Trade Intensity

[darthtrader3.0beta 10]

  deanz said:

Contracts traded are contracts traded..... the total over the same time period have to be the same no matter how you view it..... don't they ?

 

I think I'm with darthtrader, especially after visiting their website..... I smell a rat.

 

Dean.

 

well I don't really agree with this...Look at the difference between TWAP and VWAP...the fact there is a big difference is obviously quite important if your taking things to this level.

Thats why all these "VSA" threads on here to me are laughable...there should be no difference between TWAP/VWAP...else you assume things that are obviously false.

[UrmaBlume 10]

  deanz said:

Contracts traded are contracts traded..... the total over the same time period have to be the same no matter how you view it..... don't they ?

I like to give people the benefit of the doubt but I think I'm with darthtrader, especially after visiting their website..... I think I smell a rat.

 

Dean.

 

The volume histogram would of course look the same but a histogram of the intensity/rate of trade would not and this calculation requires more granularity of time than is avialable in TS.

[atto 20]

So what kind of chart would show this kind of action? A 1s chart? A 100ms chart? 10ms? Because, while TS doesn't support these types of intervals, feeds like Zenfire might come close enough.

[UrmaBlume 10]

  atto said:

So what kind of chart would show this kind of action? A 1s chart? A 100ms chart? 10ms? Because, while TS doesn't support these types of intervals, feeds like Zenfire might come close enough.

 

Atto,

 

Thank you, I appreciate the thought that goes into your posts.

 

This post, however, still misses the essence of measuring trade intensity.

 

Very short term time charts would just show a succession of trades - no spikes as most of this trade is broken down into small sizes anyway. The same would go for constant volume bars.

 

During a 1 minute bar trade intensity may spike to 100,000 contracts per minute (and does several times per day) even though only a few thousand contracts were traded during that minute. Volume in this bar did not spike, intensity did. This concept scales both up and down with either time or volume.

 

In the emini S&P trade averages about 5 - 6,000 contracts per minute during the day session. This trade is not evenly distributed throughout the day.

 

For the sake of this example - We have 2 charts - a 1 minute chart and a 5k contract bar chart. Without regard to price, trade is flowing at a steady 5k contracts per minute.

 

Volume Bars are all even, all averages of volume are flat. Now if 2k of the 5k contracts traded inside one of these bars trades at a very fast rate then there will be a huge spike in our measure of trade intensity but no spikes in any other traditional measure of volume.

 

As soon as Trade Station and other packages achieve better time granularity traders will be able to plot a host of more relevant indicators of trade flow and we know of no better indicators or motivators of price change than the flow of trade.

[atto 20]

Yeah, I'm sure I understand your concept (of intensity). However, I was simply asking what kind of interval could capture this stuff. For example, I trade on NinjaTrader with ZenFire for data, which gives very good live tick data. It seems as though this "retail" feed would work just fine, as I get the actual ticks. From there, it would be rather trivial to write a Ninja strategy that uses the tick data to gauge intensity my measuring how fast the ticks are coming in against a fast clock (such as a 100ms timer). Am I missing something?

[UrmaBlume 10]

  atto said:

Yeah, I'm sure I understand your concept (of intensity). However, I was simply asking what kind of interval could capture this stuff. For example, I trade on NinjaTrader with ZenFire for data, which gives very good live tick data. It seems as though this "retail" feed would work just fine, as I get the actual ticks. From there, it would be rather trivial to write a Ninja strategy that uses the tick data to gauge intensity my measuring how fast the ticks are coming in against a fast clock (such as a 100ms timer). Am I missing something?

 

Any interval can capture intensity - the point is that it must be measured outside the bar and imported. As shown a 1 minute or 30 second bar could very easily show normal volume yet contain a huge spike in intensity.

 

When using ticks you must remember to consider the size of the tick and not just the number of ticks.

 

Using a 100ms timer and volume flow could definetly produce a useable measure of trade intensity. We measure this intensity via 3 different algorithms.

 

cheers

[BlowFish 10]

  atto said:

Yeah, I'm sure I understand your concept (of intensity). However, I was simply asking what kind of interval could capture this stuff. For example, I trade on NinjaTrader with ZenFire for data, which gives very good live tick data. It seems as though this "retail" feed would work just fine, as I get the actual ticks. From there, it would be rather trivial to write a Ninja strategy that uses the tick data to gauge intensity my measuring how fast the ticks are coming in against a fast clock (such as a 100ms timer). Am I missing something?

 

I thought much the same. I would present the high intensity and low intensity value of the 'sample period' or bar. I guess the peak intensity of that sample period is the thing most would be interested in. Possibly how long it was sustained above a threshold would be interesting.

 

Darth I know the 'why tradestation' question wasn't addressed to me but you gotta love how quick and easy it is to prototype stuff (well I do at least).

[honvly 10]

UrmaBlume,

 

The timer resolution of Windows machines is about 16 milliseconds. The jitter of the packets in an internet connection is at least 10-20 milliseconds. How do you deal with or avoid issues coming from the millisecond frequency?

[Hlm 10]

  UrmaBlume said:

Any interval can capture intensity

I take this comment as meaning it's time frame independent and instead of dealing with micro seconds one could use such information for larger intraday moves or swings? Is this statement correct and if so, have you done any research on the larger interval "intensity spikes"? If you have, could you post some charts? Thanks.

[Pepperdog 10]

Interesting thread.

 

As it turns out, I've been working on an NT indicator in C# to track "trade intensity" in a way that may be similar to what is described in this thread, actually bid - ask volume delta, i.e. derivative of volume delta: d_Delta/dt.

 

I'd be curious to know what UrmaBlume and other's opinions are on a suitable time interval. So far I have been tracking on a per bar interval on very fast constant volume charts but I don't feel this is fine enough resolution to really see the spikes. Still a work in progress.

Edited January 26, 2009 by Pepperdog

[AgeKay 10]

Hey UrmaBlume,

 

thanks for sharing this indicator. I am not sure whether I have correctly understood how you actually calculate 'trade intensity', so can you please correct me if my interpretation is wrong on how you calculate this:

 

- You calculate the time span between to the current trade and the previous trade (that's why no latency is so important)

- You then divide the number of contracts of this trade traded by this time span. For example, 40 contracts were traded after just 0.0001 minutes (=0.06 = 60 milliseconds). So you do 40 / 0.0001 = 40.000. That would mean that the 'trade intensity' would be 40.000 contracts per minute.

 

What I am now wondering is how you you summarize this information for all trades that occur in 1 minute. Do you have one bar for each second that is the average of the 'trade intensities' in that second?

[BlowFish 10]

Well I was feeling a bit bored this morning so I knocked up something in Ninja. With the results I got I feel that you might be doing a bit of smoothing after you have processed the data? I found that the Intensity can slam around a bit.

 

Agekay my hunch is that the two histograms (in UB's charts) show peak trade intensity and lowest trade intensity for the bar - max and min values if you like. Could be wrong there. Anyway it seems that the peaks are what is interesting.

 

Here's a couple of squished up charts. The 'saw tooth' is interesting this is an artefact due to data arriving with the same delta time. I chose to add this hence the saw tooth what I really should do is reset it on a new bar. Just messing around really.

[honvly 10]

Me too, I whipped one up using AgeKay's formula and averaged the trade intensities over the bar.

 

Here are the millisecond differences of a bunch of ticks. I assume the ones with 3-7 ms differences are trades that happen one after another.

 

3.0003

0

3.0003

3.0003

49.0049

3.0003

3.0003

3.0003

3.0003

2.0002

7.0007

3.0003

3.0003

3.0003

6.0006

3.0003

4.0004

3.0003

3.0003

3.0003

5.0005

3.0003

5.0005

3.0003

6.0006

3.0003

0

48.0048

2.0002

95.0095

3.0003

4.0004

3.0003

3.0003

3.0003

3.0003

3.0003

3.0003

44.0044

128.0121

 

[UrmaBlume 10]

  AgeKay said:

Hey UrmaBlume,

 

thanks for sharing this indicator. I am not sure whether I have correctly understood how you actually calculate 'trade intensity', so can you please correct me if my interpretation is wrong on how you calculate this:

 

- You calculate the time span between to the current trade and the previous trade (that's why no latency is so important)

- You then divide the number of contracts of this trade traded by this time span. For example, 40 contracts were traded after just 0.0001 minutes (=0.06 = 60 milliseconds). So you do 40 / 0.0001 = 40.000. That would mean that the 'trade intensity' would be 40.000 contracts per minute.

 

What I am now wondering is how you you summarize this information for all trades that occur in 1 minute. Do you have one bar for each second that is the average of the 'trade intensities' in that second?

 

What we do aside - measure the time for a block of a constant size and extrapolate to cpm. We take this a step further so that our work not only reveals intensity but also the degree of trader commitment.

[AgeKay 10]

BlowFish and honvly, thanks for posting, very cool. What feed did you use in that case with NinjaTrader?

 

BlowFish, what calculation did you use for your indicator? Something similar to my interpretation?

 

honvly, what do you with the 3-7 ms differences with regards to the formula I posted?

[AgeKay 10]

  UrmaBlume said:

What we do aside - measure the time for a block of a constant size and extrapolate to cpm. We take this a step further so that our work not only reveals intensity but also the degree of trader commitment.

 

UrmaBlume, thanks for chiming in.

 

Do these program trades always chop up their orders in constant sizes? Seems rather dumb to me if they want to disguise their intentions.

 

What do you mean by trader commitment?

[deanz 10]

Here is a chart I posted earlier with one difference, the blue verticals are at 1 minute intervals, but this time I added the light coloured verticals every 1 second, anyone notice the Intensity..... :-)

 

 

185 trades in 1 second, contracts = 1342

 

now also don't dismiss the 1 second before the intense second.

 

52 trades in 1 second, contracts = 985

 

so I see 2 ways to inject volume in a very short time without it being obvious in higher timeframes, lots of smaller trades, or a few large trades. It's almost like the large spikes are a test.... of liquidly?

 

You can see a similar thing happening if you use a 1 Volume chart and plot vertical line every second.

 

I found this timeline indicator and changed it to display vertical lines every second instead of minutes like it was designed to do, unfortunately it is hugely CPU intensive now and pretty much unusable on my average home laptop. If anyone knows how to make it more CPU friendly, someone asked if there was another way of drawing a vertical line in this post:

http://www.ninjatrader-support2.com/vb/showthread.php?t=13440&highlight=vertical+line

and here is the link to the indicator the moderator said to take look at how to do it.

http://www.ninjatrader-support2.com/vb/local_links.php?catid=1&sort=N&page=1&pp=15&keyid=86

 

I want to thank UrmaBlume for helping me to see this, and I would edit an earlier post if I it was possible, my apologizes UrmaBlume.

 

Dean.

TimeLines_Seconds.zipFetching info...

Edited January 27, 2009 by deanz
added missing link

[honvly 10]

  AgeKay said:

BlowFish and honvly, thanks for posting, very cool. What feed did you use in that case with NinjaTrader?

 

BlowFish, what calculation did you use for your indicator? Something similar to my interpretation?

 

honvly, what do you with the 3-7 ms differences with regards to the formula I posted?

 

I am using Zen-Fire.

 

Right now I am just dividing the contract size by the 3-7 ms differences and averaging this value across all ticks in a bar. However, I believe that the 3-7 ms differences are artifacts of NinjaTrader's bar processing speed. To improve the accuracy, I am going to try to increase the millisecond differences by filtering out trades less than 10 contracts and splitting bid/ask volume.

[deanz 10]

honvly I would be cautious of using plain old bid/ask, as UrmaBlume has said in this post:

 

  UrmaBlume said:

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.

[darthtrader3.0beta 10]

  deanz said:

If anyone knows how to make it more CPU friendly, someone asked if there was another way of drawing a vertical line in this

 

From what I've seen to do anything usefull with drawing in ninja that is complex it seems you have to override plot(). I might be way off but I think what happens when you don't do this is ninja paints all the bar data with plot then paints over that with whatever you have in onbarupdate..I get the feeling that if you do your drawing within plot, its all painted on one pass.

I'm still struggling with figuring out how to do much when it comes to the X cordinate though.

Maybe we can get a ninja coding sub forum on here?

[AgeKay 10]

  deanz said:

185 trades in 1 second, contracts = 1342

 

now also don't dismiss the 1 second before the intense second.

 

52 trades in 1 second, contracts = 985

 

Had all of these trades the same size?

[BlowFish 10]

  AgeKay said:

BlowFish and honvly, thanks for posting, very cool. What feed did you use in that case with NinjaTrader?

 

BlowFish, what calculation did you use for your indicator? Something similar to my interpretation?

 

honvly, what do you with the 3-7 ms differences with regards to the formula I posted?

 

Ninjatrader with Zenfire. Felt like a challenge and I find coding Ninja challenging It uses the .Net frameworks DateTime construct which will returun ticks (100 nano seconds resolution). Obviously far greater than is actually available in the real world!

 

I pretty much used your interpretation AK because when I thought about it, it seems like the right way to do it.

 

I am still wondering about smoothing to which end I have tried adding average intensity per bar.

[BlowFish 10]

  honvly said:

I am using Zen-Fire.

 

Right now I am just dividing the contract size by the 3-7 ms differences and averaging this value across all ticks in a bar. However, I believe that the 3-7 ms differences are artifacts of NinjaTrader's bar processing speed. To improve the accuracy, I am going to try to increase the millisecond differences by filtering out trades less than 10 contracts and splitting bid/ask volume.

 

I dunno but I think you might have an issue I get several clear flurries with a time difference of zero and a more varied spread of values. Are you using DateTime.Now.Ticks? ..........Time[0].Ticks does not work.

 

Here's some sample output (FTSE).

ninja.docFetching info...

[AgeKay 10]

  BlowFish said:

I dunno but I think you might have an issue I get several clear flurries with a time difference of zero and a more varied spread of values. Are you using DateTime.Now.Ticks? ..........Time[0].Ticks does not work.

 

Here's some sample output (FTSE).

 

So what do you do with the time differences of zero since you can't divide by zero?

 

And what do you mean by FTSE? Is that the sample output of the UK FTSE 100 stock index?

[BlowFish 10]

  AgeKay said:

So what do you do with the time differences of zero since you can't divide by zero?

 

And what do you mean by FTSE? Is that the sample output of the UK FTSE 100 stock index?

 

Stick the salami back together again That's why you get that sawtooth effect on the chart's I posted when you get a whole bunch of activity that spans several bars but with the same time stamp.