Complex arbitration

 
I think there is no sense to continue the discussion of this speculation method, especially because its definition is finally determined (sorry for tautology) in Jury Reshetov's thread.

So we have two correlated currency pairs a/b and c/b, and correspondingly three currencies a, b, and c. It is not a secret that correlation is not absolute (at least for various Forex instruments) and its delta fluctuates around zero axis. Besides zero axis we have the maximum of correlation for the same period and its minimum. The method comes down to the following:
If at the moment the values of the correlation between a/b and b/c for period N are positive and are located near their maximum, then after a while the values of the correlation between the same instruments for the same period will be negative and will reach or nearly reach their minimum. Such discrepancies between the instruments allow arbitrage trades even within one brokerage company.

Unfortunately (or maybe fortunately) there is no possibility to backtest multicurrency in MT4, so I will post the results of forward testing. The Expert Advisor is not very advanced, I finalized it only last night, but it implements my idea to the full extent. Here is the result of last night's work:
Account: 472823 Name: DrawDown Currency: USD 2007 May 10, 02:31
Closed Transactions:
Ticket Open Time Type Lots Item Price S / L T / P Close Time Price Commission Taxes Swap Profit
12838686 2007.05.09 21:07 balance Deposit 10 000.00
12839092 2007.05.09 21:18 buy 1.00 gbpusd 1.9935 0.0000 0.0000 2007.05.09 21:22 1.9936 0.00 0.00 0.00 7.00
12839094 2007.05.09 21:18 sell 1.00 eurusd 1.3527 0.0000 0.0000 2007.05.09 21:22 1.3526 0.00 0.00 0.00 10.00
12840181 2007.05.09 21:52 buy 1.00 eurjpy 162.41 0.00 0.00 2007.05.09 22:10 162.47 0.00 0.00 0.00 49.95
12840185 2007.05.09 21:52 sell 1.00 chfjpy 98.54 0.00 0.00 2007.05.09 22:10 98.60 0.00 0.00 0.00 -49.94
12839851 2007.05.09 21:42 buy 1.00 nzdusd 0.7340 0.0000 0.0000 2007.05.09 22:42 0.7342 0.00 0.00 0.00 40.00
12839853 2007.05.09 21:42 sell 1.00 audusd 0.8278 0.0000 0.0000 2007.05.09 22:44 0.8279 0.00 0.00 0.00 -20.00
12842125 2007.05.10 00:55 buy 1.00 nzdusd 0.7330 0.0000 0.0000 2007.05.10 01:07 0.7336 0.00 0.00 0.00 120.00
12842126 2007.05.10 00:55 sell 1.00 audusd 0.8277 0.0000 0.0000 2007.05.10 01:24 0.8279 0.00 0.00 0.00 -40.00

0.00 0.00 0.00 117.01
Closed P/L: 117.01
Open Trades:
Ticket Open Time Type Lots Item Price S / L T / P
Price Commission Taxes Swap Profit
No transactions

0.00 0.00 0.00 0.00

Floating P/L: 0.00
Working Orders:
Ticket Open Time Type Lots Item Price S / L T / P Market Price
No transactions

Summary:
Deposit/Withdrawal: 10 000.00 Credit Facility: 0.00
Closed Trade P/L: 117.01 Floating P/L: 0.00 Margin: 0.00
Balance: 10 117.01 Equity: 10 117.01 Free Margin: 10 117.01

Details:
Gross Profit: 226.95 Gross Loss: 109.94 Total Net Profit: 117.01
Profit Factor: 2.06 Expected Payoff: 14.63
Absolute Drawdown: 0.00 Maximal Drawdown (%): 49.94 (0.50%)

Total Trades: 8 Short Positions (won %): 4 (25.00%) Long Positions (won %): 4 (100.00%)
Profit Trades (% of total): 5 (62.50%) Loss trades (% of total): 3 (37.50%)
Largest profit trade: 120.00 loss trade: -49.94
Average profit trade: 45.39 loss trade: -36.65
Maximum ($): 3 (66.95) consecutive losses ($): 1 (-49.94)
Maximal consecutive profit (count): 120.00 (1) consecutive loss (count): -49.94 (1)
Average consecutive wins: 2 consecutive losses: 1


Due to slippages and requotes trades sometimes close with no profit, sometimes with super profit (more than planned), and sometimes with a loss (though I haven't had any such in 4 trades yet). So today I increased max and min correlations needed to make a trade, and also increased take profit to 2 pips. There will be less trades, but the reliability will increase many times over.
 
By the way, the instruments and transaction directions are chosen in such a way that the total swaps are always positive. Therefore, you can wait for the delta sign to change as long as you like, but it rarely takes more than a day.
 
DrawDown:
if at the moment values of correlations between a/b and b/c for period N are positive and located near its maximum, then after some time values of correlations between the same instruments for the same period will become negative and will reach or will almost reach their minimum.
Oh, shaitan... I'm just now meditating on the correlation graphs and trying to figure out what I can get out of these flights of coefficients of correlation from 1 to -1 and back again.
By the way, it is possible to test on history as correlations are counted, and trades should be opened on one pair, then run on another pair and add up the results.
 
timbo:
DrawDown:
if at the moment values of correlations between a/b and b/c for period N are positive and located near its maximum, then after a while values of correlations between the same instruments for the same period will become negative and reach or almost reach their minimum.
Oh, shaitan... I'm just now meditating on correlation charts and trying to figure out what I can get out of these correlation coefficients flying from 1 to -1 and back again.
By the way, it is possible to test on history, because correlations are counted, and trades should be opened on one pair, then run on another pair and add up the results.

That's what I tried. Of the above trades in the tester 2 are missing and 2 have a time difference. That's why I preferred not to believe the results of backtest on minutes. Although the demo is not an indicator, but with a "noisy" brokerage company even with average slippage and requote frequency the system is able to show the same results as on a forward demo. So it is better to perform a forward test, though with an interrupted Internet.

By the way, I do not consider the correlation itself, as I could not see anything in the flight of its coefficients. I use only increments of the latter ;) The correlation itself does not change its sign.
 
DrawDown:

This I have tried. Of the above trades in the tester, 2 are missing and 2 have a time difference. That's why I preferred not to believe the results of backtest on minutes. Although the demo is not an indicator, but in case of a "noisy" brokerage company even with average slippage and requote frequency the system is able to show the same results as in forward demo. So it is better to perform a forward test, though with an interrupted Internet.

By the way, I do not consider the correlation itself, as I could not see anything in the flight of its coefficients. I use only increments of the latter ;) The correlation itself does not change its sign.

A couple of years ago I started dealing with MQL by writing a similar Expert Advisor. If you may look here This is a dealing center robbery and nobody will stand it for a long time.
 
New:
DrawDown:

This I have tried. Of the above trades in the tester 2 are missing and 2 have a time difference. That's why I preferred not to believe the results of backtest on minutes. Although the demo is not an indicator but with a "noisy" brokerage company even with average slippage and requote frequency the system is able to show the same results as on a forward demo. So it is better to perform a forward test, though with an interrupted Internet.

By the way, I do not consider the correlation itself, as I could not see anything in the flight of its coefficients. I use only increments of the latter ;) The correlation itself does not change its sign.

A couple of years ago I started dealing with MQL by writing a similar Expert Advisor. If you may look here This is a dealing center robbery and nobody will stand it for a long time.

Well, almost, but not quite. As I understand it, the principle of your Expert Advisor is triangular arbitrage. In the beginning I also tried to realize this idea, but could not find combinations of pairs, the aggregate swap of which would give profit (as described by Michel on Kreislick.com). Therefore this method of arbitrage is possible only on futures (maybe somewhere else, but obviously not on forex). And in my case pipsips may not be (or may not?!!!), but if anything, I can set the minimum time of position holding equal to or more than 24 hours. Then it definitely cannot be called "pips", and swaps will give a profit, because I do not include them in calculations of the latter. And the last deal looks like this:

12843214 2007.05.10 03:32 buy 1.00 nzdusd 0.7330 0.0000 0.0000 2007.05.10 03:40 0.7342 0.00 0.00 0.00 240.00
12843216 2007.05.10 03:32 sell 1.00 audusd 0.8309 0.0000 0.0000 2007.05.10 03:40 0.8320 0.00 0.00 0.00 -220.00

Profit equals 1 pip, but show me the place where this trade is called a pips.
 

DrawDown, alas, is still pure pipsing - even on the results of almost every trade. What to say about the expected payoff which is less than 1.5 pips... Increasing TP to 2 pips will not change anything qualitatively. And it is not serious to review the results within one day.

triangular arbitrage. I also tried to implement this idea in the beginning, but I could not find a combination of pairs, the total swap of which would give plus
.

Well, there are such triples of pairs. Another thing is whether there is a sense in searching for them.

Take a look at, say, here: http://kreslik.com/forums/userpix/2_FPI_ControlPanel_Rings_1.png . Not necessarily all crosses are quoted, but you can check them out.

 
Mathemat:


triangular arbitrage. I also tried this idea in the beginning, but could not find a combination of pairs whose total swap would be positive

Well, there are such triangular pairs. Another thing - is it reasonable to look for them?


Interesting, interesting... Can you send to drawdown!sobaka_rambler=ru the names and addresses of DCs with such swaps?
 
Mathemat:

Look at, say, here: http://kreslik.com/forums/userpix/2_FPI_ControlPanel_Rings_1.png . Not necessarily all crosses are quoted, but you can check it out.

Huh... it's just from this resource that I started trying to find pairs with swaps, allowing to hold positions opened by the method described there. It is possible to open them - no problem, but swaps will give minus.

My last trade shows 12 points of profit in NZDUSD and 11 points of loss in AUDUSD but not 1 point of profit in any of the instruments.
 
At the very least, you can try to prohibit the Expert Advisor to close trades until the price goes away from the opening level by a certain number of points. But in this case the number of deals will be reduced by 5-10 times. But it does not guarantee that the total profit of the hedge will exceed 1-2 points. My mobile phone battery ran out today. I got home only 2.5 hours later. The deal was still open. Now I looked, it could have closed 3 times during that time. I'll see what happens to it next.
 
Can you show me the advisor?