Econometrics: why co-integration is needed - page 21
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
In the Russian-language Internet, the term highlighted in blue is referred to in this context as martingale.
Nothing personal, just martingale is a characteristic of a random process, but in no way the manner of trading.
On the Russian-language Internet, the term highlighted in blue is referred to in this context as martingale.
Nothing personal, just that martingale is a characteristic of a random process, but in no way a manner of trading.
In fact, both martingale and martingale are used to refer to a strategy of increasing bets when losing.
An example, from the forum: "[Bond] was martingale on red at table five. ... Mr. Bond is persistent, he knows how to take risks." (c) Fleming, Casino Royale.
Also, "Calling the game tonight. "They won't agree to a quittin'. One remedy: a double-- on the martingale. Your father bailed me out. "I'll shake off the old days. It's win or lose. If I'm lucky till midnight, good, if I'm unlucky, you and I will be out of the water." Martingale (from Notes of an undertaker) by Odoevsky.
Martingale may not be very modern, but it's historically literary, plus it points to the horse's roots. :)
A couple of observations. Not in the sense of criticism, but as thoughts on the subject.
Using the results of the contest for experiments is not really justified. The matter is that some other approaches are used in contests, in contrast to "normal" trading. Therefore, the analysis results are not quite correct.
On spread trading techniques. One of them is what you describe. This is not exactly "pairing" trading, but it uses some synthetic indicator and basically works out the same way. The other method is classic, trading on the movement of the indicator towards zero. I.e. at highs (or so) we enter on both instruments (in case of basket trading, buy basket), at zero we exit. Another method, do the opposite, enter at zero, exit at the highs (or at a certain distance from zero).
How many clever articles, but no use, you guys are engaged in pseudoscience)
I've tried to do that, everything is fine, except for the trend. And using cointegration is just improving the quality of exit entry, but there are other ways to solve this problem.
This is a bit odd, because the whole point of paired trading is to buy and sell two related assets at the same time. In this case, it is assumed that no matter where the price of both assets goes (a consistent trend of both assets up or down), you will make a difference when the assets converge. Basically, it is because of this assumed property that people engage in pair trading, they do not need to consider the trend. In reality, of course, things are not as rosy as we would like them to be.
The second point, cointegration was invented and justified for "fundamental" financial series. There are very serious reasons to suppose that the series are interconnected not only with statistics, but also with big fundamental factors. The currencies are interconnected, but not directly, the links are non-linear and often indirect, moreover they are separated in time. This is why if some good coefficients are obtained using currencies, do not believe it, there is no real cointegration, all these results are false.
But, paired trading itself can certainly be done, given a certain set of factors.
This is a bit odd, because the whole point of paired trading is to buy and sell two related assets at the same time. In this case, it is assumed that no matter where the price of both assets go (a consistent trend of both assets up or down), you will make a difference when the assets converge. Basically, it is because of this assumed property that people engage in pair trading, they do not need to consider the trend. In reality, of course, things are not as rosy as we would like them to be.
In theory, yes. If the difference = 0, your profit will be the difference between the currencies when you enter the market. If the gains of both currencies are the same. And different currencies have different gains when they move.
I studied the EURUSD <-> GBPUSD pair. Everything is fine when the model is built. Result: Profit factor in pips = 1.1. Profit factor in trades = 1.05. It does not include spread. I do not discuss this profit factor, though it is a very encouraging result for MM. Maybe it could be pulled out.
This is a bit strange, because the whole point of paired trading is to buy and sell two related assets at the same time. In this case, it is believed that no matter where the price of both assets go (a consistent trend of both assets up or down), you will make a difference when the assets converge.
In theory, yes. If difference = 0, your profit will be the difference between the currencies when you enter the market. If the gains in both currencies are the same. And different currencies have different gains on the move.
Maybe we are talking about different things.
Here is an example, a classic case of arbitrage.
Two assets, Blue and Green. If at point 1 you sell Blue at 6, and buy Green at 2, the profit at point 2 would be = 4. This is the simplest case. But when there is an up trend.
Again, it's easy to see that if you buy and sell at point 1, the profit at point 2 will be = 4. In this case, we will of course incur a loss on Blue, but it will be offset by the profits of Green. It is the same if the coordinated movement of the assets will be down.
Nothing changes, the total profit remains = 4. That is why they say that trends do not matter in pair trading. In the simplest case, as in the pictures, it is so, but in reality not quite.
I investigated the EURUSD <-> GBPUSD pair. Everything is fine when the model is built. Result: Profit factor in pips = 1.1. Profit factor in trades = 1.05. This is without taking spread into account. I do not discuss this profit factor though it is a very encouraging result for MM. Maybe it could be pulled out.
In principle it is, but on the other hand, here is today's situation on these two currencies.
The situations described in the figures above occur here as well. The whole issue is to draw the red line competently, the synthetics.
Why should having correlation between instruments increase the chances of them converging to zero? The idea is that two correlated random variables will form a third random variable (spread) with each other whose variance will be smaller, but its characteristics will be the same, i.e. we will have the same infinite random walks, albeit on a smaller scale.