The ACD Method

koolaid said:
Not sure if I'm scoring my NL correctly...but if you wait for the 30 NL to confirm...you're usually very late to the party.
More...

Absolutely right you are, but even my 5 day number lines were too negative :/
 
Mav & Others,

A question if I may.

Since ACD works well for making relative value analysis (each product verses similar), have you ever considered using a something different for the A Value? Right now the common method is using a percentage of the products either Daily ATR or I remember you saying that you use Daily AvgRng due to gaps in stocks.

As we all know daily range is better for modelling on a daily time frame than a lower time frame however they are still susceptible to distortion by company news, earnings etc. An idea I had was to look back historically and try to optimise the period we use to multiply by our A value. For example we could, say rather than use a full 24 hours of an fx pair for high and low we could just take the range during a period, where not much is going on then use a higher A value multiplier. HongKong lunchtime would probably meet the require as the market is generally dead then.

You could simplify it by saying that in essence you are looking at smoothing what you use for you ATR or AvgRng etc.

Just thought I would throw it out there before taking the time to code it myself, would be good to hear from anyone who has done something similar.

Thanks,
 
redbaron1981 said:
Mav & Others,

A question if I may.

Since ACD works well for making relative value analysis (each product verses similar), have you ever considered using a something different for the A Value? Right now the common method is using a percentage of the products either Daily ATR or I remember you saying that you use Daily AvgRng due to gaps in stocks.

As we all know daily range is better for modelling on a daily time frame than a lower time frame however they are still susceptible to distortion by company news, earnings etc. An idea I had was to look back historically and try to optimise the period we use to multiply by our A value. For example we could, say rather than use a full 24 hours of an fx pair for high and low we could just take the range during a period, where not much is going on then use a higher A value multiplier. HongKong lunchtime would probably meet the require as the market is generally dead then.

You could simplify it by saying that in essence you are looking at smoothing what you use for you ATR or AvgRng etc.

Just thought I would throw it out there before taking the time to code it myself, would be good to hear from anyone who has done something similar.

Thanks,
More...
I never liked the idea of using yesterday's range, it's just me, I'm not knocking it.

I felt I wanted to capture today's volatility in setting my A levels, so I did quite a bit of experimenting back when I started with ACD and settled for my current levels.

For FX, I use a 30 minute OR from the London open, 03:00 EST currently. For the A level I use a 12 period ATR for the 03:20 10 minute bar, in effect the 10 minute range for the last 2 hours taken in the last 10 minutes of the OR. Works fine, won't be changing it.

I set up other metrics for futures, can't remember them all off hand. For the DAX I posted about playing around with the OR, well the number line hasn't helped so I have now gone down to a 6 minute OR to see if it is better. DAX is for day trading, my weakness and magnificent obsession.

I use the same approach for my levels, weekly, monthly etc, but I just had the idea that maybe I want to tweak this for the monthly and longer. I have the data for analysis of profit taking as I wrote about at length recently, that same data will allow me to identify for example monthly levels that when exceeded 60% of the time result in a move higher substantial enough to be traded. So that's my trade metrics, I know the various stops I use in ATR for all my trades, so if I use a 0.6 ATR stop I'd want the move to be a multiple of that. The question I would ask is what level would commonly see price move further 60% or 70% of the time and would allow a profitable trade. Just tweaks in search of improvement.
 
justrading said:
I never liked the idea of using yesterday's range, it's just me, I'm not knocking it.

I felt I wanted to capture today's volatility in setting my A levels, so I did quite a bit of experimenting back when I started with ACD and settled for my current levels.

For FX, I use a 30 minute OR from the London open, 03:00 EST currently. For the A level I use a 12 period ATR for the 03:20 10 minute bar, in effect the 10 minute range for the last 2 hours taken in the last 10 minutes of the OR. Works fine, won't be changing it.

I set up other metrics for futures, can't remember them all off hand. For the DAX I posted about playing around with the OR, well the number line hasn't helped so I have now gone down to a 6 minute OR to see if it is better. DAX is for day trading, my weakness and magnificent obsession.

I use the same approach for my levels, weekly, monthly etc, but I just had the idea that maybe I want to tweak this for the monthly and longer. I have the data for analysis of profit taking as I wrote about at length recently, that same data will allow me to identify for example monthly levels that when exceeded 60% of the time result in a move higher substantial enough to be traded. So that's my trade metrics, I know the various stops I use in ATR for all my trades, so if I use a 0.6 ATR stop I'd want the move to be a multiple of that. The question I would ask is what level would commonly see price move further 60% or 70% of the time and would allow a profitable trade. Just tweaks in search of improvement.
More...
Thanks for the reply.

It seems we have quite similar ideas. Currently I have a system trading a basket of around 32 ETF's but it could also be expanded to other products if I had the time. I have around 10 setups that give me prediction of where the day will finish, accurate 59% of the time with suitable R:R(Avg 1:25) in the predicted trade direction.

Even though my accuracy is not that high I created a montecarlo simulation program to test my entries vs random entries and have been able to beat an average of 96% of the runs with a standard deviation of a little over 3%. My activation/deactivation rules are bootstrap based and so I am able to trade only when the edge is strong.

Although all the above is quite black box my trade entry is manual but using all the above to determine which direction I should be trading in and on which products. Basically I try to trade against my system in regard of finer entries and exits. It still early days but things are going well thus far and my confidence has increased 100 fold.

Just realised that I have probably detracted from my original post but its nice to exchange ideas anyhow.

Thanks.
 
Maverick74 said:
Geez, do I need to set up a self help group here for 30 day NL victims. :)
More...

Hello Mav,

30 dnl knowledge is always evolving ....... so tell us your current thinking (if you can) :)

I'm seeing more and more stories like these at various websites:

https://www.bloomberg.com/news/articles/2017-08-25/wall-street-banks-are-sending-warning-signals

https://blogs.wsj.com/moneybeat/2017/08/25/worries-grow-about-the-stock-markets-bad-breadth/

Come on man, give us some ole Maver thoughts. :)
 
Would like a word from the wise as well. Utilities and bonds have been making a lot of a-ups and c-ups for awhile. Yen seems to be strengthening. Russell have been falling all month. Spx is easier to fade rallies than selloffs. Feels like we had a mild parabolic top not too long ago. This is looking more and more like risk off type of environment. Any thoughts please? :banghead:
 
You must log in or register to reply here.
Back
Top