Equi power separation for real time intraday trading

[acepsut]

Hi all,

I am studing the "Value in time" book and I found it one of the best book I have ever read.

The concepts of EV Flow and the 3 large and small players separation methos are all clear to me, but there is something I am missing so I am asking if someone can help me to understand how to use the Equipower separation method for intra day trading.

At pages 51 to 57 Pascal explains 3 different methods, while only the last one is pointed out as the best one.

Based on yesterday's effective volume, he sort the bars from left to right starting from the tallest to the lowest, then took as large volume first 50% volume starting from the left.

That's fine, but this is valid for yesterday data, and only after market close: what about if we want to trade today?

Any day total traded volume and shares differs, so we can't rely on yesterday's LP value since today may change.

So, my questions are:

1) In order to trade stocks today as market open (@ 9 AM), how should I compute LV from SV Equi power separation?

2) How many days back should I use for separation?

3) Does the sep calculation change if I do trade on various time frame, i.e. 5 min, 15 min ,60 min or daily?

4) I have read in this forum a nice suggestion about to consider above 1 standard deviation as large and below as small, does anyone use this method? and does it performs better or worst compared to equi power separation?

Thanks for any help and suggestions.

acepsut

[Pascal]

Hi all,

I am studing the "Value in time" book and I found it one of the best book I have ever read.

The concepts of EV Flow and the 3 large and small players separation methos are all clear to me, but there is something I am missing so I am asking if someone can help me to understand how to use the Equipower separation method for intra day trading.

At pages 51 to 57 Pascal explains 3 different methods, while only the last one is pointed out as the best one.

Based on yesterday's effective volume, he sort the bars from left to right starting from the tallest to the lowest, then took as large volume first 50% volume starting from the left.

That's fine, but this is valid for yesterday data, and only after market close: what about if we want to trade today?

Any day total traded volume and shares differs, so we can't rely on yesterday's LP value since today may change.

So, my questions are:

1) In order to trade stocks today as market open (@ 9 AM), how should I compute LV from SV Equi power separation?

2) How many days back should I use for separation?

3) Does the sep calculation change if I do trade on various time frame, i.e. 5 min, 15 min ,60 min or daily?

4) I have read in this forum a nice suggestion about to consider above 1 standard deviation as large and below as small, does anyone use this method? and does it performs better or worst compared to equi power separation?

Thanks for any help and suggestions.

acepsut

Acepsut,

Thank you for your questions.

To calculate EV in real time, you simply use the total volume at the last minute as if that last minute was the end of the day. This means of course that the separation volume must be recalculate every minute.

Also, make sure to use the smallest possible intervals. One minute is fine. 15 minutes is not good. I also found out that stocks that trade less than 300K shares per day display unreliable EV patterns.

The EV formula is below. It is straightforward to implement.
The LEV/SEV separation calculation is an iterative process that itself is rather fast: I calculate that separation in less than 2 sec for about 1000 stocks.

(ABS (Closei-1 - Closei )+ PI) / (Highi - Lowi + PI)

Closei-1 = Closing price corresponding to Time Interval (i-1): TIi-1
Closei = Closing price corresponding to Time Interval (i): TIi
Highi = Max (Highi , Closei-1)
Lowi = Min (Lowi , Closei-1)
PI = Price Interval (usually US$ 0.01).
ABS= Absolute value
If (Closei-1 = Closei ) then EV = 0


Pascal

[acepsut]

Hi Pascal,

thanks for your reply.

Both the logic for Effective Volume Flow and the 3 separation methods are clear to me, what I still have to clear understand is how to consider LV from SV separation.

Back to your book example, you say that for yesterday's Tellabs example, total effective volume shares was about 6,750,000 (page 52).

Then, using the Equi power method, you come to split to a separation for 42,500 shares as Large Players, all the rest as Small Players.

In other words, only the bars showing EV >= (ABS) 42,500 can be listed as Large Players.

That's fine, but it is about yesterday once we know how many shares was traded and how much volume was involved.

But what about today? When market opens, should I still have to consider this 42,500 threshold as LV - SV split?

For what about time frame, I store stock data in mySQL database @ 1 min time frame, so there is no problem for me to compute row by row the Effective volume, but it is not clear on how much back data I should compute the separation.

Do I have to accomplish separation based on yesterday's data and use this result for today's trades?

Or should I use a moving window of back data and update it for any incoming new data?

Is it correct to say that, for 60 min trading time frame, to sum up all the single 1 min EV for that hour interval?

As last question, can the standard deviation method be a good alternative approach for separation? And if so, on how many 1 minute bars back should I compute the SD?

My best regards,

acepsut

[Pascal]

To calculate EV in real time, you simply use the total volume at the last minute as if that last minute was the end of the day.

For example: if the last minute in RT is 11:32, then you calculate the separation volume between 9:30 and 11:32 as if 11:32 was the final trading minute of the day. You repeat that process for every minute until you reach 16:00

I did not use SD methods and hence cannot tell you how it works. I think that results would be very similar.


Pascal

[acepsut]

Thanks for your suggestions: I am going to develope your approach in Java, hope to have your support for further details and explainations.

acepsut