| Pages:
1
..
32
33
34
35
36
..
98 |
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
Thanks Peter, this is quite interesting.
I'm finding that using this does a pretty good job of ordering results from best to worst. (using Medium)
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by RandyT  |
Thanks Peter, this is quite interesting.
I'm finding that using this does a pretty good job of ordering results from best to worst. (using Medium) |
It was designed to get around systems curve fitting to 2007 to 2008 high volatility. It was a big problem just after this period, but not so much now
as so much data has past.
Is it better than just np*at?
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by RandyT |
Thanks Peter, this is quite interesting.
I'm finding that using this does a pretty good job of ordering results from best to worst. (using Medium) |
It was designed to get around systems curve fitting to 2007 to 2008 high volatility. It was a big problem just after this period, but not so much now
as so much data has past.
Is it better than just np*at? |
It is just my perception at this point but it seems to be a pretty reliable sort as I am digging through SI market.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Randy
46.49 has the average of thje custom metrics, displayed under stats now.
For example, This means you could build using vbasemedium, and make custom field np*at
you can then compare this to the identical test thats built using np*at
Just remember identical tests can give significant variation. If there is occasional big variation, I found it to vary on the upside. So if you get a
big increase, do the test a second time.
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
Unique Systems Period
Could someone provide their thoughts about the best setting for "Unique Systems Period" in App Settings?
Appears the default is "Training". Wondering what impact there is of other settings or if the conclusion is that "Training" is best.
Thanks
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
Optimizing Entry Points
I've done a quick analysis of entry types for SI market and would appreciate suggestions on how to interpret/use this.
Quite a dropoff from the Compare1/2 entry types. Is this significant enough to lean toward eliminating anything other than Compare1/2 in system
development or is it the case that if any of them work, it is useful to explore?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by RandyT | I've done a quick analysis of entry types for SI market and would appreciate suggestions on how to interpret/use this.
Quite a dropoff from the Compare1/2 entry types. Is this significant enough to lean toward eliminating anything other than Compare1/2 in system
development or is it the case that if any of them work, it is useful to explore?
|
Hi Randy
I may change my mind on this, but I prefer going through he whole methodology with each entry type. The first two are by far the easiest for GSB to
find, which can account for popularity. Hence why I like to test one at a time.
So try
Compare1
Compare2
Cross. // If this isnt good, ignore all other cross's
AnyIndicatorCross
NoConflictCross
The others are beta (alpha) user features only, and likely shouldn't be used.
The alpha features are things I test, and dont always stay around if they arnt worth having)
For CL< I found compare1&2 best. (likely Silver too) ES The cross's best. (mildly dual cross best)
D
It all takes time, but the rewards are worth the research.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by RandyT | Could someone provide their thoughts about the best setting for "Unique Systems Period" in App Settings?
Appears the default is "Training". Wondering what impact there is of other settings or if the conclusion is that "Training" is best.
Thanks |
As im following the methodology, the basic filters are in training.
I dont like to filter unique systems as we want to see the good and bad.
I like the families feature which shows how popular a family is and lets you pick the top family member
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by RandyT | I've done a quick analysis of entry types for SI market and would appreciate suggestions on how to interpret/use this.
Quite a dropoff from the Compare1/2 entry types. Is this significant enough to lean toward eliminating anything other than Compare1/2 in system
development or is it the case that if any of them work, it is useful to explore?
|
Hi Randy
I may change my mind on this, but I prefer going through he whole methodology with each entry type. The first two are by far the easiest for GSB to
find, which can account for popularity. Hence why I like to test one at a time.
So try
Compare1
Compare2
Cross. // If this isnt good, ignore all other cross's
AnyIndicatorCross
NoConflictCross
The others are beta (alpha) user features only, and likely shouldn't be used.
The alpha features are things I test, and dont always stay around if they arnt worth having)
For CL< I found compare1&2 best. (likely Silver too) ES The cross's best. (mildly dual cross best)
D
It all takes time, but the rewards are worth the research. |
For SF-Entry Mode, does it only have ability to use the enabled Entry Modes?
Reason I ask is that on current SI run, I am finding that the Cross entries seem to be used in the better performing systems, but the SF-Entry is
predominately Compare1/2. (image attached)
And perhaps I don't understand what the Secondary Filter is, but why is it using Data1? Should that not be Data2, or does "Secondary" not mean
secondary data stream?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by RandyT | | Quote: Originally posted by admin | | Quote: Originally posted by RandyT | I've done a quick analysis of entry types for SI market and would appreciate suggestions on how to interpret/use this.
Quite a dropoff from the Compare1/2 entry types. Is this significant enough to lean toward eliminating anything other than Compare1/2 in system
development or is it the case that if any of them work, it is useful to explore?
|
Hi Randy
I may change my mind on this, but I prefer going through he whole methodology with each entry type. The first two are by far the easiest for GSB to
find, which can account for popularity. Hence why I like to test one at a time.
So try
Compare1
Compare2
Cross. // If this isnt good, ignore all other cross's
AnyIndicatorCross
NoConflictCross
The others are beta (alpha) user features only, and likely shouldn't be used.
The alpha features are things I test, and dont always stay around if they arnt worth having)
For CL< I found compare1&2 best. (likely Silver too) ES The cross's best. (mildly dual cross best)
D
It all takes time, but the rewards are worth the research. |
For SF-Entry Mode, does it only have ability to use the enabled Entry Modes?
Reason I ask is that on current SI run, I am finding that the Cross entries seem to be used in the better performing systems, but the SF-Entry is
predominately Compare1/2. (image attached)
And perhaps I don't understand what the Secondary Filter is, but why is it using Data1? Should that not be Data2, or does "Secondary" not mean
secondary data stream?
|
Entrymodes are for the primary filters, not secondary.
secondary will chose the two modes likely closeD-close or CloseD/close
Secondary filter terminology has nothing to do with data stream.
I understand your confusion
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by RandyT | | Quote: Originally posted by admin | | Quote: Originally posted by RandyT | I've done a quick analysis of entry types for SI market and would appreciate suggestions on how to interpret/use this.
Quite a dropoff from the Compare1/2 entry types. Is this significant enough to lean toward eliminating anything other than Compare1/2 in system
development or is it the case that if any of them work, it is useful to explore?
|
Hi Randy
I may change my mind on this, but I prefer going through he whole methodology with each entry type. The first two are by far the easiest for GSB to
find, which can account for popularity. Hence why I like to test one at a time.
So try
Compare1
Compare2
Cross. // If this isnt good, ignore all other cross's
AnyIndicatorCross
NoConflictCross
The others are beta (alpha) user features only, and likely shouldn't be used.
The alpha features are things I test, and dont always stay around if they arnt worth having)
For CL< I found compare1&2 best. (likely Silver too) ES The cross's best. (mildly dual cross best)
D
It all takes time, but the rewards are worth the research. |
For SF-Entry Mode, does it only have ability to use the enabled Entry Modes?
Reason I ask is that on current SI run, I am finding that the Cross entries seem to be used in the better performing systems, but the SF-Entry is
predominately Compare1/2. (image attached)
And perhaps I don't understand what the Secondary Filter is, but why is it using Data1? Should that not be Data2, or does "Secondary" not mean
secondary data stream?
|
Entrymodes are for the primary filters, not secondary.
secondary will chose the two modes likely closeD-close or CloseD/close
Secondary filter terminology has nothing to do with data stream.
I understand your confusion |
Peter, I'm confused by the values shown in Parameters which shows an indicator value for "SF-Entry Mode". Image attached.
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
Prospecting in NQ
Struggling to generate systems on NQ30.29.31 data. Does anyone have suggestions?
I've run some indicator validations and am focused on about 15 indicators on NQ as a result of that. I have run most all entry modes which may be a
contributor to my frustration. Trying again here limiting to Cross, CrossSingle and CrossDouble.
Any hints would be much appreciated.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by RandyT | Struggling to generate systems on NQ30.29.31 data. Does anyone have suggestions?
I've run some indicator validations and am focused on about 15 indicators on NQ as a result of that. I have run most all entry modes which may be a
contributor to my frustration. Trying again here limiting to Cross, CrossSingle and CrossDouble.
Any hints would be much appreciated. |
Its been a long time, and i found it a harder market.
Others have more expertise on this, i found the profits were mainly in 3 periods. year 2000 2008 2018 from memory.
best try entry type compare1* compare2,
then try cross double
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by RandyT | Struggling to generate systems on NQ30.29.31 data. Does anyone have suggestions?
I've run some indicator validations and am focused on about 15 indicators on NQ as a result of that. I have run most all entry modes which may be a
contributor to my frustration. Trying again here limiting to Cross, CrossSingle and CrossDouble.
Any hints would be much appreciated. |
Its been a long time, and i found it a harder market.
Others have more expertise on this, i found the profits were mainly in 3 periods. year 2000 2008 2018 from memory.
best try entry type compare1* compare2,
then try cross double
|
Getting slightly better results on a compare1/2 only run.
Looking forward to hearing other people's thoughts/experience on this market. Great volatility in this market. Should be tradeable.
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Randy,
Yes, it is certainly tradeable in recent conditions.
But GSB strategies on NQ only generate signals during 2000-2001 and after 2016.
Thanks received (1):
+1 RandyT at 2020-01-24 13:00:53
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Randy
you could try excluding the volatile years in the build process, but also use them as another oos period.
To build on extreme volatility makes systems that only trade and make profit in highly volatile years
|
|
|
gdk13
Junior Member
Posts: 6
Registered: 6-6-2019
Member Is Offline
|
|
Duplicating GSB Walk Forward Process in Real Time Trading
I want to be sure when I go into implementation that I am scheduling re-optimizations in accordance with what GSB Walk Forward was doing and then
showing in graphs.
GSB appears to train/optimize on around 2 years of data and apply to the following year in the Walk Forward (on an anchored basis). Is this an
accurate interpretation of how GSB is doing the Walk Forward?
Therefore, if true, I would re-optimize at the end of a year?
Thank you.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by gdk13 | I want to be sure when I go into implementation that I am scheduling re-optimizations in accordance with what GSB Walk Forward was doing and then
showing in graphs.
GSB appears to train/optimize on around 2 years of data and apply to the following year in the Walk Forward (on an anchored basis). Is this an
accurate interpretation of how GSB is doing the Walk Forward?
Therefore, if true, I would re-optimize at the end of a year?
Thank you. |
Personally I rarely ever re optimize. Look at GSBSys1ES. The oldest GSB system. Parameters have never been changed and its great OOS 2 + years.
However users may have strong opinions on both sides of the argument.
If you have achieved high parameter stability, parameters are not likely to change.
There are exceptions to this of course.
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Fills and executions.... Is there anyone that has a record of live fill vs backtest for our GSB traded markets using our strategies?
In my setup i use MC and IB, and its very difficult to compare real fills compared to backtest in any automated way using this setup i use, at least i
have not found any way to do it, anyone uses Multichart and IB trading several strategies on same symbol, have found a way? For now i try to keep
track manually but its not optimal.
I know that Peter uses a custom API between TS and IB, and i believe this tracks this somehow? correct Peter?
Do you have any statistics of efficiency for fills for each market ?
Any input here is appreciated
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Daniel UK1 | Fills and executions.... Is there anyone that has a record of live fill vs backtest for our GSB traded markets using our strategies?
In my setup i use MC and IB, and its very difficult to compare real fills compared to backtest in any automated way using this setup i use, at least i
have not found any way to do it, anyone uses Multichart and IB trading several strategies on same symbol, have found a way? For now i try to keep
track manually but its not optimal.
I know that Peter uses a custom API between TS and IB, and i believe this tracks this somehow? correct Peter?
Do you have any statistics of efficiency for fills for each market ?
Any input here is appreciated |
There is a few issues. Slippage, execution issues, trades that do or dont occur on ts/mc after a chart is refreshed. Do you trade all days esp
Christmas + 2 weeks etc
Human error, ie you forgot to roll a contract and it failed to trade.
When you factor all of that in, i cant give solid numbers, esp as Im not putting any time in to fix my tracking if there is an error.
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
This brings up another issue that I would personally like to see as the default in the GSB system script and that is limit orders rather than market.
I've generally had good luck entering on a limit that allows one tick slippage rather than entering on market.
EDIT: Just to follow this up, I have an ES system trading live on Tradestation as broker. I got full point slippage on each side today. That will
never fly.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by RandyT | This brings up another issue that I would personally like to see as the default in the GSB system script and that is limit orders rather than market.
I've generally had good luck entering on a limit that allows one tick slippage rather than entering on market.
EDIT: Just to follow this up, I have an ES system trading live on Tradestation as broker. I got full point slippage on each side today. That will
never fly. |
I see that as an execution issue, not a GSB issue.
Your welcome to change the code in gsb for live trading.
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by RandyT | This brings up another issue that I would personally like to see as the default in the GSB system script and that is limit orders rather than market.
I've generally had good luck entering on a limit that allows one tick slippage rather than entering on market.
EDIT: Just to follow this up, I have an ES system trading live on Tradestation as broker. I got full point slippage on each side today. That will
never fly. |
I see that as an execution issue, not a GSB issue.
Your welcome to change the code in gsb for live trading.
|
Just trying to reduce the amount of code that needs to be changed...
The more code that is changed to get to live trading, the harder it becomes to maintain.
|
|
|
gdk13
Junior Member
Posts: 6
Registered: 6-6-2019
Member Is Offline
|
|
| Quote: Originally posted by gdk13 | I want to be sure when I go into implementation that I am scheduling re-optimizations in accordance with what GSB Walk Forward was doing and then
showing in graphs.
GSB appears to train/optimize on around 2 years of data and apply to the following year in the Walk Forward (on an anchored basis). Is this an
accurate interpretation of how GSB is doing the Walk Forward?
Therefore, if true, I would re-optimize at the end of a year?
Thank you. |
Please comment on how GSB is doing Walk Forward (WF) as far as length of period used for IS and OOS.
Also, is it true that WF re-optimizes for each period on an anchored basis and then uses this for the subsequent period?
If this re-optimization is what we are seeing for results on the graph report, to duplicate this performance, wouldn't we need to also re-optimize at
end of the current period for the upcoming period?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by gdk13 | | Quote: Originally posted by gdk13 | I want to be sure when I go into implementation that I am scheduling re-optimizations in accordance with what GSB Walk Forward was doing and then
showing in graphs.
GSB appears to train/optimize on around 2 years of data and apply to the following year in the Walk Forward (on an anchored basis). Is this an
accurate interpretation of how GSB is doing the Walk Forward?
Therefore, if true, I would re-optimize at the end of a year?
Thank you. |
Please comment on how GSB is doing Walk Forward (WF) as far as length of period used for IS and OOS.
Also, is it true that WF re-optimizes for each period on an anchored basis and then uses this for the subsequent period?
If this re-optimization is what we are seeing for results on the graph report, to duplicate this performance, wouldn't we need to also re-optimize at
end of the current period for the upcoming period? |
Im rusty on the maths. I made a method of dates I liked better, but we changed to be identical to TS WF dates. I think its 40% in sample, then 10 runs
with 20% out of sample. I never will change this, though it can be done.
Yes its anchored, anchored or rolling can be used. I suspect that rolling is a more stressful test, but many tests on GSB show anchored is best for
live trading.
I normally stop my WF date at 2015.6.30 and do not re-optimize after this.
So im not a big fan of re-optimzing - but many people would disagree.
Please comment on how GSB is doing Walk Forward (WF) as far as length of period used for IS and OOS. - see enclosed picture
|
|
|
| Pages:
1
..
32
33
34
35
36
..
98 |
![[*]](images/davis/default_icon.gif){kind=icon}
