Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

FX statistical arbitrage 11 replies

Statistical arbitrage position closer required 3 replies

Statistical Arbitrage Trading by EA 34 replies

Statistical arbitrage, hedge, or bluffing myself? 16 replies

statistical arbitrage 1 reply

Attachments: Statistical arbitrage and pairs trading on Forex
Exit Attachments

Statistical arbitrage and pairs trading on Forex

Analiteg
As you explained, when spread goes to +2 we want to sell pair 2; so is it not better to attach the trading panel to chart of pair 2? In that case we can press the sell button which corresponds with "sell" action of pair 2.
And when spread goes to -2, we press on buy button which corresponds with buy action of pair 2 again.
Regards
As for me, it is not so important, to which schedule the panel is attached. The main thing is not to cross the direction. You can do as you like. I know that some of my traders generally open only one chart with a panel and do not look at the second chart.
Hi, analiteg. What is the best date-range to use on your real-time dashboard for short time trading?
Quoting rebol6
Hi, analiteg. What is the best date-range to use on your real-time dashboard for short time trading?
Hi!
I use 20 days.
  • Post #25
  • Quote
  • Edited 7:42pm Jan 4, 2022 4:54pm | Edited 7:42pm
hi analiteg, great info!
How do you calculate the lot size coefficient for different pairs? Or you always use 1:1 ?
Thanks!
Quoting analiteg
{quote} Hi, Mark. p-value is statistical value. Two time series are cointegrated if the p-value is less than 0.05. We should trade only cointegrated pairs.
Hi Analiteg,
Thank you for your response and this thread. I'm still trying to understand everything that you have presented.

I understand now why your looking for p-value less then 0.05 as 'A p-value less than 0.05 (typically ≤ 0.05) is statistically significant. It indicates strong evidence against the null hypothesis, as there is less than a 5% probability the null is correct (and the results are random). Therefore, we reject the null hypothesis, and accept the alternative hypothesis.'

My understanding is that calculating p-value is composed of 3 parts:
1. The probability random chance would result in the observation.
2. The probability of observing something else that is equally rare.
3. The probability of observing something rarer or more extreme.

Are you creating a statistical distribution from the spread of the 2 currencies to calculate p-value for co-integration? Why is spread considered stationary when price is not?

With the spread formula what is x in 'Spread = A - x*B'?

Did you use a co-integration formula to come up with the pairs that you are trading?
Quoting StJohns
hi analiteg, great info! How do you calculate the lot size coefficient for different pairs? Or you always use 1:1 ? Thanks!
Usually, I use 1:1. But you can calculate it as Pair 1 point value / Pair 2 point value.
Quoting Mark9999
{quote} Hi Analiteg, Thank you for your response and this thread. I'm still trying to understand everything that you have presented. I understand now why your looking for p-value less then 0.05 as 'A p-value less than 0.05 (typically ≤ 0.05) is statistically significant. It indicates strong evidence against the null hypothesis, as there is less than a 5% probability the null is correct (and the results are random). Therefore, we reject the null hypothesis, and accept the alternative hypothesis.' My understanding is that calculating p-value...
Hi, Mark!
More details about Cointegration and Stationarity you can find here.

A commonly untested assumption in time series analysis is the stationarity of the data. Data are stationary when the parameters of the data generating process do not change over time.

As we predict the spread, it should be a stationary value. We can check the spread for stationarity withAugmented Dickey Fuller test.
I have added to the dashboard spread stationarity value. It should be less than 0.01.

A cointegration value shows how strong the connection between currency pairs is. It should be less than 0.05.
If pairs cointegrated their spread is stationary.
One good person suggested to me, that if the spread is stationary, then the currency pairs are also cointegrated. This means, that there is no need to simultaneously check currency pairs for cointegration and spread for stationarity.

So in the dashboard, I will leave only the indicator of spread stationarity. If it is below 0.01, it means that the currency pairs are cointegrated, and you can watch the deviation of the spread from the mean and look for signals for trades.

Cointegration calculation is very expensive and requires a lot of server resources (CPU and RAM).
hi everybody,
this question is not only for analiteg.
I am trying to find a pairs spread deviation indicator since I'd like to try to automate the strategy so I could back-test it. Couldn't find anything like that though.
Any chance guys you could point me to such an indicator? Or give me some formula on how to calculate it?
I guess first one needs to calculate the mean value and then the standard deviation from that mean. Have no idea how to do it.
Thank!
Quoting StJohns
hi everybody, this question is not only for analiteg. I am trying to find a pairs spread deviation indicator since I'd like to try to automate the strategy so I could back-test it. Couldn't find anything like that though. Any chance guys you could point me to such an indicator? Or give me some formula on how to calculate it? I guess first one needs to calculate the mean value and then the standard deviation from that mean. Have no idea how to do it. Thank!
Hi!
If I knew how it can be done in MT4 or MT5, I would have done it there. But for calculations, I use python and libraries for data analysis. It's much easier and more reliable that way.

I have several ideas on how to send signals for a deal to an advisor in MT4 or MT5, and there the advisor will open deals. But for now it is in the plans for the future. This implementation requires some resources and time.
In the future, you can even train the neural network to trade using this strategy. But this is a little later.
Quoting analiteg
{quote} Hi! If I knew how it can be done in MT4 or MT5, I would have done it there. But for calculations, I use python and libraries for data analysis.
thanks for a quick reply analiteg!
I know Python. If it's not too much to ask, what libraries do you use to calculate the pair's spread mean and deviation?
Quoting StJohns
{quote} thanks for a quick reply analiteg! I know Python. If it's not too much to ask, what libraries do you use to calculate the pair's spread mean and deviation?
pandas, numpy, statsmodel
Today trades
Buy GBPJPY
Sell USDJPY
Profit +3%
Attached Image (click to enlarge)
Click to Enlarge Name: USDJPY-GBPJPY.png Size: 53 KB
Quoting analiteg
Today trades Buy GBPJPY Sell USDJPY Profit +3% {image}
great trade man!
3% of the position or account?
Quoting analiteg
Today trades Buy GBPJPY Sell USDJPY Profit +3% {image}
could you please show spread gragh to understand how you entered?
Quoting Merka
{quote} could you please show spread gragh to understand how you entered?
i think this was the scenario:
Attached Image (click to enlarge)
Click to Enlarge Name: enterexit.jpg Size: 156 KB
Quoting StJohns
{quote} i think this was the scenario: {image}
Is this the real spread chart related to that trade?
Quoting Merka
{quote} Is this the real spread chart related to that trade?
yes, but i don't want to steal the thread, let's hear from analiteg
Quoting StJohns
{quote} great trade man! 3% of the position or account?
3% from account.
1 trader viewing now