How to understand the execution speed of LMAX?

Feb 05, 2017 at 07:30
2,845 Просмотры
37 Replies
Участник с Sep 11, 2016   18 комментариев
Feb 05, 2017 at 07:30 (отредактировано Feb 03, 2017 at 10:29)
Hi traders,

I am considering to use LMAX, since I have heard the execution speed is quite high.

On its website, LMAX says the execution speed is less than 5ms. Then how can I understand this figure?

If the latency from my VPS to LMAX server is 4 ms, then the time from order send to be implemented would be: 4ms + 5ms = 9ms. This calculation is correct?

For example, now I am using another ENC broker and when I see my trade records, it seems it takes about 600ms from sending OrderModify to receiving 'order was modified' message. Can I understand the 600ms will be 9ms if I use LMAX? or I am missing some key items?

Thank you.
Участник с Feb 12, 2016   522 комментариев
Feb 06, 2017 at 09:18
beebird posted:
Hi traders,

I am considering to use LMAX, since I have heard the execution speed is quite high.

On its website, LMAX says the execution speed is less than 5ms. Then how can I understand this figure?

If the latency from my VPS to LMAX server is 4 ms, then the time from order send to be implemented would be: 4ms + 5ms = 9ms. This calculation is correct?

For example, now I am using another ENC broker and when I see my trade records, it seems it takes about 600ms from sending OrderModify to receiving 'order was modified' message. Can I understand the 600ms will be 9ms if I use LMAX? or I am missing some key items?

Thank you.

Hello,
If the execution time between registering the order in the LMAX system and filling the order is for 5ms, and the latency from your VPS is 4ms until the order reaches LMAX system, then your calculation should be correct.

From the example above I assume that perhaps currently you are using MT4 - and due to the trader server in MT4 env. the latency for order execution cannot be less then 300ms - to 400ms. On top of this, if there is a bridge provider between your broker and the LP or LPs, then you will have some additional delay in terms of ms.

If you are trying to find faster execution, just have in mind that LMAX does not like post news trades (spike trades) or latency arbitrage. :)
Участник с Sep 11, 2016   18 комментариев
Feb 06, 2017 at 09:54

Thank you Baldo. Helpful explanation. I understand a lot.
If that is the case, I think switching to LMAX has a huge advantage to reduce slippage. I am using a lot of stop orders and stop loss order. These seem to tend to slip a lot, and undermining my EAs.

[quote/]
If you are trying to find faster execution, just have in mind that LMAX does not like post news trades (spike trades) or latency arbitrage. :)

The comment above makes me a little concerned. If they do not like, what exactly happens? They would stop trading?

Участник с Sep 20, 2014   365 комментариев
Feb 06, 2017 at 11:45
LMAX is fast, but apparently very expensive to hold positions over night. Might want to factor that in.
 
Участник с Sep 11, 2016   18 комментариев
Feb 06, 2017 at 12:55
Thank you, theHand!

You mean, the overnight interest rate is high, right?
Then LMAX want traders to do shorter trades rather than longer trades?
Участник с Sep 20, 2014   365 комментариев
Feb 06, 2017 at 15:45
I wouldn't know why they do it, it's just what they do.

I don't suggest over complicating the issue either. LMAX is expensive at night, but fast. This broker is this and that broker is that. You will never find the perfect broker.

Just choose what works mostly for you.

But honestly I don't see why speed is an issue. You never mention why. There is no strategy you could possibly have that requires that kind of speed or you'd be paying brokers $100 000 000 a year just for the feed. And if you could do that you wouldn't be posting here.

So why the haste ?
Участник с Sep 11, 2016   18 комментариев
Feb 07, 2017 at 08:24
I did not know about the $ 100 M, for just for having feed. This is a huge money..

My strategies are kind of scalping, and takes 2-3 pips for profit. Using current broker, the execution speed is not super fast, may be due to Last Look, and there are about 1+ pips negative slippages for each trade in average. I feel if I can reduce this slippage, the profit of the strategies will be double, triple or more. This is the reason I am seeking the speed.

Участник с Jan 31, 2017   10 комментариев
Feb 07, 2017 at 09:40
Hello,

Getting direct account with LMAX is good but it depends upon your requirements. According to me open an account with LMAX has the only advantage of “ low Latency”.

However, open an account with LMAX is not an easy, they do have minimum capital requirement to open an account.
 In addition to that they do have maximum leverage restriction, according to what I heard from colleges, the maximum leverage they can provide is 100X in some currencies, 50X in some and 33X in some.

And also with LMAX, you wouldn’t get the custom trading benefits that you can easily get with the other forex brokers like higher leverage, Deposit bonus, Lower Stop out levels, low account opening balance requirement and lower swap charges.
Участник с Sep 11, 2016   18 комментариев
Feb 07, 2017 at 11:04
Thank you, hedgescalp.

I understand I need to consider various aspects as you mentioned. At least it seems that low latency seems to be a firm advantage.

As for stop out, of LMAX, it seems to be zero when I see Demo.
Участник с Sep 20, 2014   365 комментариев
Feb 07, 2017 at 11:08
@beebird

You can also reject a trade. Remember that. For what you're describing once your initial conditions are met I'd put the thing in a loop and check for the rest of the conditions before I send the order. Especially exits.

There's no reason why you can't code conditions, I want to be out at + 1 pip at spread x - keep trying till I get it.

O is your broker. They will never reject a trade or re-quote but their spreads are dynamic.

But I do this all the time, I tell the computer how I want to exit or enter, and you might find price moves around the levels all the time, doesn't matter, machine sits there and looks for it and when it's right it executes. You'll likely find what you're looking for doesn't just exist once, but many times before the market moves away from the level.

But I have to say on 2 or 3 second trades the broker makes more from it than you. Is it really worth it ?
Участник с Sep 20, 2014   365 комментариев
Feb 07, 2017 at 11:15 (отредактировано Feb 07, 2017 at 11:23)
And I have to say, if you're getting slippage it's just your broker **** with you. The fx is so liquid. Unless you're doing $10mil/$20mil per order you should not be getting slippage.

But the platform becomes really important. Forget about MT for this. Has to be API where the execution is handled their side, not your side.
Участник с Feb 12, 2016   522 комментариев
Feb 07, 2017 at 11:32
beebird posted:


[quote/]
If you are trying to find faster execution, just have in mind that LMAX does not like post news trades (spike trades) or latency arbitrage. :)

The comment above makes me a little concerned. If they do not like, what exactly happens? They would stop trading?


Hello,
What I mean was that, if you trade with auto-click or similar software sending the order just milliseconds before the official notification of the release and / or arbitrage software, most probably you will get a call from them to close your account or, they will set all trades from your connection to be rejected. There is another option, to get delay with execution and some additional slippage.

As you explained you will scalp the market for a couple of pips and in this case I do not believe you will have an issue with LMAX.
Участник с Sep 11, 2016   18 комментариев
Feb 07, 2017 at 14:46
Thank you theHand and Baldo,

Hearing your advice, now I am feeling I should use LMAX and there should be no problem.
I will not use the news trade as like whaf you mentioned, will be using 10 lots ( 1 million ) at most.

Regarding the platform, what exactly happens if I use MT4 and how the impact on latecy would be?

Using API is necessary in the end I found.Then I think there would be additional issues, such as:
- what language should I use? - C# seems to be nice since it can be used in cTradef also?
- how the backtesting should be? - i think MT4 backtesterr is convenient and fast, variable spread testing canbe done with a plugin.

Участник с Feb 22, 2011   4862 комментариев
Feb 07, 2017 at 15:23
theHand posted:
And I have to say, if you're getting slippage it's just your broker **** with you. The fx is so liquid. Unless you're doing $10mil/$20mil per order you should not be getting slippage.

But the platform becomes really important. Forget about MT for this. Has to be API where the execution is handled their side, not your side.

well even reputable brokers like FxPro could have sllipage like 5000ms. They told me they have some algorithm which evaluated trade and proceed it 5000 ms later. Sounds like dealing desk to me. But overall I am satisfied with them
Участник с Sep 20, 2014   365 комментариев
Feb 07, 2017 at 20:29
@togr

Slippage can't be in time, it's in price and generally caused by there being no counter party to your trade, so you can't get a fill and need to look at wider prices for the fill. So the best way not to get slippage is to send and order with a 0 slippage parameter and run the risk of the order failing. Then you have to send it again.

@beebird

With MT 4 you have the internet between you and the broker, so every action is individual and you have to wait for a response. So lets say you're trying to close a trade as I outlined above, with 0 slippage and it fails it might take 2 or 3 tries, each time you wait 200 or 300 milliseconds for a reply and you can't move on to other pairs. You're stuck trying to deal with that pair.

With the API or MT 5 you can simply fire the order and move on, come back to that pair later see if I got it. No waiting for a response from the server. Also trying to close multiple trades you send a single order and they close at their best server speed, you don't have to handle each order individually and wait for a response.

I wouldn't worry about back testing mate. It's useless, certainly in MT, just test or you'll never know what the results were. Back testing is only of use to see if your code works.
Участник с Feb 22, 2011   4862 комментариев
Feb 07, 2017 at 21:22
theHand posted:
@togr

Slippage can't be in time, it's in price and generally caused by there being no counter party to your trade, so you can't get a fill and need to look at wider prices for the fill. So the best way not to get slippage is to send and order with a 0 slippage parameter and run the risk of the order failing. Then you have to send it again.

@beebird

With MT 4 you have the internet between you and the broker, so every action is individual and you have to wait for a response. So lets say you're trying to close a trade as I outlined above, with 0 slippage and it fails it might take 2 or 3 tries, each time you wait 200 or 300 milliseconds for a reply and you can't move on to other pairs. You're stuck trying to deal with that pair.

With the API or MT 5 you can simply fire the order and move on, come back to that pair later see if I got it. No waiting for a response from the server. Also trying to close multiple trades you send a single order and they close at their best server speed, you don't have to handle each order individually and wait for a response.

I wouldn't worry about back testing mate. It's useless, certainly in MT, just test or you'll never know what the results were. Back testing is only of use to see if your code works.
@theHand
I am surprised you dont understand this
Участник с Sep 20, 2014   365 комментариев
Feb 07, 2017 at 21:56
I actually got another lesson on it just the other day from a Bitcoin site of all things, so I'm quite sure I understand it, it's a function of price, not time as you claim.

How can you have slippage of 5000 ms? That makes no sense at all.

FYI

https://www.investopedia.com/terms/s/slippage.asp
Участник с Feb 22, 2011   4862 комментариев
Feb 08, 2017 at 07:10 (отредактировано Feb 08, 2017 at 07:30)
theHand posted:
I actually got another lesson on it just the other day from a Bitcoin site of all things, so I'm quite sure I understand it, it's a function of price, not time as you claim.

How can you have slippage of 5000 ms? That makes no sense at all.

FYI

https://www.investopedia.com/terms/s/slippage.asp

OK so to be more precise
execution delay is 5000 ms
Time between I click sell or buy and trade is executed
This is of course causing slippage of price as you buy or sell at different price
Участник с Sep 11, 2016   18 комментариев
Feb 08, 2017 at 07:48
Thank you.

Slippage happens because of lack of counter parts. This happens if I trade the hige lots such as 10million right?
If my ordef is smaller than market depth (for example there are 10 million at the price I order in the market and I trade only 1 million) and order is executed immediately such as 5ms, there will be no alippage. My understanding is correct?

In terms of execution latency or slippage, APi and MT5 is equivalent? If so, switching to MT5 seems convenient in terms of the cost of learning a new language and the convenience of using a sophisticated platform.

Just concern i heard MT5 cannot control each order individually.
Участник с Feb 12, 2016   522 комментариев
Feb 08, 2017 at 08:00
beebird posted:
Thank you.

Slippage happens because of lack of counter parts. This happens if I trade the hige lots such as 10million right?
If my ordef is smaller than market depth (for example there are 10 million at the price I order in the market and I trade only 1 million) and order is executed immediately such as 5ms, there will be no alippage. My understanding is correct?

In terms of execution latency or slippage, APi and MT5 is equivalent? If so, switching to MT5 seems convenient in terms of the cost of learning a new language and the convenience of using a sophisticated platform.

Just concern i heard MT5 cannot control each order individually.

Hello,
That is correct. Just to add also that slippage described above have also another name: Price Adjustment. This means when you send order with 10 mil, it will be executed, but with few orders instead of 1 and on different prices and you will get the average price of execution. This is happened, because you see the best ask and best bid available, but this is just top of book, not the depth of market and usually, the top of book is tinny (some brokers may guarantee you certain level in average), but in fact no one will offer you tight spread with solid market depth.
If you intend really to trade large orders, would be better to contact them to create for you flow that will meet your order sizes. The flow will have a bit larger spread, but in this case the real slippage / price adjustment will not be away from what you see and your orders will be filled in milliseconds.
You must be connected to Myfxbook in order to leave a comment
*Коммерческое использование и спам не допускаются и могут привести к аннулированию аккаунта.
Совет: Размещенные изображения или ссылки на Youtube автоматически вставляются в ваше сообщение!
Совет: введите знак @ для автоматического заполнения имени пользователя, участвующего в этом обсуждении.