Jump to content

Welcome to the new Traders Laboratory! Please bear with us as we finish the migration over the next few days. If you find any issues, want to leave feedback, get in touch with us, or offer suggestions please post to the Support forum here.

  • Welcome Guests

    Welcome. You are currently viewing the forum as a guest which does not give you access to all the great features at Traders Laboratory such as interacting with members, access to all forums, downloading attachments, and eligibility to win free giveaways. Registration is fast, simple and absolutely free. Create a FREE Traders Laboratory account here.

UrmaBlume

How Does This Compare to Your Approach

Recommended Posts

Our approach is based on 2 very simple principles:

 

1. Processing raw data into Information = Equity

All technical information about the condition of the market starts with the tick and those who best process those ticks will have the edge. The better the information that is the basis for the transaction, the more likely it is that the equity of the position will increase.

 

2. An imbalance in actual buying and selling volumes will precede most price moves. If we can determine and normalize the exact buy/sell pressure/imbalance then we should be able to predict the extent of the move.

 

We also feel that in most cases basic trade decision support technologies haven't changed all that much in the last decade or two, while the technologies and techniques that are available to process this data have made tremendous advances. We feel that this opportunity is further enhanced by a gap in communication between those who understand the market domain at a very high level and those who understand these technologies at a very high level.

 

Our multi-faceted approach includes, among other points, consideration of:

 

Block Trades - In most markets the bulk of the volume is done by locals or other scalpers who close their position before the session's end. While the longer term, more committed market participant is a much smaller percentage of the day's trade, this is the trade that determines the extremes and drives price. We have built software that picks out these blocks (in the S&P less that 1% of trade is for 150 contracts or more) and shows us number of blocks, total block trade and block trade imbalance at every price during the trading day in real-time. We feel there is no better determinant of support and resistance during the session than those areas that show size trade by committed traders and size imbalance by those same traders.

 

Money Flow - Using some of these algorithms and other work, we feel that we have developed formulae that accurately indicate very short term money flow in and out of the targeted market.

 

Time/Price/Volume Continuum - We feel that time is much too important a component of the time/price/volume continuum to be treated as a constant so we use no time based charts. All of our charts are based on actual volume traded which helps us to better calculate and locate imbalances. It is constantly amazing to me how much more clear a properly adjusted volume chart is when compared to a time chart covering the same time period as the volume chart.

 

While we are finding that the information described above is adequate decision support for profitable intra-session trading, we also use pre-processed and normalized data from these methods as inputs to:

 

1. Our own Time Series Analysis tools - Time Series Analysis is a branch of math sometimes referred to as spectrum analysis. Algorithms from this area of study are much used in missile and anti-missile tracking and interception. While missiles can move in more than one dimension, it is fortunate, for us, that price only moves in one dimension (up or down) and may be one of the most trackable/predictable of the components of the time/price/volume continuum.

 

Speaking technically, what the software we have under development does is "transforms the one-dimensional time series to the trajectory matrix by means of a delay procedure, performs Singular Value Decomposition of the trajectory matrix so that it can reconstruct the original time series."

 

Translated - we input a sequence of volume and price based inputs that make a "signal" that "looks" like price, our software breaks this signal down into several digital components, reproduces the signal without the delay and then, specifically, in our case, produces its best calculation of the mid point of the range for the next six bars.

 

While most all of the mathematical components of such an approach are in the public domain, our approach, including input pre-processing, number of selected eigenvectors and target formulation are not.

 

2. Neural Networks - we use our own genetic routines to optimize the parameters and architecture of these networks, often use cascades of these networks and have the expertise with this technology to avoid overfitting.

 

3. Models developed in MARS (Multivariate Adaptive Regression Splines) from Salford Systems and other regression analysis tools are very easy for us to deploy as we have built tools to transform the finished functions into Trade Station's easy language so that a finished model is up and running online as quick as we can cut and paste.

 

We determine whether to use linear (regression) or non-linear (neural networks) in any given situation by analyzing the results of runs made on test data that was not used in the model's development.

 

We feel that it is a poor state of affairs that traders today still use charts that depict bars, candles and profiles when there is technology that can do a much better job of consolidating the information described above into a more precise trade decision support screen.

 

We export tick data from Trade Station to a PostgreSQL db that drives the screen that is attached. This screen presents constantly upgraded:

 

Projections of session High, Low & Close

Total Volume, Block volume and Block and total volume imbalance at every price

Percentage of normal volume for several time frames normalized as to time of day

Volume Velocity in a gauge that shows precise contracts/share per minute

Volume imbalance in at least 4 time frame via dynamic color pie charts

Projections of the mid point of the range for the next six bars.

 

We are a private company and offer no product or service to anybody, just trying to get this discussion off the ground.

 

mhud.jpg

Share this post


Link to post
Share on other sites

Very interesting. So, how do you trade off this? How much better are your projections than random?

 

I did have a few questions, but I'll start with two.

Using some of these algorithms and other work, we feel that we have developed formulae that accurately indicate very short term money flow in and out of the targeted market.
Since all trades are 1 for 1 (a buy and a sell), what are you calculating?
..transforms the one-dimensional time series to the trajectory matrix by means of a delay procedure, performs Singular Value Decomposition of the trajectory matrix so that it can reconstruct the original time series..
I understand the technical lingo, but don't really understand what you're trying to do. Price isn't like a missle, so you can't calculate a trajectory. Could you elaborate?

Share this post


Link to post
Share on other sites

We calculate cumulative imbalance over volume frames not time frames and we use an algorithm similar to but not the same as that used by most data venders to differentiate up volume from down volume.

 

Tracking price via time series (SSA) analysis is very much like tracking a smart missile that can change course in multi-dimensions except that it is easier becasue price only moves in one dimension - up and down.

 

This time series analysis is much like signal or wave analysis in that a center/least squares fit line is drawn and the amplitude and frequency of the departures from the line are digitized and then a very small sample is projected from a very large sample of sequential inputs.

 

Pretty standard signal, time series and wave analysis.

Share this post


Link to post
Share on other sites

I'm curious where you see this discussion going...

 

I assume this is your site - http://www.tradepointtechnologies.com

 

Forgive the skepticism but if you are not here to sell a product, what exactly is the intent? The software looks intriguing but since you are not selling it and I assume will not disclose the inner workings of it, exactly where do you see a possible discussion going...

Share this post


Link to post
Share on other sites
Guest forsearch
Price isn't like a missle, so you can't calculate a trajectory. Could you elaborate?

 

Actually, since you have three vectors - Price, Time and Volume - you certainly can!

 

Take a look a some of Gann's writings (google it) and you'll see how he was able to accurately predict the future moves of a given security, using only 2 of the 3 vectors, IMHO.

 

I still haven't fully grasped how the Square of Nine and the Gann Wheel really works, but I suppose that I needn't know the mechanics of a system to make money, do I?

 

-fs

Share this post


Link to post
Share on other sites
Guest forsearch
I'm curious where you see this discussion going...

 

 

Maybe he and RussellDayTrader ought to hookup offline and leave us all out of it, since they seem to both be on the same wavelength....

Share this post


Link to post
Share on other sites

Interesting post. May I ask how you attempt to detect buy and sell pressure and order flow? I am guessing some sort of volume analysis or maybe looking at volume@bid/ask?

Share this post


Link to post
Share on other sites

Most Data Vendors designate trade on the ask as UpVolume and trade on the bid as DnVolume. We have found flaws in that approach and use something close but, still different.

 

We calculate buy and sell imbalances over several volume/time frames and then further pre-process this data into inputs to both linear and non-linear analysis.

Share this post


Link to post
Share on other sites
Most Data Vendors designate trade on the ask as UpVolume and trade on the bid as DnVolume. We have found flaws in that approach and use something close but, still different.

 

We calculate buy and sell imbalances over several volume/time frames and then further pre-process this data into inputs to both linear and non-linear analysis.

 

 

I have to ask...who is "we"?

Are you representing the company listed on that image posted?

It appears that's the case, and if so I'll have to ask you to stop advertising services here. I'll await your reply to decide what action to take.

 

Thanks,

MC

Share this post


Link to post
Share on other sites

I would agree that it is flawed. In fact at turning points the reverse tends to be true it seems to me. Do you do any sort of order book analysis? I have not, but have a hunch that looking at pulled orders as well as filled orders may be useful in determining short term direction.

 

Do you use volume over fixed time frames or over variable periods determined by market structure?

Share this post


Link to post
Share on other sites

I offer no product or service to anybody and neither does Trade Point Technologies. It is not a company, we are a small private association of of tech and market afficianados.

 

We advertise nothing and offer nothing to anybody but an elevation of the otherwise bleak technical discussion on this forum.

Share this post


Link to post
Share on other sites

I'm interested in your approach. How accurate are your "projections"? What kind of trading do you do with them (how big of moves are you trading)?

Share this post


Link to post
Share on other sites

I also would like to chime in and request that the company name not be mark anymore in this post.That being said I think that a post like this does have importance. The stuff they are doing is getting into quant or financial engineering and that's something we don't get to many discussions about here on TL so maybe this will bring more of that. Although I have just started diving deep into this type of research I believe this type of cross over between tech/computer capabilities and markets is IMO the future of trading.

Edited by stanlyd

Share this post


Link to post
Share on other sites

In reguards to your PM to me...I won't retract any accusation of "spam". The reality is you (or WE as it reads) have a site to promote, there is a "services" tab on the site. It's not populated yet, but it's there. What's going to be offered as a service in the near future? There wouldn't be a services tab there if it wasn't going to be used now would it?

 

Could you share background on your user name? "Urma blew me" is how I would read that out. That dosen't seem very legit or credible, maybe I'm reading that wrong though so help me out here. That's what caught my eye and made me look into your posts to be honest.

 

I also like the jab at the posters here and our "bleak technical discussions", real classy and a sure way to make friends. :\

Share this post


Link to post
Share on other sites

Atto,

 

We primarily trade the equity futures and find betweent 30-45 round trips in the emini S&P per 405 minute day session.

 

Browns,

 

Thanks for the welcome. With all of the intelligent processing tools availabe today I find it astounding that traders still look and bars, candles and profiles.

 

My OP defines the tools we use and my only purpose here was to try and elevate a really dull, not state of the art technical level above VSA analysis.

 

MC & Stanlyd,

 

When I say company I really mean an association of tech and market freaks. Our income comes from our own trading and we don't offer any product or service to anybody.

 

Stanlyd,

 

You are completely correct in that better, smarter and more intense processing is the future of all technical trading.

 

All market technical data starts with the tick and the trader who has the benefit of the best and smartest processing of these and related ticks wins.

 

Much of the public thinks that averages, relative strengths, stochastics, profiles, candles, VSA and spread analysis is the state of the art and it isn't even close.

 

I am not selling anything here and am just trying to demonstrate what the vast majority of those who read this forum would never see otherwise. I resent the spam accusations.

Share this post


Link to post
Share on other sites

No problem MC if you don't consider the technical discussion on this forum bleak then that speaks to your experience and understanding. Obviously a discussion of the tools and techniques I use is not welcome by management so --

 

cheers

Share this post


Link to post
Share on other sites
I am not selling anything here and am just trying to demonstrate what the vast majority of those who read this forum would never see otherwise. I resent the spam accusations.

 

You used the term spam, I did not. If I felt it was full blown spam it would have been deleted and you banned on the spot. I am interested in the approach you are suggesting and would like you to elaborate. So far you've been pretty tight lipped and given one liners and bait to bring the fish out.

 

If you provide more info and value here, I will gladly retract my questions in relation to your motive. ;)

Share this post


Link to post
Share on other sites

Hi UrmaBlume

 

I would like to hear more...interesting...

 

How do you find the TS data export...? have you experienced delays in data and hence execution of your model..? curious here...

 

Also Its interesting you have found a way that better approximates the "Up/Dn volume" at bid/ask...very interesting...how did you do that..?

The reason I ask is it looks like I may be using an inferior approximation to up/dn volume picture...could you explain this a little more please...

 

Much appreciated and look forward to the elevation in my trading !

 

All the Best

 

John

Share this post


Link to post
Share on other sites
Our approach is based on 2 very simple principles:

 

1. Processing raw data into Information = Equity

All technical information about the condition of the market starts with the tick and those who best process those ticks will have the edge. The better the information that is the basis for the transaction, the more likely it is that the equity of the position will increase.

 

2. An imbalance in actual buying and selling volumes will precede most price moves. If we can determine and normalize the exact buy/sell pressure/imbalance then we should be able to predict the extent of the move.

 

We also feel that in most cases basic trade decision support technologies haven't changed all that much in the last decade or two, while the technologies and techniques that are available to process this data have made tremendous advances. We feel that this opportunity is further enhanced by a gap in communication between those who understand the market domain at a very high level and those who understand these technologies at a very high level.

 

Our multi-faceted approach includes, among other points, consideration of:

 

Block Trades - In most markets the bulk of the volume is done by locals or other scalpers who close their position before the session's end. While the longer term, more committed market participant is a much smaller percentage of the day's trade, this is the trade that determines the extremes and drives price. We have built software that picks out these blocks (in the S&P less that 1% of trade is for 150 contracts or more) and shows us number of blocks, total block trade and block trade imbalance at every price during the trading day in real-time. We feel there is no better determinant of support and resistance during the session than those areas that show size trade by committed traders and size imbalance by those same traders.

 

Money Flow - Using some of these algorithms and other work, we feel that we have developed formulae that accurately indicate very short term money flow in and out of the targeted market.

 

Time/Price/Volume Continuum - We feel that time is much too important a component of the time/price/volume continuum to be treated as a constant so we use no time based charts. All of our charts are based on actual volume traded which helps us to better calculate and locate imbalances. It is constantly amazing to me how much more clear a properly adjusted volume chart is when compared to a time chart covering the same time period as the volume chart.

 

While we are finding that the information described above is adequate decision support for profitable intra-session trading, we also use pre-processed and normalized data from these methods as inputs to:

 

1. Our own Time Series Analysis tools - Time Series Analysis is a branch of math sometimes referred to as spectrum analysis. Algorithms from this area of study are much used in missile and anti-missile tracking and interception. While missiles can move in more than one dimension, it is fortunate, for us, that price only moves in one dimension (up or down) and may be one of the most trackable/predictable of the components of the time/price/volume continuum.

 

Speaking technically, what the software we have under development does is "transforms the one-dimensional time series to the trajectory matrix by means of a delay procedure, performs Singular Value Decomposition of the trajectory matrix so that it can reconstruct the original time series."

 

Translated - we input a sequence of volume and price based inputs that make a "signal" that "looks" like price, our software breaks this signal down into several digital components, reproduces the signal without the delay and then, specifically, in our case, produces its best calculation of the mid point of the range for the next six bars.

 

While most all of the mathematical components of such an approach are in the public domain, our approach, including input pre-processing, number of selected eigenvectors and target formulation are not.

 

2. Neural Networks - we use our own genetic routines to optimize the parameters and architecture of these networks, often use cascades of these networks and have the expertise with this technology to avoid overfitting.

 

3. Models developed in MARS (Multivariate Adaptive Regression Splines) from Salford Systems and other regression analysis tools are very easy for us to deploy as we have built tools to transform the finished functions into Trade Station's easy language so that a finished model is up and running online as quick as we can cut and paste.

 

We determine whether to use linear (regression) or non-linear (neural networks) in any given situation by analyzing the results of runs made on test data that was not used in the model's development.

 

We feel that it is a poor state of affairs that traders today still use charts that depict bars, candles and profiles when there is technology that can do a much better job of consolidating the information described above into a more precise trade decision support screen.

 

We export tick data from Trade Station to a PostgreSQL db that drives the screen that is attached. This screen presents constantly upgraded:

 

Projections of session High, Low & Close

Total Volume, Block volume and Block and total volume imbalance at every price

Percentage of normal volume for several time frames normalized as to time of day

Volume Velocity in a gauge that shows precise contracts/share per minute

Volume imbalance in at least 4 time frame via dynamic color pie charts

Projections of the mid point of the range for the next six bars.

 

We are a private company and offer no product or service to anybody, just trying to get this discussion off the ground.

 

mhud.jpg

 

 

MC if all you got from the above post is "one liners and bait" I suggest you re-read to see if you can find the discussion of Information = Equity, Volume imbalance, applicability of block trades and block trade imbalance to determine support and resistance, linear and non-linear processing, price/time/volume continuum, not to mention two descriptions of the process of time-series analysis and more. And that was just to get the conversation going. Oh yes, and a look at an application that less than 10 people in the world have seen. If we ever did decide to bring anything to the market we would certainly stop posting about it. We have no such plans.

 

You said "one liners and bait" and if that's all you got, I can certainly understand that too.

 

Like I said - No wonder that the technical discussion heretofore has been so bleak - could it be the seemingly high nit ratio?

Share this post


Link to post
Share on other sites

urma - I'm still wondering how there can be a meaningful discussion when the inner workings of the software is proprietary.

 

As I said via PM to you, the software does indeed look interesting but you are asking us to hold a conversation w/ you when we've seen one screenshot and have no idea how the output is constructed. It's similar to me providing a screenshot of my trade blotter of one trade in one moment in time and asking everyone here to talk about the trade... Would be rather difficult.

Share this post


Link to post
Share on other sites

Can we all try to keep this civil? I don't think he's here to sell anything, and his ideas aren't presented here often. Let's try to make this productive, or keep quiet. :thumbs up:

 

Assuming you want to talk about this further... so, how effective are your projections? By the figure you gave, I'd assume you're scalping. What kind of risk/reward do you use, and what's your win %?

 

Would you be willing to discuss some of the inner workings here? For instance, I'm intrigued by your different way to measure up/down volume. Could you elaborate?

Share this post


Link to post
Share on other sites

First of all, glad to have some more fellow quants on the forum. :)

 

However...I am a little confused as to which way this discussion should go. Are you planning to go into more details about each topic? Or are you only wanting a general discussion about the use of these quant topics? If we ask specific questions are they going to be answered with specific answers? Maybe we could create a new thread for each of the three items you have listed and dive deeper into each one. This may be the best way to keep things clean and on topic. I think the main problem is that no one really knows exactly which way you are meaning/wanting to take this discussion.

 

Again, welcome.

Share this post


Link to post
Share on other sites

 

We determine whether to use linear (regression) or non-linear (neural networks) in any given situation by analyzing the results of runs made on test data that was not used in the model's development.

 

mhud.jpg

 

In the above quote, when you mention "runs" are you speaking of wave cycles?

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.