Discussion of high-frequency trading on MT5 - page 26

 
hrenfx:

When a competitor comes along, others have to retaliate.

Well that's all platitudes and corollaries.

But I'd like to hear the reasons.

So you said that their "thing" is the client terminal + api to their aggregation server. And there is no mention of FIXes.

And the fact that other LPs are improving their retail services, is it only the merit of LMAX? Or is it the general trend of vectoring towards retail and they just came out in the spirit of the times prepared in time?


As it happens, there's no systematisation in the posts. Spoken where one has to.

then we will have to repeat everything and more :)
 
sergeev:


then you'd have to do it all over again :)
Maybe you'd better look it up. Or who needs it... you or him?
 
Unfortunately google throws the answer back to the previous page :) already indexed.
 

Unfortunately, even for myself, I cannot clearly articulate why I am writing all this. But I have no desire to repeat myself.

The guys tried to create a FOREX exchange called MTF (= ECN). They wrote their own technology. They made some agreements with the banks to put their limits under the guarantees of their execution. In short, worked hard and created something innovative and notable at the time. At the moment they are a good LP for aggregation. Considering them as a standalone broker is not a good decision.

Ненормальный Agile в финансах
Ненормальный Agile в финансах
  • habrahabr.ru
Фирма, в которой я работаю, разработала свою трейдинговую платформу типа MTF. В этой системе ежесекундно производятся десятки тысяч торговых операций, и с помощью паттерна Disruptor, средняя скорость выполнения трейда не превышает 20.5 миллисекунд. В проекте задействованы сложнейшие интеграции с третьими сторонами — крупными банками, Лондонским...
 
hrenfx:

Unfortunately, even for myself, I cannot clearly articulate why I am writing all this. But I have no desire to repeat myself.

The guys tried to create a FOREX exchange called MTF (= ECN). They wrote their own technology. They made some agreements with the banks to put their limits under the guarantees of their execution. In short, worked hard and created something innovative and notable at the time. At the moment they are a good LP for aggregation. Considering them as a standalone broker is not a good decision.

Oh, that's great.
 

Just needed to dispel doubts about the correctness of the ECN/STP system. Below is the code of an MT4 script which buys from itself:

void start()
{
  double Lots = 1;

  OrderSend(Symbol(), OP_BUYLIMIT, Lots, NormalizeDouble((Bid + Ask) / 2, Digits), 0, 0, 0);  
  
  RefreshRates();
  OrderSend(Symbol(), OP_SELL, Lots, Bid, 0, 0, 0);
  
  return;
}

On STP (without ECN) this will not work.

P.S. And this action will cause you to buy twice from yourself. In this case you will end up with no open positions, and the profit will be equal to four one-way commissions ( = two two-way commissions):

void start()
{
  double Lots = 1;
  double PriceOpen = NormalizeDouble((Bid + Ask) / 2, Digits);

  OrderSend(Symbol(), OP_BUYLIMIT, Lots, PriceOpen, 0, 0, PriceOpen);  
  
  RefreshRates();
  int Ticket = OrderSend(Symbol(), OP_SELL, Lots, Bid, 0, 0, 0);

  RefreshRates();
  OrderClose(Ticket, Lots, Ask, 0);
  
  return;
}
 
hrenfx:

Just needed to dispel doubts about the correctness of the ECN/STP system. Below is the code of an MT4 script which buys from itself:

On STP (without ECN) this will not work.

P.S. And this action will cause you to buy twice from yourself. In this case you will end up with no open positions, and the profit will be equal to four one-way commissions ( = two two-way commissions):

It's nighttime already - your brain is "off" :)
 
A certain nub has broadcast a hilarious theory on the hubra. Take LMAX out of the narrative and you will laugh.

Let hrenfx rock, but there are no fish in this hole.
 
Renat:
On the hubra, a certain nub broadcast a hilarious theory. Take LMAX out of the narrative and you will laugh.

Let hrenfx rock, but there are no fish in this hole.

Yes, that article was more of a PR piece. Let's skip it and don't open it again.

But it does have links to another one there at the beginning, about the LMAX processing mechanism. And even with pictures.

...to achieve execution speeds in excess of 10,000 transactions per second, developed .... a multi-threaded parallel transaction processing framework featuring high throughput and very low latency.

...<on> replaces java.util.concurrent.ArrayBlockingQueueue and outperforms it by up to a factor of 80.


But the first comment from the hubra repeats my thoughts as well.

...ну кольцевой буфер вы привлекли для этого, очень хорошо, и что из того?
Сначала думал что у вас там хитрая lock-free стратегия какая-то, но вроде её тоже нет.
Вы недоговариваете про то как реализуется конкурентный доступ к тем же счётчикам.
А как насчёт fault-tolerance? Ну скажем этот unmarshaller не просто вышел покурить, а укурился совсем и его увезли на скорой в больницу?
Как-то неполно и нового ничего нет…
...мне приходилось самому разрабатывать такие системы в реальности (>30tps 24/7) и мне трудно поверить что на яве можно сделать что-то работающее в этом контексте.



All in all, everything looks PR.


---------

So all the same, what did they do?

worked hard and created something innovative and notable at the time. At the moment they are a good LP for aggregation.

hrenfx, can you explain where they have something special that others don't have?

 
Renat:
A certain nub has broadcast a funny theory on the hubra. Take LMAX out of the narrative and you will laugh.

Let hrenfx rock, but there are no fish in that hole.
Yes, the story also struck me as "a bit theoretical". What does your last sentence mean, can you expand on that?