Yet Another Statistical Arbitrage

[zkogan]

I don't know if this was done before, most likely it was, if so, please direct me there..
The idea is - we have major pairs, we have crosses, in my example I will use EURUSD/USDCHF/EURCHF and EURAUD/AUDUSD/EURUSD..

1. Everyone knows that cross can be synthetically traded via majors, like Buy EURUSD/Buy USDCHF = Buy EURCHF (may be completely wrong with directions here, but hope I'm not).

2. I charted spread of synthetic pair-cross pair and had some interesting results using ChartBuilder:

(note these are not close prices but highs-lows)

3. Assuming 1 pip spread + 0.5 pip (5$ roundtrip) comission (average ECN broker) we get about 5 pip negative for opening a hedging triangle.

4. If you look at the upper chart, you can see difference go up to 30 pips each side (50+ total definitely)

5. Can this be automated? Like catching those peaks, quickly opening a triangle (or just one-side trade) and getting profit? If all majors+their crosses could be monitored, that would be pretty great as many more situations occur each day I assume

Please post your thoughts and assumptions, and thanks!

[sonik]

Are you sure your math is correct? I don't see a 50 pip arb opportunity. I've written and tested a triangular arb EA before and the arb opportunities are barely high enough to cover spreads usually so I threw it in the garbage. And this was on a demo test. You could be holding a triangle for several days to get from one extreme to another to make profit, and by then the swaps will have nullified any profit.

[zkogan]

It is correct (I am 80% certain), there are possible problems though:
1) slippage, which can be pretty high (those arb situations almost everytime occur during HUGE volatility spikes)
2) lack of liquidity (everyone sells, noone buys for example, simply not enough lots in level 2 to be filled)

I found an EA concerning arbitrage, which is pretty cool: http://codebase.mql4.com/6245
but it should probably be modified and when I open the code, my eyes become like that 0_0 because of its complexity..

[WideEyed]

Buy EURUSD/Buy USDCHF = Buy EURCHF (

When you go long EU you're buying one Euro and selling approx 1.3 USD. When you go long USDCHF you're buying 1.00 USD.

The difference between 1.00 USD you're buying and the 1.3 USD you're selling represents risk - a net 0.3 short exposure on USD. You're not fully hedged and this is therefore not arbitrage.

[zkogan]

Yeah, haven't thought of that... Thanks!
Nonetheless, testing that Trade-Arbitrage expert proves it profitable, so I'll later test it on a small live account (which will die fast I believe due to slippage. lol) and post the results if they are positive

The thread can be closed I assume..

[eigenvector]

It is correct (I am 80% certain), there are possible problems though:
1) slippage, which can be pretty high (those arb situations almost everytime occur during HUGE volatility spikes)
2) lack of liquidity (everyone sells, noone buys for example, simply not enough lots in level 2 to be filled)

I found an EA concerning arbitrage, which is pretty cool: http://codebase.mql4.com/6245
but it should probably be modified and when I open the code, my eyes become like that 0_0 because of its complexity..

There are no 30-50 pip arb opportunities in triangular arbitrage. You are off by a factor of 10. I have drilled down on the triangular arbitrage concept very far and I can tell you with certainty that the supposed opportunities that you see in your charts are mythical when it comes down to execution. The most you see is the difference that three pairs can randomly fluctuate. When they move more than random fluctuation allows, the banks/dealers move their prices and thus the arb is eliminated. What you are seeing very likely is 3-5 pips but you are probably not accounting for spread either so it may show 5 pips but may be -1 pips of profit potential when you actually execute. At times it is more but often it is gone before you can take action. 99.9% of the time Prices are already moving when you see the opportunity - thus I say mythical. If there were free arbitrage opportunities then the banks, dealers and brokers would go belly up because in this case money made by you the retail trader comes out of your counterparty's pockets. It is important to understand that there is no free money here because the arb opportunity originates with the counterparty and not with the market - the same counterparty who can control whether you get an extra 2 pips of slippage or delayed fills etc. You are fooling yourself.

To find a real market edge you will need to dig deeper than the surface triangular arbitrage trade. And yes, you will need to figure out stuff like how to properly size the pairs such as Triangular Arbitrage Lot Size.

But starting with and understanding how the arbitrage trade works is the basis for all real inefficiency based edges in the markets. But for us retail guys, it will need to be more based on statistical arbitrage, meaning allowing for some directional risk with larger rewards than 1 pip here and there. We can't compete on execution speed but we can compete on knowledge based statistical strategies that take advantage of the same basic arbitrage dynamics. Three Pairs Hedging or any of the related or spin off threads are good places to start, but you will need to dig a little deeper and devise your own strategy that takes advantage of the triangular arbitrage dynamics to develop your own market based edge.

[WideEyed]

Don't forget that many brokers don't like arbitrage and will go out of their way to frustrate you. FXCM, for example, say in their terms that they will increase your spreads massively if they think you're trying to exploit arbs.

[zkogan]

Wow, thanks a lot eigenvector, that clarified many things..

WideEyed, as far as I'm concerned, if the broker is TRUE ECN one, it should be grateful if a customer generates a fortune in commissions

[Jason Rogers]

Don't forget that many brokers don't like arbitrage and will go out of their way to frustrate you. FXCM, for example, say in their terms that they will increase your spreads massively if they think you're trying to exploit arbs.

Hi WideEyed,

FXCM's spreads are automatically set based on the best bid and ask prices being quoted to us from 10+ liquidity providers. Nowhere in our terms does it state that spreads will be widened as you claim.

Jason