| Pages:
1
2
3
4 |
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
NON CHERRY PICKED SYMBOLS; NEED CONTRACTS LIST INFO; FRANTIC, PLEASE HELP!
Can't attach the Excel for someone to fill in and re-post.
Only 4 days left on eval !!! PLEASE HELP! To assess the real power of Builder's formula combining (while also optimizing inputs), I want to
press forward on the idea these symbols are not "cherry picked" for the evaluation. I'm going to look at ones that DIDN'T come with the evaluation
platform: these:
SOMEBODY PLEASE fill in the boxes! It appears each detail must be in place for the GSB run to go forth correctly.
Oh, apparently it DID attach as an Excel. PLEASE HURRY!
Attachment: Login to view the details
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Super Admin, I'm curious, your opinion of using @EMD (Trade Station here) in place of $IDX in all cases?
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Good question.
This is a valid idea. GSB can optimize data streams so if you use both, once a systems is build you can optimize the data streams. Use at least 250 x
250 population / generations. However the few times ive done this, i get better summary metrics, but worse equity curve. ie its not so linear.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Good question.
This is a valid idea. GSB can optimize data streams so if you use both, once a systems is build you can optimize the data streams. Use at least 250 x
250 population / generations. However the few times ive done this, i get better summary metrics, but worse equity curve. ie its not so linear.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Hello, Mr. Zwag, in an interview with you out there or somewhere I recall you with degrees in electronics. Some of the very best traders seem to be
good in the area. Among other reasons, I'm certain it's their understanding of frequency, waves, amplitude, osclillation etc. When merged with
statistics, they are the best.
My question: If you were consigned to pick only one of the following, and had to do it evermore:
1. Trade off a 30 min chart as data1 with data2 being 15 min.
OR
2. Trade off a 15 min chart as data1 with data2 being 30 min.
Which would you pick?
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by parrdo101  | Hello, Mr. Zwag, in an interview with you out there or somewhere I recall you with degrees in electronics. Some of the very best traders seem to be
good in the area. Among other reasons, I'm certain it's their understanding of frequency, waves, amplitude, osclillation etc. When merged with
statistics, they are the best.
My question: If you were consigned to pick only one of the following, and had to do it evermore:
1. Trade off a 30 min chart as data1 with data2 being 15 min.
OR
2. Trade off a 15 min chart as data1 with data2 being 30 min.
Which would you pick?
|
Yes, I love electronics & arduino's. I was a computer engineer years ago, but with hind site wish i started as a programmer.
Option1 is bad as Gsb & TS will treat data 2 as 30 min bars.
data 2 must be the same or bigger time frame, never smaller.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
In TS, when creating the chart to view and save the Data Window into the DropBox, should one always use Exchange Time as circled? Need a clue
whether one can get away with Local Time instead of Exchange time. Thanks.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | In TS, when creating the chart to view and save the Data Window into the DropBox, should one always use Exchange Time as circled? Need a clue
whether one can get away with Local Time instead of Exchange time. Thanks.
|
If you have more than one data stream on a chart, TS forces you to use local.
Many people will not do this, but I set my trading computer to Central USA time.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Hello, Super Admin, Can you, for the symbol SM, (soybean meal, futures), approximate and post to us a custom session that mimics the effect created by
the ES custom session of 8:30 AM to 3:00 PM which you have chosen? What session hours for SM would you posit to mimic this ES situation? Then, after
that, what should one specifically enter in Contracts List for SM under MOC1 FROM AND MOC1 TO?
What would YOU do to mimic this, going now with SM? Thanks.
Additional question:
Are there instances in Contracts List, where one could conceivably type in something in MOC1TO and MOC1FROM that
would prevent the GSB from finding systems? Or does "anything one chooses to put in here" not affect, stop or physically halt the GSB from finding
systems? Perhaps better asked as...stop the GSB from at least finding some sort of system? - as to whether the systems are proper of not,
not my question or issue. Only whether these MOC settings can halt it.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | Hello, Super Admin, Can you, for the symbol SM, (soybean meal, futures), approximate and post to us a custom session that mimics the effect created by
the ES custom session of 8:30 AM to 3:00 PM which you have chosen? What session hours for SM would you posit to mimic this ES situation? Then, after
that, what should one specifically enter in Contracts List for SM under MOC1 FROM AND MOC1 TO?
What would YOU do to mimic this, going now with SM? Thanks.
Additional question:
Are there instances in Contracts List, where one could conceivably type in something in MOC1TO and MOC1FROM that
would prevent the GSB from finding systems? Or does "anything one chooses to put in here" not affect, stop or physically halt the GSB from finding
systems? Perhaps better said as stop the GSB from at least finding some sort of system - as to whether the systems are proper of not, not my question
or issue. Only whether these MOC settings can halt it.
|
I dont have data for sm, but the volume should give the clue to the correct session time. Moc shouldnt halt anything.
Are there instances in Contracts List, where one could conceivably type in something in MOC1TO and MOC1FROM that
would prevent the GSB from finding systems?
No, this feature is needed for tick data where ther may be no data say at 15:00:00 but it might come at 15:00:20
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Say I'm optimizing, ("configuring indicators"), 3 datastreams in a GSB run.
Any final result can be examined for the instances of a calculated value per datastream.
Been any thought to forcing GSB "optimize"-run to giving exact even weight of some kind to each datastream?
One way I'm leading up to - tho many ways I'm sure - would be to require it to produce as close as possible - or exactly - even instances of
calculations-values that contribute to the signal entry.
If madata1(length1) and ma2data2(length2) and ma3data3(length3)...generally, produce 1 value for each datastream that then begets the entry signal.
Now, throwing in all 25 indicators churning a config and forumula, you might fall short of the combined generated values of all datastreams being
divisible by 3
(in TS: If Mod(TotalInstancesOfValuesProducedByFormulas,SumOfEachOfTheseFromAllTHREEdatastreams,3) = 0, but
the good programmer might allow, e.g., (if the GSB churned out this many):
10 values data1
9 values data2
10 values dat3
The above is the instance where it was 1 value short by natures of the GSB production, of each datastream having the same number of values generated
upon it - instances of any generated values. The programmed event would force the instances of produced values, values themselves could be anything
produced by formula, to be as close as possible to the same.
It would not allow:
10 values data1
8 values data2
10 values data3
nor allow: not allow:
10 values data1
9 values data2
8 values data3
This would be ok too:
9 values data1
9 values data2
10 values data 3
If you get what I'm pointing to....close as possible to same.
The odd fellow datastream(s), short a produced value relative the other or others, would still require no datastream(s) to have instances of
values-generated that was more than 1 of any of the other datastream(s). An evenness...all datastreams. by programmed force, evenly employed.
Which leads to my question: Is the GSB run, as currently structured, forcing any kind of even usage of all three datastreams? Please describe in some
layman's terms how, if so.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
There is no way to weight any one data stream, and no plans to do so. The GA will find whats best, and you can later on optimize data streams.
Ive only optimized data streams 10 or so times, and Ive never had an improvement. Sometimes the metrics improved, but the linearity of a curve got
worse.
Often you get the same final parameters. You do need a lot more iterations when you optimize data streams too.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Super Admin:
"You do need a lot more iterations when you optimize data streams too. "
The best ways for extra iterations?
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Suggestion: Each new 14 day trial evaluator goes through this:
1. New evaluator is asked to have 3 RANDOM symbols ready for Data1,2,3. This is ALL the new person selects, and the new person must select them
without Admin's knowledge or help; Admin will select EVERYTHING ELSE, including time frames.
2. Connects 1 time (at least) with Big Super Admin via Team Viewer.
3. Begins recording with something like this the moment is connected with Admin (see image for Flashback Express - a very simple to use and sturdy
video recorder).
4. Is directed by Admin how to gather and save all the data for each of the 3 symbols, but NOT for the recent 10 years.
5. Yes, some things are related/correlated and the night has a thousands tricks of correlation among the instruments, but today Admin's KEEN, keen eye
will show up to try bring order out of chaos (random symbols). (You've got a Doctorate in Wave theory and Statistics onboard guiding you through.) Let
Admin decide ALL the settings; he's got the knowledge and the experience, and we expect he'll apply it all to try to bring the order of profit out of
these random, but inviting-to-be traded-right-now for the thousands by every Exchange out there, symbols. What an invitation; what a tempt!
6. Admin steps through each setting, under recordation, you won't forget it, and the GSB is finally fired up to crank it out - to buoy the trade on
these 3 with all the known - at least best that you could get in here - and it's good and certified - statistics and a programmed-platform.
7. When the script is up (Trade Station), you plop it into the
chart in Trade Station for the recent 10 years. Make sure to add the right slippage. I will bolster this by including some of my 20 years of study for
a rather precise slippage of several symbols that real trading gets when live (see the Excel).
THIS should be the starting point for all, whether the GSB fails in the Out-of-Sample or not.
Optional: A thread is started, just for conversation and curiosity:
Subject: My 3 startup symbols: Mark Won or Failed:
Posters just front 1 word each in their posting regarding the Out-of-Sample that was produced in their session with Admin, and a tally goes forth:
Won.
Won.
Failed.
Won.
Failed.
Failed.
E.G. asf
Keep learning.
Attachment: Login to view the details
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | Suggestion: Each new 14 day trial evaluator goes through this:
1. New evaluator is asked to have 3 RANDOM symbols ready for Data1,2,3. This is ALL the new person selects, and the new person must select them
without Admin's knowledge or help; Admin will select EVERYTHING ELSE, including time frames.
2. Connects 1 time (at least) with Big Super Admin via Team Viewer.
3. Begins recording with something like this the moment is connected with Admin (see image for Flashback Express - a very simple to use and sturdy
video recorder).
4. Is directed by Admin how to gather and save all the data for each of the 3 symbols, but NOT for the recent 10 years.
5. Yes, some things are related/correlated and the night has a thousands tricks of correlation among the instruments, but today Admin's KEEN, keen eye
will show up to try bring order out of chaos (random symbols). (You've got a Doctorate in Wave theory and Statistics onboard guiding you through.) Let
Admin decide ALL the settings; he's got the knowledge and the experience, and we expect he'll apply it all to try to bring the order of profit out of
these random, but inviting-to-be traded-right-now for the thousands by every Exchange out there, symbols. What an invitation; what a tempt!
6. Admin steps through each setting, under recordation, you won't forget it, and the GSB is finally fired up to crank it out - to buoy the trade on
these 3 with all the known - at least best that you could get in here - and it's good and certified - statistics and a programmed-platform.
7. When the script is up (Trade Station), you plop it into the
chart in Trade Station for the recent 10 years. Make sure to add the right slippage. I will bolster this by including some of my 20 years of study for
a rather precise slippage of several symbols that real trading gets when live (see the Excel).
THIS should be the starting point for all, whether the GSB fails in the Out-of-Sample or not.
Optional: A thread is started, just for conversation and curiosity:
Subject: My 3 startup symbols: Mark Won or Failed:
Posters just front 1 word each in their posting regarding the Out-of-Sample that was produced in their session with Admin, and a tally goes forth:
Won.
Won.
Failed.
Won.
Failed.
Failed.
E.G. asf
Keep learning.
|
Your idea is problematic. Starting with unknown symbols is going to waste a lot of everyone's time, including my support time. This is going to slow
down GSB development which affects all users. You should start with what works and then slowly branch out.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
I've noticed a quirk. Tried to save 14 mos. of 30 minute bars, chart ending Jan 1 2002 in Trade Station, then use in GSB.
GSB won't run. It points to that holiday at end chart day. Seen this on some other days, too, I think, that weren't holidays. Turns out these latter
were later in the week, and all, maybe most rather, of these no-runs were actually fixed (then ran in GSB) by moving the chart end date to the nearest
Monday; data save then ran in GSB optimize.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Question: $IDX starts around 1985, $SPX.X much later and @ES around 1998.
Is a Trade Station chart set to Jan 13, 2001, 20 years back, then saved for a run in the GSB going to work? Just noticing it's giving me unique
systems in trying just that. But is there a problem with this?
Being careful on align....
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | I've noticed a quirk. Tried to save 14 mos. of 30 minute bars, chart ending Jan 1 2002 in Trade Station, then use in GSB.
GSB won't run. It points to that holiday at end chart day. Seen this on some other days, too, I think, that weren't holidays. Turns out these latter
were later in the week, and all, maybe most rather, of these no-runs were actually fixed (then ran in GSB) by moving the chart end date to the nearest
Monday; data save then ran in GSB optimize. |
There are a few issues here.
Ive seen an issue on I think it was google stock where it stops premature. Its listed as a bug to be fixed.
You should never run GSB on only 14 months of data, unless there is no choice. Even then I wouldn't do it as the sample size is to small to be valid.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | Question: $IDX starts around 1985, $SPX.X much later and @ES around 1998.
Is a Trade Station chart set to Jan 13, 2001, 20 years back, then saved for a run in the GSB going to work? Just noticing it's giving me unique
systems in trying just that. But is there a problem with this?
Being careful on align.... |
That should run fine
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
If I recall, in at least 2 YouTubes with Peter, he set up and ran the GSB optimizer without any slippage. I seem to recall an adjunct statement like,
"That's not the way quite a few do it, but I do it this way."
Assuming I follow that (no slippage) please fill in the blank:
Best to avoid any system on a walk forward with less than _____ % Parameter stability. (I'm also assuming the more important stability is the Rolling
Stability?)
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | If I recall, in at least 2 YouTubes with Peter, he set up and ran the GSB optimizer without any slippage. I seem to recall an adjunct statement like,
"That's not the way quite a few do it, but I do it this way."
Assuming I follow that (no slippage) please fill in the blank:
Best to avoid any system on a walk forward with less than _____ % Parameter stability. (I'm also assuming the more important stability is the Rolling
Stability?)
|
Good questions and I need to add this into the docs.
I want at least anchored stability for the last 3 runs. This is close but not the same as 30%.
lets say we get 4 rows with all parameters exact, and the other all erratic.
the anchored stability score would be 4 * 9 = 36
I would be happier with 3 * 10 - 30
Basically the more parameters that are not stable in a row, the less the score.
From memory if more than 3 parameters in a row are not stable, the score then goes to zero.
If you get say 2 rows only that are stable, I would expect out of sample performance to be more likely to drop. When I did apple stock systems, I
often got this. (over night system) At the heart of the issue was my apple systems has a bit over 100 trades. I am a bit unconformable with only 100
trades, unless the anchored stability is very high.
I am less concerned with rolling score.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Just running into another question:
Continuing to use my 8:30am to 15:00pm Exchange Time custom session (390 minutes) I saved 3 datastreams into the GSB for its run. The 3 were NQ on 30
min; YM on 120 min; and ES on 480 minutes. The results looked somewhat bad, then I happened to notice the Trade Station script did not use Data2
nor Data3. The Trade Station chart was fitted with the script and run into a large Out-Of-Sample period.
Can you provide insights into how I presume I dumbed up something with this one? This is something one should not do, the sessions vs bar size etc.?
The script did produce some entries.
|
|
|
curt999
Junior Member
Posts: 51
Registered: 24-7-2017
Member Is Offline
Mood: No Mood
|
|
you would likely have more luck trying multiple timeframes of the same instrument..ie nq30 nq 120 nq 480 or closer spaced like 30 60 90
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by parrdo101 | Just running into another question:
Continuing to use my 8:30am to 15:00pm Exchange Time custom session (390 minutes) I saved 3 datastreams into the GSB for its run. The 3 were NQ on 30
min; YM on 120 min; and ES on 480 minutes. The results looked somewhat bad, then I happened to notice the Trade Station script did not use Data2
nor Data3. The Trade Station chart was fitted with the script and run into a large Out-Of-Sample period.
Can you provide insights into how I presume I dumbed up something with this one? This is something one should not do, the sessions vs bar size etc.?
The script did produce some entries. |
You sound like you just chose a system with a number of trades.
You need to run GSB for some time, then chose the systems with the best metrics. GSB will chose the best data streams that it finds best. Rule of
thumb is dont change too many things all at once - otherwise you dont know what works and what doesnt.
Bar times larger than data1 on the same sybol is ok, and other symbols lie nq etc would be ok to try.
RC2 er/es which was is in the top 10 of futures truth had a number of data streams futures and or cash, but they were all 30 min. It is ok to try
bigger time frames than your data1.
|
|
|
parrdo101
Junior Member
Posts: 71
Registered: 18-11-2017
Member Is Offline
Mood: No Mood
|
|
Now in a custom session of 390 min (8:30am to 3pm Central) I saved data files into GSB, putting a 480 min as data1 (on that cust session),
the 120 and 30 as secondary. Won't produce systems.
Then saved a 390 file in place of the 480 and saved a 195 min in place of the 120, keeping the 30 as is, then put the 390 as
data1 (Primary) the 195 and 30 as secondary. Won't produce systems. I hope I remember this.
DOUBLE CHECKED:
All data files, 30, 195 390 on this 390 min cust sess were exactly the same period as shown:
The script continues to TOTALLY ignore the existence of the 195 min data file in all instances. Scrolled through many random scripts, findings were
all as below in triple quotes.
Script consistently still shows complete ignoring of the existence of the 195 min bar data: Out of the scripts:
"""
// Settings
// ID: 20171208-051627-498797-YxRhy
// Price Data: Data1: ES.30.Minute.20yback.endJan29.2010.txt, Data2: ES.390.Minute.20yback.endJan29.2010.txt
// MaxBarsBack: 500
"""
This is just one sample, but I've seen several instances of "new world possibilities" using stuff like 15, 30, 30 min datastreams discussed in these
forums, and I'm just wondering if something like the problem above has escaped attention to here - on those.
Changed one variable to the above; changed the ES 195 data file to YM 195 data file, then inserted among secondary datastreams as had done the ES
before, then ran GSB optimize; it's completely ignoring the existence of the 195 min bar data file - exact same result for this:
"""
// Settings
// ID: 20171208-051627-498797-YxRhy
// Price Data: Data1: ES.30.Minute.20yback.endJan29.2010.txt, Data2: ES.390.Minute.20yback.endJan29.2010.txt
// MaxBarsBack: 500
"""
Something in the program? It should ignore a 195 min bar on a 390 min custom session?
(Above needs to be significantly re-exhibited and worked, botched much methodical sci and demos - use for in-flight and gist only - low on time -
accidentally deleted a .jpg - lost -no time - may come back.)
|
|
|
| Pages:
1
2
3
4 |
![[*]](images/davis/default_icon.gif){kind=icon}
