What do you consider noise?

Quote from oldtime:

I don't know that much about Elliot, yeah I browsed through the book. The name has been coming up more and more often on this site. I think it is all fashion, things come and go, and if you stick with it long enough you will be hip, then and old fogie, and finally hip retro. That's why I never throw away an old tie. Sooner or later they will be back in fashion.

Correct me if I'm wrong, but they way I read your last post is, individual transactions are noise, at least in liquid markets, but mass psychology is the move. Sounds good until you try to figure out where individual transactions have crossed the line and become a mass move.

Some of us try to make a living off noise, since it requires really no prediction which takes any skill other than the common sense your mother hoped you were born with. But paying the bills is one thing, to make up for all the mistakes I make along the way I need a big move, or as you call it, "Mass Psychology." I don't know how Ellliot figures it out, for me it's just wake up one morning and the realisation, "Wow, I don't think this is just noise! We might be on to something here. And it's about time, because this noise has been chopping me to death." (and that's what noise will do to you, it will make you money a lot longer than the trend traders can hold on, but in the last days it will take away everything it gave to you and then some. Hopefully by then you are still solvent and can ride the noise free trend.)

There are enough materials out there on Elliott that if you found yourself interested, there is plenty to read. As I became more immersed in Elliott, I realized that he had made some crucial errors, rendering his specific approach pretty useless, but, as a way of thinking about the market, his approach turned out quite useful to me, specifically the ideas of "mass psychology" and not needing anything outside of price data.

It's not so much that individual transactions are "noise", it's the reasons for them that are noise. When price goes up and triggers a trade for me, the specific transaction that enabled price to hit my trigger may be from a hedge fund, a guy putting down his last $400 on a day trade or an insurance company trying to hedge variable annuity exposure. It doesn't matter. All that matters is that price hit where my stop limit was parked and now I'm in a trade.

The science of it is exactly that, to figure out when individual transactions become a trend. For that, you need to understand the logic of the market's fluctuations, or "waves". I have devised a set of rules that give me better than even odds of identifying that point.

Interestingly, since you mention noise chopping you up, the rules are so sensitive to market action that rather that put me into bad trades, it appears that the rules actually simply stop giving trade triggers when the environment isn't conducive to profitable trading. I've spent most of the past 4 months on the sidelines, but when I do get in, the trades are still profitable.
 
Thanks for all the very interesting replies. I was actually wondering about the subject from the perspective of a beginner, and was seeking more practical advice.
For example, depending on the kind of trading you do (intra-day, swing) or what you trade (stocks, forex) are there time frames you ignore? Also, although I am not suprised that indicators look different in different timeframes does that matter. Can that really be used? I might see a setup I know in one time frame. Lets say a break above the upper bollinger when the macd is above the signal and zero lines in a 15 minute chart. Here I might consider a buy. But a 5 minute or hourly chart might look completely different. Should I care? How far on either side of your normal chart timeframe do you bother to look. And do you consider extremely fast charts to be less useful because of increased volitility of the data ie noise.
Thanks again. Interesting replies though.
 
Quote from oldtime:

Yeah, once a butterfly in Japan flapped it's wings and nobody even paid any attention to it. Then a few years later there was a horrible typhoon in the Indian Ocean. They tried to figure out what caused it and sure enough, they traced it back to the butterfly in Japan. So then they set up a whole array of microphones to record the flapping of butterfly wings in Japan. But when they listened to the recording, all they heard was noise.
The Butterfly Effect is a complete crock. Invented by some mathematician with no sense of the physical world. Show me the butterfly that can affect the weather by flapping its wings and I'll show you Mothra. :D
 
Quote from braincell:

Take a look at G-Bot, seriously.
I did look, seriously. Your link doesn't illustrate any profits extracted from a random series.
 
Noise to me is an HFT algo quote stuffing just to slow the feed down. Pres obama or Bernanke speaking is noise, cnbc and news is noise. Most of ET is noise. Most of twitter is noise. When my little voice in my head is talking to myself when I'm trading , it's noise :)
 
Quote from mcteague:

Thanks for all the very interesting replies. I was actually wondering about the subject from the perspective of a beginner, and was seeking more practical advice.
For example, depending on the kind of trading you do (intra-day, swing) or what you trade (stocks, forex) are there time frames you ignore? Also, although I am not suprised that indicators look different in different timeframes does that matter. Can that really be used? I might see a setup I know in one time frame. Lets say a break above the upper bollinger when the macd is above the signal and zero lines in a 15 minute chart. Here I might consider a buy. But a 5 minute or hourly chart might look completely different. Should I care? How far on either side of your normal chart timeframe do you bother to look. And do you consider extremely fast charts to be less useful because of increased volitility of the data ie noise.
Thanks again. Interesting replies though.
start from the wide angle and work in,for the simple reason that the big money moves the market and because of their size it's a gradual signal,they have the dough,they have the largest positions,they have the most risk,they would be the in most vulnerable spot,it would be in their spirit of survival to hide their vulnerabilty,this would involve,bernake,cnbc,ecb,.etc..your timeframe perspective needs to encompass those bosses,the big picture with a lot of small pics showing you the breakdown,if you understand the big picture ,you will understand the millions of brush strokes that are necessary to complete the picture,start from the top and work down,one way to understand this is market profile,learn the intraday flow and extrapolate it to the multi year timeframe..i dont know elliott w but i think if you applied ir the same way,the layups would present themselves,all new traders think the business is about making a killing,you learn after awhile(several small accounts) it's about not getting killed
 
Noise is any price action that is not tradable. Where a tradable pattern isnt identifiable. Obviously, what is considered noise varies greatly between traders.

As far as Elliot wave and other technical analysis beyond simple support and resistance/chart patterns is waste of time in my opinion. Elliot Wave is beautiful in hindsight but broken in real time. Its something nice and pretty that is easy to market and attractive to new traders. They are all just functions of price movement anyway. Its more important to understand orderflow and price action. Knowing the news and fundamentals. MACD. Stochastics, RSI, CCI, Elliot Wave all bs in todays market.

Watch the level 2, read the tape, know the news and key support/resistance. Trading is incredibly difficult so don't kid yourself that a short cut lies in technical analysis. It takes thousands of hours in front of the market. Not reading books or backtesting endless combinations of technical analysis indicators.

Just my opinion. Take it with a grain of salt.
 
Quote from kut2k2:

I did look, seriously. Your link doesn't illustrate any profits extracted from a random series.

Ok well, if you downloaded and used it (you need permission), you would see that you can analyze a strategy on a random price series. In fact it's the only way to really analyze strategies in that program. It will create random prices and give you series of analysis of how the algo would fare. There are many that give very nice PnL, sharpe ratios, etc, but drawdown can be a problem. I've checked the data series on which it works and they really are randomly created. It gives you an option of doing N iterations to get distributions of PnL and other descriptive statistics. Like i said, some of the settings will make good PnL distributions and indicate they are robust in extracting profit from random prices. If you don't believe me I can get screenshots or a video or something. There are also analogous whitepapers written (mathematical) that describe that same possibility.
 
Quote from braincell:

Ok well, if you downloaded and used it (you need permission), you would see that you can analyze a strategy on a random price series. In fact it's the only way to really analyze strategies in that program. It will create random prices and give you series of analysis of how the algo would fare. There are many that give very nice PnL, sharpe ratios, etc, but drawdown can be a problem. I've checked the data series on which it works and they really are randomly created. It gives you an option of doing N iterations to get distributions of PnL and other descriptive statistics. Like i said, some of the settings will make good PnL distributions and indicate they are robust in extracting profit from random prices. If you don't believe me I can get screenshots or a video or something. There are also analogous whitepapers written (mathematical) that describe that same possibility.
this might be a little off topic. I'm sure the random programs you use today are much better than the ones we used in the past, but the whole idea of testing a system on random data was to find a strategy that would break even minus the commissions. If it made money, or lost money it was disqualified as being an acceptable strategy to build on.

The ones that consistently broke even could then be employed not as a trading strategy, but as a sound money management system. You could trade them with confidence that no matter the drawdown they would get you back to even. And this is where we probably part company, but my idea was always (and still is) to tweak them in your favor with discretionary trading usually based on some kind of fundamental bias.

sorry if I went off topic, back to noise
 
Quote from ammo:

noise is what the parent's of a successful musician have to listen to while the kid's growing up,if he/she makes it,it becomes music

No offense, but this is an oversimplified view that ends-up wrong. You are confusing signal lack-of quality with noise. I would bet you are not a musician.

Wikipedia has a long article on noise, of which the 1st sentence is a good start for the purpose of this thread: "In physics and analog electronics, noise is a mostly unwanted random addition to a signal" - I would argue that the word "random" isn't a requirement to define noise, and I will use this moving foward: "noise is an unwanted addition to a signal".

Which should make it clear that noise is always relative to a signal, and cannot be defined in the absence of a definition for the signal itself.

When it comes to trading, I contend that the tape contains a variety of "signals" mixed together ... the "signals" are the footprints of many different type of players & trading strategies - I'll refer to these signals as "signal-1, ... signal -2, .. signal-n":

tape = signal-1 + signal-2 + signal-3 + ... + signal-n.


Let's use a single strategy viewpoint, and say that it is focusing on signal-1 to derive its actions. From that single strategy viewpoint, we can write:

tape = signal-1 + noise-1
w/ noise-1 = signal-2 + signal-3 + ... + signal-n.


If we were to look at another strategy focused on signal-2, we would have:

tape = signal-2 + noise-2
w/ noise-2 = signal-1 + signal-3 + ... + signal-n.


If you insist on a music analogy, then play simultaneously the 9 Beethoven's symphonies, and try to focus listening to symphony #6 for example (which happens to be my favorite of those 9) ... this will be impossible for most people on earth, and I bet would drive (literally) crazy a good number of musicians. Trading is no different.


Go back to Brass's post ... He summarized it all in 2 sentences:

Quote from Brass:

Anything that annoys me I consider to be noise. Of course, one man's noise can be another man's music.
 
Back
Top