m-lerner / jbooktrader

Automatically exported from code.google.com/p/jbooktrader
0 stars 0 forks source link

Measuring "motion" profile of market forces enhancement #4

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
>What steps will reproduce the problem?
No problem. New feature request.

>What is the expected output? What do you see instead?

I expect to address bookbalance FirstDerivative, SecondDerivative and
ThirdDerivative.

Under the application namespace it can be called:
Indicator.FirstDerivative(), Indicator.SecondDerivative(),
Indicator.ThirdDerivative();
or
Indicator.DepthBalanceVelocity(), Indicator.DepthBalanceAcceleration(),
Indicator.DepthBalanceJerk();

>What version of the product are you using? On what operating system?
JBT 2.08

>Please provide any additional information below.

1. Short theory.
I analyzed the application logic.
We try to measure "motion" profile of market forces.
The motion profile usually consists of up to 7 segments defined by the
following equations:
   1. acceleration build-up, with constant jerk = maxJerk
   2. move with constant acceleration, with jerk = 0
   3. approaching the desired maximum velocity, with jerk = -maxJerk
   4. move with constant speed, with jerk = 0 and acceleration = 0
   5. approaching the desired deceleration, with jerk = -maxJerk
   6. move with constant deceleration, with jerk = 0
   7. approaching the desired position, with jerk = maxJerk

2. What is to be calculated.
We are interested in:
a) finding points in time when motion is approaching the desired maximum
velocity if we want to be contrarians (take a position against market
forces looking for market correction).
b) finding points in time when motion is just started and is strong enough
to follow it. That case should correspond acceleration build-up stage, with
constant jerk = maxJerk.

3. What is to be implemented:
Indicator.FirstDerivative(), Indicator.SecondDerivative(),
Indicator.ThirdDerivative() and sample strategies using this indicators.

4. Future innovation and implementation.
Then we can calculate the best points for ENTRY and EXIT for strategy a and
strategy b for specific types of the market (e.g. ES or NQ futures) and
prove it statistically.

Kind regards,
Andrey Bogomolov.

Original issue reported on code.google.com by andrey.b...@gmail.com on 11 Apr 2008 at 6:10

GoogleCodeExporter commented 9 years ago
I agree that the derivates of market depth could be useful. I'll add them in 
the 
next release.

Original comment by eugene.k...@gmail.com on 11 Apr 2008 at 10:47

GoogleCodeExporter commented 9 years ago
Andrey, I am not sure what's the best way to approximate derivates from time 
series. 
If we had a nice smooth signal, it would be sufficient to take firt, second, 
and 
third differences to approximate the first, second, and third derivates. But 
because 
market depth is so noisy, the resulting derivates would be noise, too. Let me 
know 
if you have any ideas of how to do this better.

Original comment by eugene.k...@gmail.com on 13 Apr 2008 at 8:54

GoogleCodeExporter commented 9 years ago
1. Please make a noisePreFilter() function and call it before the first, 
second, and
third derivatives functions.
2. For the first try this function may count simple arithmetic average of 
market forces.
3. Later on we should experiment and create another function for noise 
filtering. I
consider it to be alike GARCH model. We should give more weights for high
"volatility" - in our terms more weights for high  market book balance.

Original comment by andrey.b...@gmail.com on 20 Apr 2008 at 2:29

GoogleCodeExporter commented 9 years ago
What noise correction methods do you think out from your Visual Recurrence 
Analysis
programming experience?

Original comment by andrey.b...@gmail.com on 20 Apr 2008 at 2:39

GoogleCodeExporter commented 9 years ago
Resolved in 3.04.

Original comment by eugene.k...@gmail.com on 8 Jun 2008 at 7:16