Your favourite building blocks

Hi!

What indicators, operators etc. have proven to work best for you in building blocks?

Depends on what you are trying to build, short term range bound strategy, longer term trend follower, breakout strategy.

I find it is best to have an idea for a strategy, maybe based on a London open breakout for example.  SQ works best for me if I set it up with one type of strategy in mind…

I see. Thanks for your reply.

How about the number of building blocks used in strategy generation?

Have you observed what would be too many or too few tickboxes checked?

I see. Thanks for your reply.

 

How about the number of building blocks used in strategy generation?

Have you observed what would be too many or too few tickboxes checked?

Its all about failing and learning upon your mistakes..

then you just create a favorite template of your settings and use it to create your strategies with it,

definitely you should approach SQ with an idea first and let SQ generate different combinations of your idea to see whats works best for you,

Most definitely consider purchasing the data from asirikuy this is M1 data from 89, that way you will get even further information about what you testing rather then just using dukascopy’s data from 2004,

this way your can use 89~2004 as InSample and 2004~today as an OutOfSample with asirikuy data, and plus to that you can use the dukascopy’s TICK data to make a double check on your OutOfSample on your candidates..

i think i told your the biggest secret here hehe:),

so yeah…, basically just use that long data from 89 to get a huge edge, because if some strategy worked from 89 till today you most definitely got a good working strategy, (depends on how you do stuff and dont curve fit it to the data it self)

go with simple strategies that way it would not be curve fitted, use InSample + OutOfSample, Montecarlo and etc to make sure..)

just saved you a couple of months of trial and error there :), good luck!

Its all about failing and learning upon your mistakes..

 

then you just create a favorite template of your settings and use it to create your strategies with it,

 

definitely you should approach SQ with an idea first and let SQ generate different combinations of your idea to see whats works best for you,

Most definitely consider purchasing the data from asirikuy this is M1 data from 89, that way you will get even further information about what you testing rather then just using dukascopy’s data from 2004,

this way your can use 89~2004 as InSample and 2004~today as an OutOfSample with asirikuy data, and plus to that you can use the dukascopy’s TICK data to make a double check on your OutOfSample on your candidates..

 

i think i told your the biggest secret here hehe:),

so yeah…, basically just use that long data from 89 to get a huge edge, because if some strategy worked from 89 till today you most definitely got a good working strategy, (depends on how you do stuff and dont curve fit it to the data it self)

 

go with simple strategies that way it would not be curve fitted, use InSample + OutOfSample, Montecarlo and etc to make sure..)

 

 

just saved you a couple of months of trial and error there :), good luck!

Read the eBook.

Huge thanks guys!!

I read the e-book, the one by Zdenek Zanka, right? (see attached screenshot). Great book, concise and useful.
Just thought I’d try out to ask this also to see if you seasoned SQ-users had some more wisdom regarding this to share. Indeed got new valuable info here, I am very grateful!  🙂

File: Capture.JPG

I find it’s best to be really specific, as few building blocks as possible….

For example, if I’m doing trend following, first I decide what sort of trend following, like a MA cross over, or a linear regression, etc… If I’m doing linear regression, I would probably only select linear regression, and ADX. If it was MA cross over, probably only the 4x types of included MAs, and MACD. If it was breakout, just the highest/lowest in range, and open a close for the day. So, very specific. Also, select building blocks that go together with a strategy theme, don’t just randomly select stuff.

There’s somethings that I keep no matter what I’m trying, like profit trailing, stop loss, etc..but generally, as few building blocks as possible, because I want it to develop on the idea I have, and explore that idea specifically, not just random stuff. I will sometimes mix it up, like try a trailing stop loss instead of trailing profit, but I always have some stops, either profit or loss. I don’t really use profit/risk ratios. ATR, I keep pretty loose, to not stop out on noise.

Oh, yeah, your operators questions, I use them all except or/illogical. I’m don’t use or, but I use all the others, sure.

Also, keep the indicator and price look back periods really short. I find keeping it short is best, at least. I mean..I develop pretty much mostly on 60 minute or 1D, if I go with 620 indicator and price look back, that’s YEARS ago, and probably pure luck if it finds something. Much better (in my experience) to have look back really, really recent, like highest in price range of a few hours or a day or so ago….a break out is something like today is higher than yesterday and day before, not today is higher than Thursday the year before last, or 2.5 months ago.

So, very few building blocks, very recently, very specific….then let it generate for a week+….you want THOUSANDS of strategies, because 80% will fail the first OOS, then the next 19.99999% will fail the next OOS…

(Removed in retrospect because it was an embarrassing beginner question and didn’t add any value, haha)

Depends a bit on what I’m looking for. So, if for example it’s something like a MA cross over on a m60, then something like a few days back, tops.

So…a little dependant. Not super specific, sorry, it’s something you’ll get through experience. For example with moving average cross over trend following a SMA, a 200 period SMA is a pretty standard filter, so probably 200 would be max in that sort of system. If I’m doing a momentum breakout, then it’s probably a couple of days back, max a week…

Here’s some more commentary on look back periods (maybe not super helpful, but a more authoritative voice than mine commenting on luck vs edge in look back periods)….you could always start with however many bars is ~a week, based on your time frame, and go from there….

hi,

I tried to buy Zdenek Zanka ebook on amazon. Its not available. Where I can buy it?

Hello, yes if you do not own SQ licences the ebook still can be obtained through Amazon https://www.amazon.com/gp/product/B01BDTOCU0 

Hello

I`m new in here

Is this Ebook free for SQ pro licences or I have to buy this?

Hello,

SQ license owners can get it for free. If you wish to receive a copy, send me a request to [email protected]