Trade Intensity

[UrmaBlume 10]

In most markets and certainly in the equity index futures markets there is a small set of very big traders whose entries into the market often have an immediate and very tradeable impact on price.

 

These traders go to great lengths to disquise their trade - they have automated routines to send a series of small transactions to the market at intervals close to 1000th of a second.

 

Because many of these traders operate from price based models they must get their execution within a very narrow price/time range which means they must execute as much of the volume they want as fast as they can while price is in that range. This results in a huge spike in what we call the intensity of trade.

 

To calulate the indicator shown below you must have a very expensive and almost zero latency data feed and be able to measure time in 1000ths of a second. The calculation of the indicator is merely so much volume over so little time.

 

The charts shown are in TradeStation. While TradeStation by itself doesn't have the ability to get nearly as granular with time the are dll's etc that can enable that most able platform for this work. This software is NOT for sale or lease.

 

To trade this indicator at optimal levels requires automated execution.

 

These plainly visable spikes in short term trade intensity occur between 12 and 30 times per session in the S&P.

 

Notice the time axis on the bottom of this first chart - the whole chart is a look inside 1 minute of trade.

 

 

 

 

This chart shows a similar spike in the intensity of trade

 

 

 

The same spikes happen at tops - this trade was good for 5 points in only 3 minutes - a great days trade in only 3 minutes

 

[BlowFish 10]

I wonder who's data feed you use? I've always found ZenFire to be good. I also wonder what sort of connection to the internet you have or do you run your trading application in a data centre? I'm pretty sure most of any latency I get under normal conditions is simply down to geographical location and the speed of light

 

I don't wish to quibble (but am going to anyway) but I think your 1000th of a second is probably off by an order of magnitude or two. Light would only travel about 3Km in that time for example, and the average hard disk seek time is almost 2 orders of magnitude greater than that.

 

Having said that I have no doubt that looking at volume with respect to very short time frames provides interesting data but wont a a simple 1 second chart with volume show a similar thing?

[UrmaBlume 10]

Having said that I have no doubt that looking at volume with respect to very short time frames provides interesting data but wont a a simple 1 second chart with volume show a similar thing?

 

Price is motivated by volume, not time. The granularity of an optimal data vessel is determined by units of trade and not units of time.

 

While I do believe that a 1 second bar chart would indeed reflect much of this same information, much would be lost by having any measure of time as the structure constant rather than units of trade.

 

In the emini S&P, which until very recently has traded an average of something over 2m contracts during the session, we demonstrate this data in a 25 contract bar which generates something just short of 100,000 bars per session. A 1 second chart would only generate 24,300 during the standard 405 minute day session and their rate of presentation would be constant rather than accelerating and and slowing down consistant with the rate of trade at the moment.

 

Time as a constant with regard to this kind of processing fails to report many very pertinent facts with regard to the balance and flow of trade and money in any market. A very compelling demonstration of this fact is accomplished by our Buy/Sell Volume Harmonic which I will present in a future post.

[Hlm 10]

I still don't understand the purpose of this thread since since...

 

To calulate the indicator shown below you must have a very expensive and almost zero latency data feed and be able to measure time in 1000ths of a second. The calculation of the indicator is merely so much volume over so little time.

This software is NOT for sale or lease.

To trade this indicator at optimal levels requires automated execution.

...sounds like a tooting of ones horn.

 

However...going back to what BlowFish said and you response (just for fun), what about a small tick chart value which is NOT time dependent. Also, I would like to note that you have not specifically described how this is used for trading. Without that information no one can compare how much "special" information is being extracted versus other simple methods. So again, what is the purpose of this thread?

[UrmaBlume 10]

Hlm,

 

Tick charts are NOT useful in the analysis of trade flow becasue they treat all transactions equally. In a tick chart, each transaction is treated as a single tick regardless of size and I assure you that a 1000 contract trade in the S&P has a bigger affect on price than a 1 lot.

 

As to zero or near zero latency data feed and automated execution, they are the standard for many technically astute commercial traders. Most of the work we do requires automated execution.

 

For the technically astute, the charts make it obvious that you trade the spikes.

 

The special information presented is completely described in the text as the intensity of trade by special size traders. I know of not other way to demonstrate this information and have never seen it demonstrated before, have you?

 

Most all retail trade uses indicators that have price as their only input - adaptive moving averages, exponential moving averages, RSI, Stochastic, Bollinger Bands, CCI etc. Regardless of time frame or combination of time frames these indicators have no chance of ever defining trade or leading it as they are based on price which responds to trade flow.

 

Price is motivated by trade not price itself.

 

The indicator I posted does not use price in any way and defines trade by the very secretive traders whose trade alone is enough to motivate/propel price and the non-retail traders I know find that very useful indeed.

 

Another purpose is to try and elevate the discussion beyond the triteness and futility of discussions about the utility of trade based on the consideration of multi-time frames of price based indicators.

 

Be assured that most successful commercial trade is not based on a few parameter changes on off the shelf indicators but is based on processes that have never been discussed or even mentioned on this forum.

 

The intensity of trade indicator I demonstrated here is such an indicator and I know of no forum, book or discussion that has ever mentioned it before.

 

I created this indicator and am proud of its basis and utility and don't mind sharing my work. You seem to have a big history of many very short, mostly content free posts that offer nothing new and contribute very little. Where is something new and useful that you have created that you are willing to share?

[Hlm 10]

Thank you for the extra information.

 

You seem to have a big history of many very short, mostly content free posts that offer nothing new and contribute very little. Where is something new and useful that you have created that you are willing to share?

:o You're kidding right? I guess people can have different defintions of "sharing". As for useful, that depends on the user and information supplied.

 

As a medium for sharing new ideas and things I am working on, I naturally prefer a more live atmosphere via the chat room or private message. My participation on the forums is mainly focused at answering specific questions one may have.

 

Enjoy your thread.

[atto 20]

I'll post my thoughts later on this, as it is potentially an interesting idea. Besides the motivations question (since you aren't selling), I think you have misread Hlm. He's contributed quite a lot to TL, both in the forums and especially in live chat. Let's try to keep this civil, and about your concept of trade intensity, rather than launch personal attacks. I can vouch that he knows his stuff and very frequently is helping other traders.

 

Speaking of live chat, we'd love if you could drop in during market hours (the TL chat link is on the navigation bar at the top). While we couldn't see exactly what you're talking about, as you use very expensive and fast data feeds, you could show examples of your premise and discuss with other traders (we usually have 15-16 in the morning US session).

[darthtrader3.0beta 10]

Pretty cool, I've also noticed something like this but haven't figured out a way to play with it yet.

One thing I would be interested to know is how much does your datafeed cost?

One interesting way to see this if you have ninjatrader is to pull up a time and sales window and set Timer refresh to false. Setting it to true means it updates ever 250 ms...when you set it to false though you see how many trades during those "blasts" that 250 ms is actually far to slow and actually misses tons of trades.

The fact that ninja defaults to not seeing this to me is an interesting statement on our retail tools. It just strikes me that our tools all default as if we are trading against a very very large pit and not against algorithms.

[JBWTrader 10]

Hi UrmaBlume,

Thanks for the post. You are correct. I have not seen this concept before.

 

You mention that this type of signal can occur between 12 and 30 times on the ES. Have you had the chance to calculate the win/failure ratio when taking the correct side of this signal ? in other words have you proven an edge to this concept.

 

I agree any new way to represent data visually always provides a mindset change and hopefully a new set of rules to "see" the market more profitably.

 

By the way sounds like your group is one (of the many) well organised companies that are into fully automated strategies.

 

All the best

John

[Hlm 10]

I wonder if you would get similar results by using a constant volume chart and having the histogram value defined by the amount of time needed to create a new bar. So a spike would be created when a volume comes in quicker in relation to other times.

[OAC 10]

All this reminds me of MarketDelta when it first came out about four years ago, It had this whole new way of looking at volume at the bid ,volume at the ask ,and their cumulative values arranged a new way known as the footprint. It really appealed to the logical minds of many. But in reality. how many of you today are using this amazing tool to trade?

[sevensa 10]

I wonder if you would get similar results by using a constant volume chart and having the histogram value defined by the amount of time needed to create a new bar. So a spike would be created when a volume comes in quicker in relation to other times.

 

Wouldn't volume coming in quicker result in NOT a spike, since the bar will be created quicker? So you would look at the troughs instead of the peaks in the histogram.

[atto 20]

Wouldn't volume coming in quicker result in NOT a spike, since the bar will be created quicker? So you would look at the troughs instead of the peaks in the histogram.

I think that's what he meant. Very low time bars (seen through a small "time" indication) would indicate very fast orders coming in. Something else to look in would be the derivative of that, as you'd get the change as it develops.

[Hlm 10]

Wouldn't volume coming in quicker result in NOT a spike, since the bar will be created quicker? So you would look at the troughs instead of the peaks in the histogram.

I guess you are right (not top of my game early Sunday morning)...you would just have to display the inverse. Same difference.

[darthtrader3.0beta 10]

All this reminds me of MarketDelta when it first came out about four years ago, It had this whole new way of looking at volume at the bid ,volume at the ask ,and their cumulative values arranged a new way known as the footprint. It really appealed to the logical minds of many. But in reality. how many of you today are using this amazing tool to trade?

 

I think Market Delta is to blame for this though. For one they don't have a forum, something that just completely blows my mind. There is no place to bounce ideas off other people and no place for people not using the software to get really interested like what a forum would provide.

Considering the price of the software your basically on your own to come up with how to use it and I think this is why most people try it, like it, but don't use it.

[BlowFish 10]

Interesting stuff and quite similar to some research ideas I have been playing with.

 

One interesting visualisation I showed a while back is a 1 tick constant range chart with a volume histogram below, that shows quite clear spikes.

 

Good stuff keep it coming

[sevensa 10]

Interesting stuff and quite similar to some research ideas I have been playing with.

 

One interesting visualisation I showed a while back is a 1 tick constant range chart with a volume histogram below, that shows quite clear spikes.

 

Good stuff keep it coming

 

Hi BlowFish

 

I am sorry if this sound like I am nitpicking on the posts here, but this is really not my intend. What is a 1 tick constant range chart? 1 tick chart I understand, but the "constant range" part confuses me a bit.

 

Constant range charts to me means that the bars are all the same range, but I don't understand how that will work on a 1 tick chart. Did you maybe mean just a 1 tick chart with volume histogram?

[Hlm 10]

Hi BlowFish

 

I am sorry if this sound like I am nitpicking on the posts here, but this is really not my intend. What is a 1 tick constant range chart? 1 tick chart I understand, but the "constant range" part confuses me a bit.

 

Constant range charts to me means that the bars are all the same range, but I don't understand how that will work on a 1 tick chart. Did you maybe mean just a 1 tick chart with volume histogram?

It means price has moved from 859.25 to 859.50. With a one tick chart you could have several bars created while just sitting at 859.50. With a constant range chart of one tick, volume will keep increasing until price actually moves up or down a tick hence the ability for a "spike" to be formed at the previous price/bar.

 

Added: There are two different ways to draw constant range charts. A new bar could be created at the time of the one tick move up, or it could wait until trading has continued within the range (moves beyond your specified range). I personally prefer to wait until price moves beyond my specified range which creates less bars.

Edited January 18, 2009 by Hlm
Added information

[darthtrader3.0beta 10]

also as far as trading ms..this is a really interesting podcast interview from an algo guy at Deutsche Bank on latency.

http://www.algotradingpodcast.com/MP3/ATP_LowLatency.mp3

[janus 10]

The concept of 'trade intensity' is not new. There is no truth in saying that it is not discussed/used anywhere.

 

MarketDelta has implemented IOAMT's Trade speed ideas ( I am not affiliated with any of them) in TradeSpeed Indicator.

 

http://www.marketdelta.com/kb/article.aspx?id=10585

 

http.http://www.marketvolume.com has products/pending patents based on this idea of spikes in volume. Their SBV Indicator uses the idea of trade intensity/spikes in volume.

 

I request the thread starter to elaborate and comment if his ideas are different from the above.

[BlowFish 10]

Funnily quite similar to some research ideas I have been playing with, though it seems that you work has come to fruition

 

One interesting visualisation I showed a while back is a 1 tick wide constant range chart with a volume histogram below, that shows quite clear spikes.

 

Good stuff keep it coming

[BlowFish 10]

Hi BlowFish

 

I am sorry if this sound like I am nitpicking on the posts here, but this is really not my intend. What is a 1 tick constant range chart? 1 tick chart I understand, but the "constant range" part confuses me a bit.

 

Constant range charts to me means that the bars are all the same range, but I don't understand how that will work on a 1 tick chart. Did you maybe mean just a 1 tick chart with volume histogram?

 

Hi a constant range chart that is 1 tick of the instrument in width. So for ES it would be a .25 point range for each bar. Hope that makes it clear.

 

Oh btw HLM's suggestion (constant volume bars with time as a histogram) gives an interesting take on things. Something else that I have played with in the past. Kind of weird though as things are back to front, for example a tiny time histogram bar shows fast activity.

[Head2k 10]

I would dare to say that the concept or idea is old at least hundred years. It is simply a high volume reversal, or a climax. A lot of volume in small amount of time in a small price range. I would also say that how chopped or not chopped are the orders of big guys are is not that relevant. No indicator is needed for detection of such an event. But if somebody wishes to react fast enough to participate at this very top or bottom, then some sort of computation and automization can be useful. Yet since the thread starter states that the way he does it is secret, this thread serves only as an example that something like that is maybe possible, at least with super-hyper-datafeed and processing capacity

[darthtrader3.0beta 10]

I would also say that how chopped or not chopped are the orders of big guys are is not that relevant.

 

I don't know...In the abstract it is an interesting question if a 500 lot trade moves the bid/ask as 500 1 lot trades does.

I can't totally put my finger on it but it seems a bit asburd to believe that 500 trades will act the same as 1 trade, no matter what the size.

[Head2k 10]

I don't know...In the abstract it is an interesting question if a 500 lot trade moves the bid/ask as 500 1 lot trades does.

I can't totally put my finger on it but it seems a bit asburd to believe that 500 trades will act the same as 1 trade, no matter what the size.

Market would most likely react diffenently, at least on micro-scale, but the most significant fact is that there were 500 lots traded. How the trades were splitted is IMHO not that important. But maybe if somebody wants to pick the very bottom or top it actually may be important, because picking the extreme is in fact a micro-scale work.

And last but not least, I am a beginner so what do I know?