Forex Latency Arbitrage software from scratch PART 7

For connection to MT4 and MT5 I used Expert Advisors in MQL5 and MQL4 and ZeroMQ for communication.

[@jorgecamara1030]

Hello I do love all your videos info..Great support please don't stop helping the community 💗 I can't wait for your Expert Adviser for testing / HFT software

[@mikepapinski]

Thx 🙏 btw , where are you from ?

[@juanhernandezrepilado9630]

Hey Mike, again really good content here!! Thanks for this material !!

[@mikepapinski]

Thank you 🙏 tomorrow new video on that series 😃

[@jombonny1202]

how to get the software

[@CompilerZ]

Hey Mike, Why most broker using metaTrader 5 the minimum tick speed ( 95ms + ) at real-time quotes ! any Idea why not receive less than 95 ms I tried a lot broker using MT5 minimum tick speed ( 95ms + ) !!! ,Any idea ?

[@mikepapinski]

this is interesting, maybe they have a loop on their server side with a wait spin of +/- 100ms , They populate their internal order book from their own data feed real time and scrap the quotes and distribute to the clients every 100 ms? They cost of doing this would be much more cheaper than REAL real-time quotes... This is just my free thinking. I have seen in many application either wait spins with 100 ms or 50 ms, no idea why not 49 ms or 98 ms , maybe its some unconscious bias to keep numbers round. But anyway I would bet they might have wait spin on every 100ms from what you describe. This is interesting, I never measured the interval speed for a provider.

[@juergenx298]

@@mikepapinski Thats correct - and therefore latency arb didn't work very good. Triangulare arb would be an good alternative.

[@HFTHaidra]

Downloading Video Done.

[@mikepapinski]

😂

[@HFTHaidra]

For Example If you Have 100 Strategy run You need to Open Lmax Account Master[ifCopyTrade]/Fast[if1Leg] 100 Time , and 100 Slaves/Slow Account In Total 200 Account, Not 1+100=101 , Each Strategy has Owen FastSlow/MasterSlave Accounts In this way, each strategy is separate from the other And do the same thing with the Symbols, each Symbol you separate from the others Does this solve the problem?

[@mikepapinski]

this approach is ok but the issue I see is when you separate the symbols from each other you would most probably make a strategy that will loop with some 1ms wait spin to get all the actual quote data and perform a calculation, there is a risk of loosing data? When I was testing High volume of data, Some times I had missing ticks as I had few updates every 100-200 ns so the 1 ms loop was missing quotes. I ended up using dedicated quote stream using lmax disruptor circular buffer pattern also sticking to value types helped a lot when it comes to increasing latency and throughput. Currently I have few handlers on my circular buffer on separate threads that are running specific strategies and also storing all the quote ticks to back office with correct timestamps for future simulations.