 56
Fav. B: 58
Fav. B: 59
Fav. B: 60
Fav. B: 76
Fav. B: 78
Fav. B: 102
Fav. B: 103

On issue is on YM is these indicators gave really good results, but don't match TS. the issue is in GSB not ts code
CloseToHighLow8
CloseToHighLow7
Im now trying other sf on YM again, but keen to get those faulty indicators fixed

admin - 11-2-2021 at 06:09 PM

When I get a new system on a new market, I like to optimize each input individually in TS over a very wide range, find the rough range to be used
then optimize them all together in EWFO
This takes ages to do in human and cpu time, but has been very worthwhile at times
There are two better ways to do this.
1) use the ts API. Docs included.
2) put inputs into a array multi dimensional array in ts, with each parameter set having a index number
then all inputs in ts can indirectly be optimized in one pass.
This is not explained well but you might get the idea.
Is anyone willing to take this project on?

Attachment: Login to view the details

Attachment: Login to view the details

sfuser108 - 11-2-2021 at 08:42 PM

Can someone please clarify what "inputs" need to be optimized in TS? The systems that get generated by GSB don't have any inputs?

Not sure what optimization is being referred here.

Thanx

Carl - 11-2-2021 at 10:33 PM

Hi sfuser108,

Please see the GSB code in TS.

There is a block called // inputs
And then you can see the text "vars:"

Change vars: to inputs: and the inputs will become visible in TS

admin - 12-2-2021 at 02:01 AM

Can someone please clarify what "inputs" need to be optimized in TS? The systems that get generated by GSB don't have any inputs?

Not sure what optimization is being referred here.

Thanx

they don't need to be optimized as GSB does this, but first time on a new market I like to investigate a system in much more detail

admin - 12-2-2021 at 03:11 AM

Im doing a big run on ym.
3 more days work to test all sf.
email me share key if your willing to get the info and contribute you CPU

Likely I will email out a summary of gold in a weeks time. Gold developemtn is stalled as I'm waiting on updates to GSB - custom parameters.
Regardless if I stoped work on gold now, I would be very happy. Have finalized my gold portfolio with 9 systems, caped at max of 5 contracts.

Piet - 12-2-2021 at 03:46 AM

send you an email for my 8 machines. Like this kind of contribution!

Steps After Families are found

sfuser108 - 12-2-2021 at 04:26 PM

Fundamental question - How do you go about picking the right family to trade in real time after you have created the families and you get several..

Lets say off the 100 Fav B, the families result in the following

Family/Number of members
1/19
2/8
3/6
4/6
5/4
6/2
7/2
8/2
etc

What procedure do you do after to arrive with the ultimate system to take to TS for possibly live trading?

admin - 13-2-2021 at 01:44 AM

Fundamental question - How do you go about picking the right family to trade in real time after you have created the families and you get several..

Lets say off the 100 Fav B, the families result in the following

Family/Number of members
1/19
2/8
3/6
4/6
5/4
6/2
7/2
8/2
etc

What procedure do you do after to arrive with the ultimate system to take to TS for possibly live trading?

Great question
I like the top few families as they are most robust as less sensitive to parametaters.
wf them
look for good astab-c and less important rasta-c
look for brown curve (oos) that doesn't go down or flat
look for a linear system equity curve

Here are some very good examples.

Note GSB now users astab-c and rstab-c in the graphs which is superior to astab and rstab

admin - 15-2-2021 at 07:21 PM

I just emailed out a lot of info on building gold systems to those who contributed to the gold project.
If you did contribute, and your not on this email list - then email me. I can then verify if you gave me your share key.
Whats included in the email out is what secondary filters work on the multiple session times. (All were tried = a lot of CPU power went into this)
Gold basically has 2 very different session times.
I showed out of sample results of all systems, real time results of all systems, and some tricks that are unique to gold to get better out of sample results.

For those who did not contribute, I'm fairly open about the 2am to 330am info, but not sharing about the other time frame.
I just think that's the fair thing to do - reward those who contributed, and if I gave everything away - then that's no reward for the contributors.
the 2am to 330 am systems are the most profitable, but other time frame & secondary filters give significant diversification.

If you missed out, you can at least contribute to the dow project... email me your share key.

admin - 15-2-2021 at 07:31 PM

I was asked, how doe the share key work.
open a worker in c:\gsb\workers
go to app settings, share key
enter say your email+random code
joeblogs@gmail.com981234
save app settings.
kill all workers (via gsb resource manager)
start some workers via gsb resource manager)

you can use any name, but using email makes it easier for me to track the contrubutors

admin - 15-2-2021 at 09:46 PM

Here is another YM system. I dont have high faith in this system, and not sure I will put the time into it until I have better YM setup

The oos is in blue. pre 2017 and post 20200930

Its a wonder to behold its current winning streak.
counter trend and trend following.

Unlike Bruces YM system, it didn't do well b4 2007. This is an orange flag.

meldinman - 16-2-2021 at 07:59 AM

If you where more interested in this equity curve what further work would you do? Would you individually WF each parameter in EWFO to see if parameters are robust? would you first try other similar iterations in GSB?

admin - 16-2-2021 at 04:17 PM

If you where more interested in this equity curve what further work would you do? Would you individually WF each parameter in EWFO to see if parameters are robust? would you first try other similar iterations in GSB?

There is lots to this system I'm cautious about. ie it used moonsine which I'm skeptical on. Moonsinre rated highly in indicator testing - which surprises me.
So I would optimize each input one at a time in ts, and look for the correct range, bell curve in results and not red flags. ie parameter just randomly affects results with no clear ideal range.
I put some work into this system allowing 3 inputs to be optimized as if it was done by itself in ts.
Rough proof of concept.

Im going to add this into gsb AI WF. basically might have gsb make a HTML file with the optimization curves in one document, and the table of the results

admin - 17-2-2021 at 12:38 AM

Im changing how I do dow
dates 200811 to 20020228 to avoid extreme volatility.
Big part of the profits made were march 2020
Im also trying ES on similar dates, as I think ES will be easier and what works well on ES, may work well on YM
Next build new entry types that allow ClosedminuscloseBPV to work with secondary filter GA.
All this not needed if tertiary filters were working.
Those who donated to the cloud might not get any cpu usage till you have 61.17 build
There are some new ELD's in the build.

Carl - 17-2-2021 at 03:03 AM

Peter
Im going to add this into gsb AI WF. basically might have gsb make a HTML file with the optimization curves in one document, and the table of the results

Maybe something like this?

admin - 17-2-2021 at 03:20 AM

Carl,
I was thinking of just the ai wf curves all in one HTML document. The current setup is cumbersome that you have to click on each parameter.
It is however much faster than a wf of each input in TS
AI wf has enormous potential to grow, but is very complex.
Im doing ym and es systems at the moment, and finding high level wf very useful

Here is what I found today.
Just been playing around high level with ES.
Issues are do you use pre 2018 or post feb 2020 in system building.
I suspect es systems are similar to ym, just easier and faster to make.

admin - 18-2-2021 at 06:43 PM

im still 30 hours away of completing my first pass of secondary filters on ES. Bottom line is there is clear potential for systems that trading nothing like GSBSYS1ES

Also have not figured out many things. ie what years is best to use for insample data
possible start dates are 1997, 2000, 2008.11 start dates

admin - 19-2-2021 at 04:24 AM

making some progress
1997 start date here

admin - 19-2-2021 at 05:51 PM

some of these new systems are just so different from each other, and from anything else we have.
sys1 screen shot from today.

system6 report, which is till today.

Sys6 Note no trades for months, and march where just about everything made a killing... made nothing

Carl - 20-2-2021 at 08:56 AM

Got a nice one on ES today.

Build and selection fitness NP x R x NP/DD

Chart is after costs and slippage. Stoploss 2k

admin - 21-2-2021 at 04:01 PM

Got a nice one on ES today.

Build and selection fitness NP x R x NP/DD

Chart is after costs and slippage. Stoploss 2k

Looks good
whats the start date, secondary filter, entry type?

Carl - 22-2-2021 at 12:27 AM

Hi Peter, here are some details

ES Nth IS 2007-2017, crossANDclosed, SF closetohighlow3

bartek - 22-2-2021 at 09:52 AM

Got a nice one on ES today.

Build and selection fitness NP x R x NP/DD

Chart is after costs and slippage. Stoploss 2k

What is R fitness criteria?

Carl - 22-2-2021 at 12:08 PM

R = pearson

Bfan - 22-2-2021 at 07:15 PM

Hi Peter, here are some details

ES Nth IS 2007-2017, crossANDclosed, SF closetohighlow3

Hi Carl,
Were you able to match it in TS? It seems whenever the new set of indicators of Closetohighlow1 thru 9 were involved I couldn't match anything.
Thanks

admin - 22-2-2021 at 07:42 PM

Quote: Originally posted by Bfan

Hi Peter, here are some details

ES Nth IS 2007-2017, crossANDclosed, SF closetohighlow3

Hi Carl,
Were you able to match it in TS? It seems whenever the new set of indicators of Closetohighlow1 thru 9 were involved I couldn't match anything.
Thanks

GSB_CloseToHighLow8 and GSB_CloseToHighLow7 don't match and they have a conceptial bug. Problem will be fixed, but is not easy to fix. (GSB issue)
Ceptial problem in gsb and ts, and mismatch issue in gsb too

However the some of the series of GSB_CloseToHighLowx are total gems.

Bfan - 22-2-2021 at 07:48 PM

Quote: Originally posted by Bfan

Hi Peter, here are some details

ES Nth IS 2007-2017, crossANDclosed, SF closetohighlow3

Hi Carl,
Were you able to match it in TS? It seems whenever the new set of indicators of Closetohighlow1 thru 9 were involved I couldn't match anything.
Thanks

Thank you Peter!

meldinman - 22-2-2021 at 09:30 PM

agree with peter, have had a lot of success with some of those indicators on various markets.

Gold Portfolio

LucaRicatti - 1-3-2021 at 04:34 PM

Just a week of GSB research to get this "low correlated" gold future portfolio.
Many thanks to Peter and to everyone sharing CPU power !

NickW - 1-3-2021 at 04:50 PM

Hi Luca,

Well done! Are all these systems using the same timeframe or different timeframes?
Also, are they all using different secondary filters?

Nick

NickW - 1-3-2021 at 04:50 PM

dup

RandyT - 1-3-2021 at 04:58 PM

Luca,

I would add to Nick's question, are these the same session, or different sessions?

Very impressive results.

LucaRicatti - 1-3-2021 at 05:20 PM

Thanks,
In that portfolio:

- 2 different Secondary Filter
- 3 different sessions
- data1 or data1+data2

You can also add two same session systems to a single chart and limit the entry to one contract to keep dd low

sfuser108 - 1-3-2021 at 06:46 PM

@Luca
These look very impressive... Is the corr matrix OOS or the whole data set?

Cheers

admin - 2-3-2021 at 05:51 PM

@Luca
These look very impressive... Is the corr matrix OOS or the whole data set?

Cheers

Pretty sure this is the case.

LucaRicatti - 3-3-2021 at 05:44 AM

@Luca
These look very impressive... Is the corr matrix OOS or the whole data set?

Cheers

Yes, followed Peter's method, standard gold macros / dates.

bartek - 3-3-2021 at 12:44 PM

Have you tried bitcoin futures with promising results?

Bruce - 10-3-2021 at 12:55 AM

An NQ update, I completed some new builds earlier in the year and over the past week or so have had some awesome results from real trades, one day last week delivered a $20k day.

These builds were delivered using AU, build dates 2007/01/01 to 2017/06/30, 8 indicators for the Indicator test, Build = Goldv1.1, WF all data to 2020/02/28. SF = CloseToHighLow3V2, appears to be the go to SF for ES, NQ and YM. Performance report (with Comm and slippage) below is for OOS only (2017/06/30 to 2021/03/09) and I've included a real-world equity chart for the past couple of weeks.

Peters made some further advancements on these systems and I suspect with the most recent GSB release there's undoubtedly more that could be achieved.

I hope this helps...

Screen Shot 2021-03-10 at 7.30.10 PM.png - 208kB

Screen Shot 2021-03-10 at 7.29.21 PM.png - 122kB

Screen Shot 2021-03-10 at 1.58.58 PM.png - 63kB

admin - 10-3-2021 at 02:56 AM

Awesome bruce. How many contracts / systems traded to get your 25k of wins?

Systemholic - 10-3-2021 at 03:38 AM

hi bruce, i don't seem to be able to find CloseToHighLow3V2 in my AU. i did see CloseToHighLow3 but not the V2. Are they the same?
Also Peter i also cannot find your closeLessHighLowv3 . Is it available already?

tks

Carl - 10-3-2021 at 08:59 AM

Great news, Bruce. Thanks for sharing.

I got some great strategies on ES by using the CloseToHighLow3V2 secondary filter, crossANDclosed, ES 30 min, session 0830-1500.

Did you use the usual session time 08:30 to 15:00 as well or did you get even better results using alternative session times?

All the best

@Systemholic, CloseToHighLow3V2 differs from CloseToHighLow3.

CloseToHighLow3
if (close > prevDayClose) then
result = hhNorm
else if (close > prevDayClose) then
result = -1 * llNorm;

CloseToHighLow3V2
if (close > prevDayClose) then
result = hhNorm
else if (close <= prevDayClose) then
result = -1 * llNorm;

RandyT - 10-3-2021 at 09:20 AM

@Systemholic, CloseToHighLow3V2 differs from CloseToHighLow3.

CloseToHighLow3
if (close > prevDayClose) then
result = hhNorm
else if (close > prevDayClose) then
result = -1 * llNorm;

CloseToHighLow3V2
if (close > prevDayClose) then
result = hhNorm
else if (close <= prevDayClose) then
result = -1 * llNorm;

Actually, CloseToHighLow3V2 is just the latest (corrected) version of CloseToHighLow3. If you are using current version of GSB, you will get the V2 indicator.

admin - 10-3-2021 at 03:19 PM

Great news, Bruce. Thanks for sharing.

I got some great strategies on ES by using the CloseToHighLow3V2 secondary filter, crossANDclosed, ES 30 min, session 0830-1500.

Did you use the usual session time 08:30 to 15:00 as well or did you get even better results using alternative session times?

All the best

@Systemholic, CloseToHighLow3V2 differs from CloseToHighLow3.

CloseToHighLow3
if (close > prevDayClose) then
result = hhNorm
else if (close > prevDayClose) then
result = -1 * llNorm;

CloseToHighLow3V2
if (close > prevDayClose) then
result = hhNorm
else if (close <= prevDayClose) then
result = -1 * llNorm;

830 to 1500 should be used
CrossandCloseD works, but crosssingle is best.

Bruce - 10-3-2021 at 03:50 PM

Thanks Carl. These results are with the normal 830_1500 session times, however, I do also use 930_1500 for some of my systems as I have a personal bias that 830_930 is typically 'amateur hour' though that may be dispelled statistically!

Carl - 11-3-2021 at 06:09 AM

830 to 1500 should be used
CrossandCloseD works, but crosssingle is best.

Thanks, Peter.

I have tried PF entry mode crosssinglelevel, but my results are not as good as with crossandclosed.

I wonder what difference in our GSB settings is causing this different outcome.

admin - 11-3-2021 at 07:01 PM

830 to 1500 should be used
CrossandCloseD works, but crosssingle is best.

Thanks, Peter.

I have tried PF entry mode crosssinglelevel, but my results are not as good as with crossandclosed.

I wonder what difference in our GSB settings is causing this different outcome.

i was getting on es about 100 fav b with cross and close and about 200 with cross.
WIll publish my nq settings in a while. Perhaps due to your tight stop?? I use $2000

Carl - 12-3-2021 at 07:45 AM

i was getting on es about 100 fav b with cross and close and about 200 with cross.
WIll publish my nq settings in a while. Perhaps due to your tight stop?? I use $2000

For my first test runs I used a 2000 USD stoploss as well.

I suspect it's my trainings filter and fitness function that is guiding me in a different direction.

For short only ES strategies based on crossANDclosed and closetoHL3 I am getting 180 in FavB.
For long only strategies much less.

admin - 12-3-2021 at 02:46 PM

@carl Fitness we use np/dd How many long / short in fav B. Did you start at 1997 on ES data

Carl - 12-3-2021 at 04:44 PM

@carl Fitness we use np/dd How many long / short in fav B. Did you start at 1997 on ES data

Hi Peter,

I've tried several build fitness functions.

Crosssinglelevel gets me 248 strats in FavB. crossANDclosed only 142.
My tests show better quality strategies with crossANDclosed, the average trade is higher than with crosssinglelevel.

Short only about 180 in FavB, long only just around 20 in FavB

Only 1 GSB run for each test, so the reliability can be improved by doing more of the same GSB runs.

admin - 12-3-2021 at 05:20 PM

@carl Fitness we use np/dd How many long / short in fav B. Did you start at 1997 on ES data

how do you define "My tests show better quality strategies with crossANDclosed"
Its an interesting comment.

you can also build with cross single, and after system build put in a close + offset>Closed filter

RandyT - 12-3-2021 at 05:28 PM

@carl Fitness we use np/dd How many long / short in fav B. Did you start at 1997 on ES data

Carl, I've seen this exact behavior and I do consider Average Trade to be an important metric when ranking systems. This has all led to the voices in my head having a very heated conversation about what is the best measurement of the results we are seeing.

NickW has started looking at these results in the context of Count x AT x FavB#. I think this is an interesting way to measure, but again, I worry that we are overlooking better systems that may not occur that often by weighting them with higher numbers found instead of the truly unique market filters.

The approach of building short/long only systems is also interesting. I've found that certain entry modes work better for shorts vs. longs and it has made me question my approach of trying to build short/long combined systems. I've looked at doing short/long only in the past but never really embraced it.

admin - 12-3-2021 at 09:46 PM

@Carl, Randy
there is merit to long only / short only, but less trades and lower peasons etc. Ive played but never felt it really worth while
you could do long, short then combine the metrics etc to compare with long & short. Not sure its worth all the effort

as for rare gems of systems that are hard to find. Im still of the mindset that the stronger the overall metrics, fav B, the higher change of good out of sample.
Even with highlow3 secondary filter, it seems easy to get systems with a really low correlation to each other on losing trades

Carl - 13-3-2021 at 01:08 AM

Interesting discussion guys!

@Peter, I find ES strategies based on crossANDclosed + closetoHL3 better, because the average trade on the long side is good enough to trade live.
For a majority strategies that end up in FavB using crosssinglelevel + closetoHL3, the average trade on the long side is only about 80 USD. After subtracting costs and slippage, only a net average trade of 50 USD remains.
Changing the training filter and/or build fitness might improve the results. I have to test this.

@Randy, "what metrics to use" or "how am I going to build and select strategies". That a difficult one, but very important. Maybe the most important issue in the whole development process.

I have done a few proof of concept tests by using Python scripts on GSB data.
Let machine learning models look for relationships between the GSB indicators, the in sample metrics and the outcome in out of sample.
It seems that when using all indicators the machine learning models indicate the indicators are the most important.
When I select a smaller group of indicators, the ML models show that the in sample metrics are becoming more important.
It is even possible to look for the best predicting metrics to come up with a "best" fitness function.
The issue is that results are changing for every GSB run, so the results aren't stable.
And what about the curve fitting risks by doing this?

Here is one of the "feature importance" tables from randomforestregressor in Python:

OUrocketman - 13-3-2021 at 03:21 AM

This is an interesting discussion indeed.

Carl, interesting work using machine learning in an effort to elicit the most probable predictors of future success!

At a high level, I think it's interesting that for large number of indicators the importance is on the indicators. Seems to support the overall notion that when we down select the indicators, we are beginning to sniff out interesting features that while may not easily be intelligible by human inspection, are in fact there. Once this is accomplished, it seems to make intuitive sense that the big money has knowledge of this as well, and seeks to make more money in the future in the most efficient way possible, based on the underlying features that tend to move a market--thus the higher weight on performance.

So, it seems like, at a high level, your feature extraction verifies in some way Peter's two pass approach. I'm curious to know have you considered ASTAB-C, RSTAB-C, WFE, OOS fitness as features as well and they didn't make the cut on the importance scale in your screenshot above?

Also, have you checked out Microsoft's open source Light Gradient Boosted Method? It's rumored to out perform randoforest, but I'll confess I'm not currently familiar enough with the topic to assess it--longer term goal of mine.

MS stuff is here: https://lightgbm.readthedocs.io/en/latest/

admin - 13-3-2021 at 03:32 AM

This is an interesting discussion indeed.

Carl, interesting work using machine learning in an effort to elicit the most probable predictors of future success!

At a high level, I think it's interesting that for large number of indicators the importance is on the indicators. Seems to support the overall notion that when we down select the indicators, we are beginning to sniff out interesting features that while may not easily be intelligible by human inspection, are in fact there. Once this is accomplished, it seems to make intuitive sense that the big money has knowledge of this as well, and seeks to make more money in the future in the most efficient way possible, based on the underlying features that tend to move a market--thus the higher weight on performance.

So, it seems like, at a high level, your feature extraction verifies in some way Peter's two pass approach. I'm curious to know have you considered ASTAB-C, RSTAB-C, WFE, OOS fitness as features as well and they didn't make the cut on the importance scale in your screenshot above?

Also, have you checked out Microsoft's open source Light Gradient Boosted Method? It's rumored to out perform randoforest, but I'll confess I'm not currently familiar enough with the topic to assess it--longer term goal of mine.

MS stuff is here: https://lightgbm.readthedocs.io/en/latest/

Im no longer doing 2 pass. I do one pass force max 10 indicators, min 4
I think its likely just as good as 2 pass, but its faster.
also on nq I found 3 indicators on the first pass was better than 2.
For other markets I'm not yet sure.
@carl.
Ave trade can be boosted by long short day of week filters, the Andrea Unger pattern filters and changing the highlow3 to GSB_CloseToHighLow3v4_offset after you have built the system

Tertiary filters now work in gsb, but at a quick test don't match. Pattern filters are being worked on now.

Carl - 13-3-2021 at 05:22 AM

Thanks OUrocketman,

There are a lot of boost models: light boost, catboost, xgboost.
My teacher always says: try an ensemble of models to see what works best on your data set.

I have used the WF metrics in previous data analysis projects.
The issue with WF is, I only have a couple of hundred rows of data.
And it is easy to get 100k rows of data with build and validation results.
By using sampling I divide the GSB test data into an in sample and an out of sample part, so a large number of data points is more reliable.

I will look for these old WF test results and will let you know.
Too high parameter stability like 100% seems very good at first glance, but isn't always the case because 100% stability could also mean only a particular set of parameter values give the best result, but making a small change in parameter value might cause the results to collapse.
So 100% parameter stability doesn't always mean the strategy is robust (I think...).
I prefer 50% to 80% GSB parameter stability, but always do an extensive rolling and anchored WF in TS and EWFO to be sure.

With all this said, not all the strategies I have selected according to this selection process were profitable going forward.

admin - 19-3-2021 at 05:37 PM

Just an update for the week.
I had many builds of GSB to try. Mainly working on custom parameters, new indicators, and learning NQ market.
Some of the new indicators are very good, and there are numerous indicators in the pipeline.
I did show some screen shots in the beta build section.
I think NQ is right now the hottest market. Metals have gone quite.
The closetohighlow3 works on many markets like ES NQ YM and likely many more
Last few days I got stuck in that I cant reproduce the results I had a few days earlier, and don't yet know why.
There things can be very time consuming to resolve.
The goal is to do the next video on NQ. ES and likely YM are 95% the same build process.

admin - 23-3-2021 at 03:50 AM

Here are some indicator testing on nq830 start, but start trading at 930
and start trading at 830. Note the accumdist variations on the top end of the list

rws - 27-3-2021 at 05:23 PM

Peak optimizers are always having risks.

There was a system in the past that had shift of parameters
from 1 to 5 up and down for all parameters in the system.
I have been testing this for the last year.
You could have all kind conditions where systems were rejected for example if the profit of the system was not 90% of the original settings with parameter shift of 1 to 5. I find 3 is often enough.
This was stratasearch a system that was discontinued but it had some great ideas and works good on portfolios of stocks where it can find good OOS especially with this principle of parameter stability. I use it to run on about 400 of the SP500 stocks and if IS is good there is a very good chance that OOS is good. Walkforward is almost not necessary with so many tickers and good parameter stability.
What you typicaly see is if 2S or 3S setting goes down very rapidly, you will often have bad OOS performance.

There is complete controll in this way how stable you would like to have your parameters for many different metrics and for the selected shift size. You will see there will be easily 10 times less systems found but the ones that are found have a much higher chance of having good OOS in different market conditions. It is one of the best ways to avoid a system that had a lucky peak IS.
The downside of this program is that it only works on daily data but I am convinced that the principle also works intraday.

@ Peter, please evaluate this kind of parameter stability it works.

One way to predict the SP500 index is to count the number of buy/sell signals of it's stocks which seems to work reliable.

parameter stability avoiding peak optimization.png - 69kB

Carl - 28-3-2021 at 04:59 AM

Hi rws,

What do you mean by "shift"?
Do you mean shifting the parameter value over bars?
So a shift of 2 is delaying the parameter value by 2 bars?
Or so mean just changing the parameter input values?

I agree an important goal is to find strategies with good parameter stability.
I think that is exactly what Peter is trying to do in the development process he describes in his Youtube video's about GC, CL and ES.

step 1. build strategies based on Nth IS price data
step 2. select strategies based on Nth IS or Nth(IS+OOS) performance
step 3. back-test the top 300 selected strategies on three separate years OOS
step 4. select the strategies that also perform well in OOS and copy these strategies to FavB
step 5. determine the different families in FavB. A family is a set of strategies that use the same combination of indicators

If a family in FavB is large, this means a lot of strategies with different parameter values (but the same indicators!) end up in the good performing strategies. So the larger the family, the better the parameter stability.

@Peter, please correct me if I am wrong in explaining your method.

rws - 28-3-2021 at 05:59 AM

Hi Carl,

With 1shift it means that an optimized RSI value of 10 is changed to 9 and 11 and results compared to 10

With 2shift it means that an optimized RSI value of 10 is changed to 8 and 12 and results compared to 10.

This has some disadvantage when parameters are very small. It will give a message when there is a parameter value of 3 and there is a shift of 3 but the building will continue. This doesn't happen very often but in effect it would avoid systems with small values of parameters so there is room for improvement. But the final result is that only systems are build with good parameter stability.

In GSB as a result of a big family with the same systems and different parameter values it would have the same confirmation but it is not as direct, visable and configurable as in Stratasearch.

In Stratasearch you can see the performance degrade as a result of the shift in parameters and/or you can rejects systems while building that have no parameter stability by just coding some simple rules like:
profit with shifted parameters by 2 > profit original params x0.8
profit with shifted parameters by 3 > profit original params x0.75
drawdown with shifted parameter by 1 > drawdown original x 1.15x

etc
$V_5SPctProfitable > $V_PctProfitable x 0.7 meaning that if parameters were shifted by 5, the % profitable is maximum degraded by 0.7 times

So it gives a very clear indication if the current parameters are not in a peak and how wide this "non peak" is. You have controll about how wide you want the "non peak" and based on what criteria and you can use it as a rejectable metric while building.
Stratasearch builds in stages, first systems, then optimizations of parameters and optional adding intermarket rules like breadth, sector strength etc. There can be many rules for all kinds of metrics. It also allows to build system on SP500 where as a metric there is a rule that the system should also work for a certain % on other markets like Nasdaq as a rejectable rule. So it does not build on Nasdaq in that case.

I know there is parameter stability in GSB but as you noted that even with 100% there still can be instability OOS so I mentioned the way it was done in Stratasearch. I don't know a direct comparison in GSB vs stratasearch in this case.

I am an early adopter of GSB but didn't use GSB for a while but within a couple of weeks I have time again and I read about many good additions.

I am not here to advocate stratasearch because it is useless for futures and commodities real time. I also don't advice to testdrive stratasearch because it takes months to test every aspects and it is a discontinued program. But after putting in many hours and long building times I am convinced it is possible to build reliable systems OOS by having this broader user configurable parameter optimization.

Of course you can build many systems and try as much parameter changes in the hope you have the broader peak but I can't see it directly in GSB or instruct GSB to have a broad peak. How do you know you have enough builds with different parameters for the system? You can confirm in Tradestation if there was no peak but
so far as I can see not a way to build a broad peak.

I wanted to test a system builder that supports intermarket relations which Stratasearch has and while searching for that I found that this way of parameter optimization in Stratasearch was more important than intermarket relations.

I think it is a valuable addition to GSB.

Hi rws,

I think that is exactly what Peter is trying to do in the development process he describes in his Youtube video's about GC, CL and ES.

step 1. build strategies based on Nth IS price data
step 2. select strategies based on Nth IS or Nth(IS+OOS) performance
step 3. back-test the top 300 selected strategies on three separate years OOS
step 4. select the strategies that also perform well in OOS and copy these strategies to FavB
step 5. determine the different families in FavB. A family is a set of strategies that use the same combination of indicators

If a family in FavB is large, this means a lot of strategies with different parameter values (but the same indicators!) end up in the good performing strategies. So the larger the family, the better the parameter stability.

@Peter, please correct me if I am wrong in explaining your method.

rws - 28-3-2021 at 10:41 AM

You can even have a custom criteria that is searching for
and optimizing stable parameters apart from profit and
average trade. That often works a bit better than only excluding
limits because it allows systems that initially weren't good because
of bad parameters to get improved.

$V_AvgTrade$V *AvgAnnReturn / $V_LossDays
*($V_MaxDrawDU / $V_1SMaxDrawDPctU)^2
*$V_MaxDrawDU / $V_2SMaxDrawDPctU
*$V_MaxDrawDU / $V_3SMaxDrawDPctU

If you want even broader stability then you could add 4 and 5 shift
*$V_MaxDrawDU / $V_4SMaxDrawDPctU
*$V_MaxDrawDU / $V_5SMaxDrawDPctU

And in addition you can set limits for
*$V_MaxDrawDU / $V_XSMaxDrawDPctU
so systems are rejected

So this is NP * AT times a relation of the drawdown
for the current parameters divided by the drawdown when
there is 1, 2 and more shifts.
$V_LossDays is the number of average days for losing trades.

Quote: Originally posted by rws

Hi Carl,

With 1shift it means that an optimized RSI value of 10 is changed to 9 and 11 and results compared to 10

With 2shift it means that an optimized RSI value of 10 is changed to 8 and 12 and results compared to 10.

This has some disadvantage when parameters are very small. It will give a message when there is a parameter value of 3 and there is a shift of 3 but the building will continue. This doesn't happen very often but in effect it would avoid systems with small values of parameters so there is room for improvement. But the final result is that only systems are build with good parameter stability.

In GSB as a result of a big family with the same systems and different parameter values it would have the same confirmation but it is not as direct, visable and configurable as in Stratasearch.

In Stratasearch you can see the performance degrade as a result of the shift in parameters and/or you can rejects systems while building that have no parameter stability by just coding some simple rules like:
profit with shifted parameters by 2 > profit original params x0.8
profit with shifted parameters by 3 > profit original params x0.75
drawdown with shifted parameter by 1 > drawdown original x 1.15x

etc
$V_5SPctProfitable > $V_PctProfitable x 0.7 meaning that if parameters were shifted by 5, the % profitable is maximum degraded by 0.7 times

So it gives a very clear indication if the current parameters are not in a peak and how wide this "non peak" is. You have controll about how wide you want the "non peak" and based on what criteria and you can use it as a rejectable metric while building.
Stratasearch builds in stages, first systems, then optimizations of parameters and optional adding intermarket rules like breadth, sector strength etc. There can be many rules for all kinds of metrics. It also allows to build system on SP500 where as a metric there is a rule that the system should also work for a certain % on other markets like Nasdaq as a rejectable rule. So it does not build on Nasdaq in that case.

I know there is parameter stability in GSB but as you noted that even with 100% there still can be instability OOS so I mentioned the way it was done in Stratasearch. I don't know a direct comparison in GSB vs stratasearch in this case.

I am an early adopter of GSB but didn't use GSB for a while but within a couple of weeks I have time again and I read about many good additions.

I am not here to advocate stratasearch because it is useless for futures and commodities real time. I also don't advice to testdrive stratasearch because it takes months to test every aspects and it is a discontinued program. But after putting in many hours and long building times I am convinced it is possible to build reliable systems OOS by having this broader user configurable parameter optimization.

Of course you can build many systems and try as much parameter changes in the hope you have the broader peak but I can't see it directly in GSB or instruct GSB to have a broad peak. How do you know you have enough builds with different parameters for the system? You can confirm in Tradestation if there was no peak but
so far as I can see not a way to build a broad peak.

I wanted to test a system builder that supports intermarket relations which Stratasearch has and while searching for that I found that this way of parameter optimization in Stratasearch was more important than intermarket relations.

I think it is a valuable addition to GSB.