I notice Joules likes to follow "FT71", who I watched a few webinars of his live trading, and he confirms himself that he does not trade on new highs/new lows, basically confined to previous ranges.Fractals are generated by iterating a function…or by applying a feedback loop to a system.
(Iterating a function means that you take the output from the equation and feed it back into the equation.)
A “Fractal” then, is what “emerges” from these feedback systems. We call this the “emergent property” of the system.
Yeah that's right, the volume there has a shelf life, and price seems to react to more recent areas in a trade-able manner.The problem with Market Profile is that it is the application of parametric statistics and inferences when the market is clearly a non-parametric system.
It's like believing in Black Scholes distribution of events or Efficient Market Hypothesis. Which is why CBOT and everyone else loved it in the 80s, they were so busy convincing themselves that price fits into a normal distribution.
If you have an understanding of non-linear/non-parametric statistics, chaos theory and entropy, if you accept the market is a chaotic system where certain events occur with much more frequency than if they were random then you know that Market Profile is not going to generate you consistent profits. Ever.
In non-linear systems, complex behaviour comes from simple iterations and non-linear feedback. That is all!
I notice Joules likes to follow "FT71", who I watched a few webinars of his live trading, and he confirms himself that he does not trade on new highs/new lows, basically confined to previous ranges.
Worthwhile reading from Taleb
www.fooledbyrandomness.com/GIF.pdf "The Bell Curve, That Great Intellectual Fraud"
If you can accept that the way price moves is random then all thats left for me to play with is simple direction, range and time.The problem with Market Profile is that it is the application of parametric statistics and inferences when the market is clearly a non-parametric system.
It's like believing in Black Scholes distribution of events or Efficient Market Hypothesis. Which is why CBOT and everyone else loved it in the 80s, they were so busy convincing themselves that price fits into a normal distribution.
If you have an understanding of non-linear/non-parametric statistics, chaos theory and entropy, if you accept the market is a chaotic system where certain events occur with much more frequency than if they were random then you know that Market Profile is not going to generate you consistent profits. Ever.
In non-linear systems, complex behaviour comes from simple iterations and non-linear feedback. That is all!
How come you get to play with it if I'm the one doing the accepting? :If you can accept that the way price moves is random then all thats left for me to play with is simple direction, range and time.
:venus:
Ultimately, if you try to assert that you know the capital-T Truth about something you'll almost certainly be wrong. But if you take instead Taleb's prescription and become a "skeptical empiricist," you'll always focus on finding something useful to do with the data and avoid as much as possible creating a model -- unless it turns out to be helpful. If you do create a model, you will be ready to throw it away at a moments notice by not ascribing any special qualities to it -- most especially Truth. This is why the Agilist approach of collecting data and using that data only to estimate what can be accomplished in the next iteration works well. This approach intuitively understands the limitation of the billiard-ball problem: just because you can make a pretty good approximation of what happens next, doesn't mean you have a spreadsheet formula that tells you what happens any time after that.
Can we just accept this as being adaptive to the unfolding auction process? i.e. don't be locked into a scenario just because your analysis tells you that x or y could happen. Use the current feedback from the market to form decisions...How come you get to play with it if I'm the one doing the accepting? :
But, yes. Agree.
http://www.reinventing-business.com/2011/03/limits-to-knowing-our-limitations.html
If you can show me the red highlight in realtime CanOz, I'd be a lot more convinced. For now I am still of the opinion that all the mathematical reasons MP was abandoned by quants in the 80s still tend to hold as valid compared to the reasons proponents give for using it.Can we just accept this as being adaptive to the unfolding auction process? i.e. don't be locked into a scenario just because your analysis tells you that x or y could happen. Use the current feedback from the market to form decisions...
Sorry Sinner, i just can't frame it up as well as you do in your complex terminology!
CanOz
Market Profile, as Peter Steidlmayer first developed is not actually traded by that many now as electronic trading has become the norm. It is still used however, as way to structure the market. Most of the price acceptance/rejection you see are areas of past high volume and areas of past low volume.If you can show me the red highlight in realtime CanOz, I'd be a lot more convinced. For now I am still of the opinion that all the mathematical reasons MP was abandoned by quants in the 80s still tend to hold as valid compared to the reasons proponents give for using it.
You see plenty of HFT algos written using DOM, sure, haven't seen any written using MP!
Agree with this.As Nick said to me one time "if it gives you the confidence to pull the trigger and you can generate a positive expectancy, then use whatever you want",
Fair enough.i guess time will tell for me. In the meantime this is something that makes sense to me, that i can relate to and that helps me frame the market. I'm trading off of support and resistance,as it shows up on the volume profile. My entries need to be confirmed by the action at the market, on the DOM and the volume ladder.
I disagree here, I know a couple of quants working prop and their work (similar to work done at many other shops according to them!) is based around principal component analysis and similar tools, not (very) discretionary and not following flows.You don't see to many quants trading at Prop shops either i suspect, most are all discretionary traders following the big boys around.
Nah, that's not the point I'm trying to make, just been waiting to raise the discussion about linear versus non-linear models since you've been discussing Market Profile a lot recently. I personally have been unable to successfully quantify intraday trading so use my models for sizing and bias rather than signals and triggers so I'd be the last person to say you need to quantify it. Just trying to make sure you don't get "fooled by randomness" Fact: The market is non-linear. Fact: you are using a linear model. That's all. (said the Bollinger band loverI'm not quite sure what your getting at overall, but I'm not a quant. Perhaps you can just say that "all discretionary trading does not work" as well? If you can't "quantify" it it cannot work...?
CanOz
That's cool, and i appreciate your view...but although i post the market structure, a basic plan and my trades. You still don't know how i am deciding to enter the trades...I don't believe for a minute that i am using my volume profile as market profile was first developed to be used. This is only for providing me with areas to pay attention to.I personally have been unable to successfully quantify intraday trading so use my models for sizing and bias rather than signals and triggers so I'd be the last person to say you need to quantify it. Just trying to make sure you don't get "fooled by randomness" Fact: The market is non-linear. Fact: you are using a linear model. That's all.
The thing is MP in its original form isn't even parametric stats, it's complete nonsense from a statistical perspective. Price values don't even come close to a normal distribution, or any distribution for that matter. Logaritmic close to close returns are modeled in basic quant models as a normal distribution, BSM model for example. I find market returns seem to fit a student t distribution as the closest quantifiable distribution, with usually between 3-10ish degrees of freedom. However variance, skewness and kurtosis are constantly changing, but the t distribution is a better fit for the fat tailed returns.The problem with Market Profile is that it is the application of parametric statistics and inferences when the market is clearly a non-parametric system.
It's like believing in Black Scholes distribution of events or Efficient Market Hypothesis. Which is why CBOT and everyone else loved it in the 80s, they were so busy convincing themselves that price fits into a normal distribution.
If you have an understanding of non-linear/non-parametric statistics, chaos theory and entropy, if you accept the market is a chaotic system where certain events occur with much more frequency than if they were random then you know that Market Profile is not going to generate you consistent profits. Ever.
In non-linear systems, complex behaviour comes from simple iterations and non-linear feedback. That is all!
I notice Joules likes to follow "FT71", who I watched a few webinars of his live trading, and he confirms himself that he does not trade on new highs/new lows, basically confined to previous ranges.
Worthwhile reading from Taleb
www.fooledbyrandomness.com/GIF.pdf "The Bell Curve, That Great Intellectual Fraud"
The thing is MP in its original form isn't even parametric stats, it's complete nonsense from a statistical perspective. Price values don't even come close to a normal distribution, or any distribution for that matter. Logaritmic close to close returns are modeled in basic quant models as a normal distribution, BSM model for example. I find market returns seem to fit a student t distribution as the closest quantifiable distribution, with usually between 3-10ish degrees of freedom. However variance, skewness and kurtosis are constantly changing, but the t distribution is a better fit for the fat tailed returns.
Maybe over PM or in a new thread would be better, but happy to.I'm very interested in your view on non-linear methods, do care to elaborate?
Not all machine learning models are up to the task of financial market prediction. Personally, Support Vector Machines I find to be very good for the job and this is backed up by plenty of research. I also have been doing some work using Decision Trees but the internals of how a decision is made are much more opaque (by design) so it's harder to trust psychologically to actually trade than with SVM which is very robust and easy to trust.havn't studied chaos theory yet but I have found many non-linear techniques to be less robust than simple linear methods. For example neural nets and MARS seem to fall apart far more often than simple multifactor OLS or logistic regression. Although regime switching methods such as TAR and related, I have found to be slightly better than their linear counterparts such as ARIMA at predicting smoothed/filtered signals.
AFAIK its equities. The components aren't necessarily what you'd think.Just out of interest, are your friends in the propshops using PCA trading fixed income?
+1However saying that, I like MP from the price rejection vs acceptance point of view. The volume profile makes much more sense to me personally rather than the old time/price method.
Hello and welcome to Aussie Stock Forums!
To gain full access you must register. Registration is free and takes only a few seconds to complete.
Already a member? Log in here.