The ACD Method

[Maverick74]

The other day I was rewatching a seminar that Mark Fisher recently gave with CQG, and thought it was chock full of great trading insights - even when Fisher doesn't say much he is still sharing a lot value...which got me thinking....would it be possible for us as a group to maybe come up with 5 or 10 questions and email them to him? He seems to keep a pretty profile, but perhaps he'd be interested in responding if he knew there was a 1300-page thread here dedicated to his work?

He knows about the thread. I've told him. He is not terribly interested in ACD these days. If any of you guys live in the Miami area, he does have an office down there and has a small group of guys there. I can point you in the right direction.

 

[Maverick74]

Sorry guys not been around for a while but just caught back up again on the thread!!

As I have mentioned quite a few times I really gravitated to R not really sure how or why, but it has done me just fine. From starting out as an absolute beginner I have now written functions for NL scoring and backtesting along with a handful of other system tests from ideas I picked up in Acrarys threads.

Although most of you guys use ACD as a kind of tool box, I was able to go beyond that slightly whereby I am using ACD from a more strict quantitive setup. Beating the same drum as Mav, the edge in ACD is being able to view market information, in away that is not necessarily the same as others would see it. There are literally thousands of way you can apply ACD data and I never become tired of testing them.

With regard to learning to code the Datacamp courses look excellent. I haven't used them myself but they look great at getting you started. The best thing I can suggest is to start coding something that will be useful to you from the start, then work your way upon from there. For example write yourself a code that grabs a load of stock data and then computes monthly A levels, then make it run though your dataset so you have some history then automate it and store the results etc. The more you struggle the more you will learn but its very satisfying creating hypotheses writing the code and testing. The more your curiosity increases the more your programming will improve.

This is exactly what we do. I'm working on my R skills as well. But we have a massive database of NLs for every product in this world and the next and it's all data analysis. I don't even look at charts anymore. LOL.

 

[Maverick74]

Is there a particular data provider that you prefer? Thanks.

DTN is good too.

 

[redbaron1981]

I don't even look at charts anymore. LOL.

I am with you on that one haven't opened up Esignal for charting in ages just use the data manager, same goes for IB. Being able to compute real probabilities and edge has been a huge confidence booster to me and I am surprised I never did it sooner.

Literally every spare hour I get goes into R, who needs social media!! When I think of something new its great being able to pull apart/reuse code that I have used before making the modifications for individual requirements. Not sure if you have tried it yet but R also has a few useful scraping packages which are quite handy for that not so easy to get data.

 

[Maverick74]

I am with you on that one haven't opened up Esignal for charting in ages just use the data manager, same goes for IB. Being able to compute real probabilities and edge has been a huge confidence booster to me and I am surprised I never did it sooner.

Literally every spare hour I get goes into R, who needs social media!! When I think of something new its great being able to pull apart/reuse code that I have used before making the modifications for individual requirements. Not sure if you have tried it yet but R also has a few useful scraping packages which are quite handy for that not so easy to get data.

Which packages are those?

 

[redbaron1981]

For most of the stuff I have coded Rvest and XML have worked fine it really depends on what your trying to get your hands on. Once you have your scraped data you can use stringr to extract certain parts of strings followed by the base functions "as.numeric" and "as.date" etc for conversion to a useable format for data frames.

By the way not sure if you have tried it yet but if you are dealing with lots of data I have started using data.table which is extremely fast although it syntax is a little more complicated than using the piping in Dplyr.

If you need any help feel free to PM me.

 

[sss12]

For most of the stuff I have coded Rvest and XML have worked fine it really depends on what your trying to get your hands on. Once you have your scraped data you can use stringr to extract certain parts of strings followed by the base functions "as.numeric" and "as.date" etc for conversion to a useable format for data frames.

By the way not sure if you have tried it yet but if you are dealing with lots of data I have started using data.table which is extremely fast although it syntax is a little more complicated than using the piping in Dplyr.

If you need any help feel free to PM me.

@Maverick74 @redbaron1981 or anyone else.

Forgive my ignorance on the programing aspect of trading.

So after you have whatever outcome your program produces from your data inputs,is that a stand alone solution ? Or is that just one factor in an investment/ trade decision ?

For ex: how would a programmed action/decision jive with a political event like the Comey hearings or a bianary event such as Brixit ? Thanks for the info.

 

[baggerlord]

He knows about the thread. I've told him. He is not terribly interested in ACD these days. If any of you guys live in the Miami area, he does have an office down there and has a small group of guys there. I can point you in the right direction.

I find it strange that he isn't very interested in ACD anymore. Do you have any more insight on this?

 

[baggerlord]

Although most of you guys use ACD as a kind of tool box, I was able to go beyond that slightly whereby I am using ACD from a more strict quantitive setup. Beating the same drum as Mav, the edge in ACD is being able to view market information, in away that is not necessarily the same as others would see it. There are literally thousands of way you can apply ACD data and I never become tired of testing them.

This is where I am at too. I can say that ACD has allowed me to do some stuff that quite frankly is beyond my natural level of ability. I don't consider myself a very sophisticated person when it comes to markets, statistics, or trading but I can honestly say that I am getting some pretty unique stuff going on my computer. I keep finding myself on the right side of the market and I feel like I am still just barely getting to understand how to use the ACD tools I've built. This thread has truly been a game-changer for me.

 

[baggerlord]

An example of what I mean by barely understanding...I had the idea today to use a NL derivative and apply it to a pair trade strategy. IE play the mean reversion when the NLs are in agreement, and play the trend when they diverge.