The ACD Method

[sleepy]

Mav,

Based on your years of experience using ACD, are number lines the best determinant of whether a market will follow-through or not? And does this also apply when trading ACD intra-day (i.e., the likelihood there will be follow-through from the OR to the ATR). Or do you suggest using something more objective like the ADX?

 

[Maverick74]

Mav,

Based on your years of experience using ACD, are number lines the best determinant of whether a market will follow-through or not? And does this also apply when trading ACD intra-day (i.e., the likelihood there will be follow-through from the OR to the ATR). Or do you suggest using something more objective like the ADX?

The number lines do an excellent job of describing the current market environment for that product. For daytrading they are best used for intra-day bias. So if you have a +20 NL I would not be looking to take short trades. Only longs.

 

[MoneyMatthew]

I said it on Monday and look where Live Cattle is at now, a case of Bad News/Good Action....

 

[redbaron1981]

The number lines do an excellent job of describing the current market environment for that product. For daytrading they are best used for intra-day bias. So if you have a +20 NL I would not be looking to take short trades. Only longs.

Hi Mav and other ACDers,

Have any of you guys run any figures on the 30 day nl score versus profit/loss?

Something like filtering nl30 > 20 then totalling up the returns from being long on those day vs returns for being short? Maybe even filtering in brackets nl30 > 10 & nl30 < 20 then totalling up returns from longs vs shorts?

It could be quite interesting to see an might provide some clearer insights about the nl30 vs return structure. Do you think that this would be of any value MAV, perhaps more from a qualitative point of view vs quantitive?

Thanks,

 

[Maverick74]

Hi Mav and other ACDers,

Have any of you guys run any figures on the 30 day nl score versus profit/loss?

Something like filtering nl30 > 20 then totalling up the returns from being long on those day vs returns for being short? Maybe even filtering in brackets nl30 > 10 & nl30 < 20 then totalling up returns from longs vs shorts?

It could be quite interesting to see an might provide some clearer insights about the nl30 vs return structure. Do you think that this would be of any value MAV, perhaps more from a qualitative point of view vs quantitive?

Thanks,

Don't let me stop you from that! I'm always interested in these experiments. I know one for certain from my anecdotal experience, my largest losses have always happened when sitting in a position with exposure against a confirmed 30NL. In every single instance I was fighting a confirmation. And I always seemed to remember at the time that the 30NL seemed to confirm so late in the move and I would say to myself, "this confirmation is late, the market is going to reverse". LOL. Or the market will come back off the confirm. It's just a bad print. Never ended well.

 

[redbaron1981]

Ok I will run this for you on my db of 32 etfs around 5 years of data. I like testing over a universe of data since I am therefore not seeing a character of an individual instrument but more of the general market.

The late confirmation is something I have already tested for. I built a system then plotted returns vs time of confirmation from open. It was as clear as night and day that the later the confirmation the lower the return for the day.

 

[justrading]

Ok I will run this for you on my db of 32 etfs around 5 years of data. I like testing over a universe of data since I am therefore not seeing a character of an individual instrument but more of the general market.

The late confirmation is something I have already tested for. I built a system then plotted returns vs time of confirmation from open. It was as clear as night and day that the later the confirmation the lower the return for the day.

I think Mav is referring to how many days into the move before the 30D NL confirms, not time of day the A confirms.

 

[baggerlord]

I have found some meaningful correlations but you need to filter for certain situations or create NL derivatives. I found 0 correlation between 5 day or 30 day NL and various future returns in general. I'm away from the computer for a while so don't have my analysis handy but I would look into momentum of the NLs, and also time of month type concepts.

 

[justrading]

ACD Meets Market Reality.

Here is a simple exercise to understand this point, particularly if you day trade.

First, let's agree that the 30D NL can stay confirmed for weeks, because it does this all the time when you have a sustained move.

Now, if you have several years of daily data, count the number of instances of 2 consecutive higher daily closes, 3, 4, 5, 6. Do the same for lower. You don't need programming skills for this, I did it in Excel, but be warned many ongoing studies of this nature gobble up RAM, I have 48GB on my workstation, learn programming if you are young enough, LOL.

You will see that while 2 consecutive is common, it starts to fall off sharply by the time you get to 4, 5 or 6. I'm away from the workstation so I can't give you any numbers, just from memory I'm telling you 5 or 6 consecutive are pretty uncommon.

If you are a day trader and you are looking at the 30D and saying it's confirmed positive, I'm going long today, good luck. Try moving averages or MACD crossovers, less work than ACD and you'd probably get similar results when applied mechanically.

You have done the work, you have your number lines, how to use them is a fine art, it is not mechanical. Derivatives that tell you the nuance are where it's at, and it's no secret, Mav has said it before, as have I.

A simple example. The 30D is confirmed, the 5D is weakening rapidly. What time frame are you trading? If longer term, just check for a pullback in progress, where likely support is, and decide what you will do with your stop.

If day trading or just trying to catch short term momentum, I'd be hesitant to go long in that environment. Waiting for the ducks to line up usually gives best results.

 

[redbaron1981]

it is not mechanical. Derivatives that tell you the nuance are where it's at, and it's no secret, Mav has said it before, as have I.

I have heard it stated multiple times on this thread that ACD is not mechanical, however that is limited only to your own current perception/experience of ACD.

From my own experiences and most probably me being a control freak, I really wanted to apply ACD from a more qualitative standpoint, also to have vast statistical evidence to backup that when a = b there is an increased likelihood of c happening etc.

Basically I wanted to have as much confidence that ACD works before committing any capital to it. For me that meant forming hypothesis, rigorous testing and calculating probabilities.

During my testing I woke up during the night with an idea of how I might be able to use ACD in a mechanical framework. Luckily enough I keep a notepad by my bedside to write down any ideas like this that pop up through the night. After a few very late nights/long days coding I finally had robust rules that can be applied going forward using ACD on multiple markets. Yes I understand markets change over time but there are certain aspects of ACD that remain robust through out time no matter what the conditions.

If I had limited my own studies by testing only topics discussed on this thread, I would never have made the discoveries that I made. In short ACD really is only limited by your own creativity.