A Kinetic Energy Idea

[MAESTRO]

Quote from Thunderdog:

Thanks. I'll just sit back and watch as the thread evolves and develops. I'll try to limit myself to just nodding knowingly, with perhaps an occasional "Hmm," rather than give myself away with a boneheaded comment.

I am running 2 year tick-by-tick test now to determine the "bounce" after the impact in different market conditions. The results I am getting are pretty interesting. I just do not want to contribute to it only by my self. I would like to hear other constructive thoughts.
Cheers

P.S. Today is a perfect day to test this idea. I am running one small real money test account (< 15K) and will update with the results shortly.

 

[MGJ]

Very well, since you asked for ideas and suggestions, here is one: Run a correlation study between indicators A and B. A is your kinetic energy indicator, and B is the DOBRR (Dumb Ordinary Boring Range Ratio) shown below. Run the correlation to see just how much A and B mimic one another.

DOBRR = Range_of_current_bar / N_bar_Average_True_Range

Perhaps the Dumb Ordinary Boring Range Ratio can explain 85% of the activity in the kinetic energy indicator. Wouldn't that be interesting?

 

[MAESTRO]

Quote from MGJ:

Very well, since you asked for ideas and suggestions, here is one: Run a correlation study between indicators A and B. A is your kinetic energy indicator, and B is the DOBRR (Dumb Ordinary Boring Range Ratio) shown below. Run the correlation to see just how much A and B mimic one another.

DOBRR = Range_of_current_bar / N_bar_Average_True_Range

Perhaps the Dumb Ordinary Boring Range Ratio can explain 85% of the activity in the kinetic energy indicator. Wouldn't that be interesting?

Yes, it is interesting. As I said before I do use ATR as a normalizer. The difference is, of course is that I use Volume as well to modulate the square of price rate of change. What it does, IMO is it filters better the "worthless" runs. All together the difference between normalized price change and the Volume weighted price change is quite apparent. Right now my efforts are focused on using a better normalization techniques. I am currently testing FIR filters to see the difference.
Thank you for your input.

 

[gehko]

I programmed my own version of what your are trying to do (i think ) and I am finding that instead of filtering out worthless runs, it is signaling exhaustion (not necessarily reversal) of current move. I attached a 1min es chart.

 

[EllisWyatt]

Quote from MAESTRO:

The purpose of this thread is discussing a little idea I had. The background is simple. One could see the markets as a pool of some viscose liquid. If someone throws a rock in to this “pool” the wave that is caused by this action will be larger or smaller depending on the weight of the rock. It is easy to demonstrate that the amplitude of the wave and its duration in time (before it fades away) will depend on the kinetic energy stored in the rock’s movement and the level of liquid’s viscosity. The kinetic energy is something that one could figure out using grade’s 5 math:

E = (m * square(v))/2, where “m” is the mass of the rock and “v” its velocity. Applying this formula to the markets is also easy. One could possibly use the following interpretation: “m” = Volume of shares, “v” = Price change in the unit of time (price’s rate of change). So the “kinetic energy” of the market moving event of any kind could be assessed by squaring the price rate of change and multiplying it by the number of shares traded in the same period of time that has been used to assess the price change. The tougher question, of course, is how do we model the viscosity? One of the ideas I had is to use a simple ATR indicator to divide the Kinetic energy by it to normalize the energy by the size of average swings observed in the market. I have built a few indicators using this idea and applied it to a chart. At some point of time I will post that chart but for now I would like to discuss the approach first and listen to the ideas that folks might have. It is quite easy to program such an indicator in any charting software to see the results. I personally plotted it as a histogram. It did produce surprisingly interesting results for such a simple indicator, but I don’t want to influence your opinion and would like to discuss this idea first.

Cheers.

Interesting idea. In your paradigm, you could consider "viscosity" to be a function of liquidity in the order book -- that is, size of the aggregate limit orders at each price level, weighted inversely for distance from best bid/offer.

For example, a market with relatively thin offers would have lower viscosity to the upside. A market with relatively thick bids would have higher viscosity (drag) working against a down move.

 

[Corey]

First, pardon any of my ignorance. I may be misunderstanding your proposed system.

I think we should probably outline some of the assumptions of the system.

For example, at time 0, do we assume that the system is 'stable'? If so, at what point do we assume to be t=0? Choosing any arbitrary point in the market discounts all activity previous to that time as non-events. I suppose this could be done with relative confidence if you were able to determine 'turnover' for whatever market you are working in, and start t=0 as several 'turnover' cycles previous to the point you would like to evaluate.

Also, what does a more 'turbulent' or 'calm' liquid tell us about the market? Do we have buy volume as amplifying forces and sell volume as dampening volume? If not -- wouldn't this metric be a simple measure of current market volatility relative to some dampening constant?

What other sorts of objects can we think of the market as? Perhaps a ball on a table that is constantly changing its angle, determined by the length of the legs of the table (buy/sell volume)? Perhaps as a tug of war? Perhaps as a large ball of clay where buy volume 'hits' the ball, adding to its mass, where sell volume subtracts from the ball.

I have really appreciated your concepts on volatility measures. I think examining volume relative to the price movement it causes is very important...in fact, I would use this as my definition of momentum (mass being volume and velocity being price movement for the time period). Unfortunately, I don't like measuring in periods of time -- as far as I can tell, time adds an arbitrary degree of freedom that can be removed using data relative to ticks or volume. The only issue in this case would be that mass would have to be redefined...

I will end my ramblings here. Perhaps something in there makes sense....

 

[syswizard]

Quote from MAESTRO:

As I promised, here is the chart with the Kinetic Energy indicator on it. I also marked trades using this indicator and targets based on the impact of the pulse.

Sorry, upon closer review of the daily and then 1 minute data and the signals generated, I don't see any tradeable edge using this indicator. Your idea may be sound, but your implementation needs more work.

 

[MAESTRO]

Quote from gehko:

I programmed my own version of what your are trying to do (i think ) and I am finding that instead of filtering out worthless runs, it is signaling exhaustion (not necessarily reversal) of current move. I attached a 1min es chart.

Yes, that is the feeling I get as well. Did you normalize the indicator by the ATR or it is just straight composition of the square of the velocity multiplied by volume? I think normalizing make the picture a bit clearer. Thank you for your input.

 

[MAESTRO]

Quote from EllisWyatt:

Interesting idea. In your paradigm, you could consider "viscosity" to be a function of liquidity in the order book -- that is, size of the aggregate limit orders at each price level, weighted inversely for distance from best bid/offer.

For example, a market with relatively thin offers would have lower viscosity to the upside. A market with relatively thick bids would have higher viscosity (drag) working against a down move.

A very nice idea! Let me try to program that in. It is resonating well with what I would like to accomplish! Thank you so much!

 

[MAESTRO]

Quote from Corey:

First, pardon any of my ignorance. I may be misunderstanding your proposed system.

I think we should probably outline some of the assumptions of the system.

For example, at time 0, do we assume that the system is 'stable'? If so, at what point do we assume to be t=0? Choosing any arbitrary point in the market discounts all activity previous to that time as non-events. I suppose this could be done with relative confidence if you were able to determine 'turnover' for whatever market you are working in, and start t=0 as several 'turnover' cycles previous to the point you would like to evaluate.

Also, what does a more 'turbulent' or 'calm' liquid tell us about the market? Do we have buy volume as amplifying forces and sell volume as dampening volume? If not -- wouldn't this metric be a simple measure of current market volatility relative to some dampening constant?

What other sorts of objects can we think of the market as? Perhaps a ball on a table that is constantly changing its angle, determined by the length of the legs of the table (buy/sell volume)? Perhaps as a tug of war? Perhaps as a large ball of clay where buy volume 'hits' the ball, adding to its mass, where sell volume subtracts from the ball.

I have really appreciated your concepts on volatility measures. I think examining volume relative to the price movement it causes is very important...in fact, I would use this as my definition of momentum (mass being volume and velocity being price movement for the time period). Unfortunately, I don't like measuring in periods of time -- as far as I can tell, time adds an arbitrary degree of freedom that can be removed using data relative to ticks or volume. The only issue in this case would be that mass would have to be redefined...

I will end my ramblings here. Perhaps something in there makes sense....

Actually, a lot of what you said makes perfect sense to me! I got the same feeling that we need to get away from the linear time to see the picture clearer. I will try to use different measures of energy. The results will be posted.