I'm a Zorro S subscriber and I've been looking into broker arbitrage.

I downloaded tick data using the download script from 4 brokers (B, C, D, E) and 3 pairs (EURUSD, GBPUSD, USDJPY) and then backtested all combinations.

All combinations were profitable and made tens of trades a day.

I then filled a couple of small accounts and had the script run live. In 24 hours not a single trade has been made. I cleared the data, downloaded it again and backtested it again and it shows many trades being made.

My conclusion is that the MT4 tick data must not be accurate. Is there anything I can do about this (like have Zorro record it live)?

Just remembered this is a beta feature and I should probably keep it in the beta thread...

I made similar experiences with test runs, esp. if using static spreads and commissions but using spread=marketVal() and recorded ticks or
downloaded tick prices (assetHistory) from FXCM result in realistic appearing scenarios (I didn't try assetHistory via MT4 so far because its history seems to be very short). But nevertheless the differences to live trading still persist.


I think this is the only way to get realistic results, esp. because the important time component (both signals generating and execution speed) of this strategy can hardly be simulated even if the tick material is valid and fine. And I fear that at last Zorro wasn't built to simulate strategies that are at least anywhere on the border to high frequency trading.

I think the biggest problem right now is broker support. MT4 just isn't suited for arbitrage with the execution speeds you see from it.

To get even closer to realism, I ended up writing a Python script that connects to cTrader's FIX API and records arbitrage opportunities. After you account for spread and commission (which eat into potential profits a lot), arbitrage opportunities are very fleeting.

Trading broker arbitrage via FIX API is also my imagination and AndrewAMD already announced to write a plugin, as his time permits.
Your Python solution is actually recording only or also able to trade?

I'm going to try writing a plugin too I think.

I've only implemented recording but there's no reason it can't trade too.

Hoo boy, people are counting on me to do this stuff?

It's a little tricky, it might become a premium product of mine.

I wouldn't code broker arbitrage in Python - C/C++/Lite-C should be preferred any day.

The Python was just because it was easy and I wanted to see if it even worked. Turns out it might. I saw some pretty high differentials throughout the day.

I've started work on C++ now and I've got quotes coming in. I'll see if I can get it into Zorro now. It's looking surprisingly achievable.

I really wish Zorro just gave me a damn callback for data though, this incessant polling is silly when I'm getting packets pushed to me over a network.

I have Zorro S but from script Download I can't dowload tick.
The file .t1 has generated but in the ZHistoryEditor I see data each minute

why?

Sounds hopeful promising! Nevertheless I would like to know why development wasn't pursued further on because from Zorro version 1.40 to ~1.5x 'FIX API template (for Zorro S)' was part of 'Planned for future Zorro versions' from where it vanished later silently.

Some brokers don't actually have tick data, even if they use MT4. When Zorro asks for it, they give you M1, which they do have.



I noticed that too!

My suspicion is that FIX would have to require code changes for certain brokers and giving away source is hard to control. Alternatively, they'd already built a FIX plugin for a client and weren't sure if it should be released or thought they'd build one for a client but the client changed their mind.