Econometrics: why co-integration is needed - page 12

New comment
 
faa1947: I don't think I'm familiar with it. Would it be possible to have a link.
There are a couple of books. Both in English, published by Wiley & Sons.
 

faa1947:

I don't think I'm familiar with it. Could we have a link.

Statistical arbitrage is thirty years old and was first developed at Morgan Stanley. The concept of cointegration is key to it.

faa1947:

Spread-trading is the certainty that from the extremes the quote will return to zero. but after how long?

This value can be estimated using stochastic calculus methods.

tara:

The spread characterises the willingness of market participants to trade and does not characterise anything else. If we are talking about the spread as the difference between the Ask and Bid of one instrument.

If you have placed a limit order in the market it means you already want to trade. If you have two Limit orders - for buying and for selling, it means that you also want to trade and you have some idea of a fair price. At the same time, you cannot specify a fair price to the nearest one pip, and this inaccuracy is the source of your risk. Obviously, you will want to control this risk and place one order below the expected fair price and another above it. In doing so, the greater your uncertainty about the value of the fair price, the further you will place your orders.

Thus, the spread reflects information about the uncertainty of market participants about the behaviour of the price.

If you do not want to trade - you simply will not place orders.

 
More on the topic of cointegration and spread http://www.russian-trader.ru/forums/content.php?r=48-pravduk-regression
 
Avals:
More on the topic of cointegration and spread http://www.russian-trader.ru/forums/content.php?r=48-pravduk-regression
No one has heard of the unit root test, and they should.
 
faa1947:
No one has heard of the unit root test, and they should.


In fact, you can actually see when it passes.) A flat is a flat. But for the sake of formality it can be run, or Hearst can be measured. The main thing is the robustness of such a spread - so that in the future it will be entered in approximately the same way.
 
Avals:

You can actually see when it's going)) A flat is a flat. But for the sake of formality you can run it, or measure Hearst. The main thing is the robustness of such a spread - so that in the future it will be entered in approximately the same way.

Robustness and non-stationarity are not compatible. I've had a quick look at Pravdyuk and didn't notice any struggle with non-stationarity. If you haven't watched it, the level of this arbitrator is simply astounding in its ignorance.
 

Cointegration based arbitrage seems questionable to me.

let's calculate the cointegrating vector between EURUSD and GBPUSD.

Let's calculate the series, which is a stationary difference between these two quotes using the formula:

genr coint = eurusd -(1.156203*gbpusd - 0.455564 - 0.000106*@trend)

Let's show everything on one chart

Profits on arbitrage where there are divergences, BUT the MOTION is AGREED. Made a profit check at several points. Only once 43 pips and usually 10 to 20, but that's close to the spread.

The whole thing is questionable.

 
anonymous:
If you have put a limit order in the glass, it means you already want to trade. If you have two limit orders, one to buy and one to sell, then you also want to trade and you have some idea of a fair price. At the same time, you cannot specify a fair price to the nearest one pip, and this inaccuracy is the source of your risk. Obviously, you will want to control this risk and place one order below the expected fair price and another above it. In doing so, the greater your uncertainty about the value of the fair price, the further you will place your orders.

Thus, the spread reflects information about the uncertainty of market participants about the behaviour of the price.

If you don't want to trade - you simply won't place orders.


I agree! (on the uncertainty of behaviour).

That's what I mean: the spread reflects the participant's desire to trade (not 'want it', but the degree of desire, or 'enthusiasm').

Orders will be placed anyway, in most cases. Example:

A currency exchange office. Oil is bouncing: People are craving for rubles. I'm buying currency at a price below the plinth. I sell it as it was yesterday.

 
faa1947:

I'm waiting. I'll be sure to answer.


San Sanych, can't you do it this way?

Files:
20120205_1.zip  29 kb
 
faa1947:

Like oversold/oversold?

Simultaneously, at the right point, we buy the euro and sell the inverse dollar index, like a loca. Then at another point, we close both operations and open reverse ones, i.e. sell the euro and buy the inverse index. Here the whole point is in the rate of return per unit time - judging by your pictures roughly it is something like 30-40 pips per week. If we exclude spread/swap in instruments and slippage at opening/closing we get 15-20 pips per week. In general it is not bad as long as it is stable and this is what we should check very attentively.

1...5678910111213141516171819...28
New comment