Hearst index - page 29

 
It's not like I was suggesting a quote :)
 
Rorschach:

I'm getting some kind of bullshit here.

We take 0.5, we take the period in ticks, then the size of the candle should be proportional to the root of the tick volume. Right?

But according to the statistics obtained it turns out that the value of the candle is proportional to the tick volume?


Not the size of the candle, but the sko (sigma). You should take Open-Close instead of High-Low. I.e. sum over all candlesticks (Open-Close)^2, divide by the number of candlesticks and calculate the root. Then this value will be proportional to the root of the tick-wave, if H=0.5. It means for example for a 100 tick increment the Sigma will be X points and for a 200 tick increment X*SQRT(2) points

P.S. By the way, here are similar studies http://forum.fxclub.org/blog.php?b=712

 
Avals:



P.S. By the way, here are similar studies http://forum.fxclub.org/blog.php?b=712


" To calculate the "cost" of ticks is simple, you need to divide the size of a candle by the number of ticks coming. The advantages of this approach in demonstration of tick variability are undoubted - now it's clearly seen that the dependence of the candle size on ticks is not straightforward.

You can see that at times when there is increased activity in the market, the 'value' of the ticks decreases."

It is difficult to call it a study...

The fact of non-linear dependence of the candle size on the number of ticks is established, isn't it?

Here's where the author would like to relate this to non-entropy. But that's in our other thread.

;)

 
Mathemat:

Rorschach, I don't understand how the data was collected. In one 24-hour period - or by significant history?


250k bars
 

While we're on the subject... I did a quick calculation.

Actually, the dependence of bar size on tick volume turns out to be quite non-trivial. So much so, that I even suspect a mistake in the code (but cannot find it:). I was calculating EURUSD1 with 250 000 bars from 10.06.11 till 13.02.12

int start()                                     // Спец. ф-ия start()
   {
      int h = FileOpen("sz_to_ticks.csv",FILE_WRITE|FILE_CSV);
      FileWrite(h,"Volume","Size");
      for(int j=Bars-1;j>=0;j--)
      {
         int v = Volume[j];
         int sz = MathAbs(Close[j]-Open[j])/Point;
         FileWrite(h,v,sz);
      }
      FileClose(h);   
   }
 

Correct me if I'm wrong -- volume (teak) should grow roughly quadratically with size.

And the picture in general is interesting. It looks very much like a reflection of a feature of the quote filter.

 
TheXpert:

Correct me if I'm wrong -- volume (teak) should grow roughly quadratically with size.

And the picture in general is interesting. It looks very much like a reflection of a feature of the quote filter.

1. Yes, +-

2. Yes, that's what I thought too. Interesting possible extensions...

 

Let me put it this way for now:

the probability of more ticks during a given one-minute bar a) depends inversely on volatility b) falls sharply after reaching a volume of 150 per minute. I.e. if volatility is high, the Day trading center simply stops ticking when some minute volume is reached.

 
alsu:

Let me put it this way for now:

the probability of more ticks during a given one-minute bar a) depends inversely on volatility b) falls sharply after reaching a volume of 150 per minute. I.e. with high volatility, the DC simply stops ticking when some minute volume is reached.

five digits?

"Swallowing" should be somewhere, up to the threshold, less...

At four digits you would need to look at the same period and asset.

Or vice versa ;)

 
avatara:

On the four marks, the same period and asset would have to be looked at.

Yes.