Broker to Trading

When placing a trade from their online trading platform, the trader may notice the time lag taken for an order to be processed (or be impressed by the lack of lag). This is the time taken for the order to be sent from the terminal to the broker and back. As might be expected there are a number of steps along the way, and optimising these steps can result in faster order processing. It is also possible to cut out some steps, such as by using a VPS.

The CFD order processing process

When a trader placed a buy or sell order on their trading terminal, the order is sent to the broker who processes it and then sends a confirmation back. This can take from milliseconds to seconds. A CFD is a Contract for Difference, which is to say it is a contract entered into by the trader with the CFD provider (broker) to exchange the difference between the value of the contract between the beginning and end of the trade. The value of the CFD (which is a derivative instrument) is tied to the value of the underlying asset being traded. There are various mechanisms to keep the CFD value and the asset value aligned.

When buying or selling a CFD, the trader does now own any of the underlying asset. What this means is that the trader when placing an order to buy an asset for example, the order processing does not involve the trader buying the underlying asset. However the CFD provider can choose to hedge this order by creating an opposite sell order in the external asset market. This might be for the asset itself or a derivative (e.g. futures or ETF) based on the asset.

The broker may choose to match the trader's CFD with another CFD at the broker (as there are not external pools of CFD orders), which is to say match a buy with a sell and vice versa. This is done with an automated process, so it can happen fast. There are also checks taken on orders such as margin requirements, but again this is normally a rapid automated process.

Trading servers

The trading servers take care of the order processing process. Some brokers explicitly cater to those who want to run higher frequency trading strategies. Examples of these are scalping and high frequency trading itself. As trading frequency gets higher it becomes harder for the human trader to execute these strategies themselves. So traders may turn to automation, i.e. online trading robots.

While strategies employed by automated systems will not necessarily be shorter frequency, they have a tendency to be used this way, which is why automated trading brokers tend to seek rapid order transmission. Trading servers can be hosted in different ways, including in house and on the cloud. However for rapid order transmission some brokers will host their servers at data centers dedicated to finance, such as the Equinix NY data center. This itself is a speed optimisation, as these centers operate at scale and can use very efficient methods and technologies to route orders.

Simple proximity becomes a factor in order processing speed, that is proximity between the broker's servers and liquidity providers and institutions. Optical cabling is typically used for optimal processing speeds within the center, as well as efficient routing algorithms and technology. So this is why at those brokers which cater to higher frequency trading, traders may find descriptions about the use of data centers and how this can optimise order processing.

Data transmission

In data centers optical cabling is used due to its superior speed of data transmission. Some data centers may leverage microwave transmission as this can be faster, particularly at critical communication points. There is also the potential for faster and more secure methods, including more advanced optical cabling to reduce latency, laser networks and faster transmission methods than microwave.

The universe has a built in speed limit for the transmission of signals (the speed of light in a vacuum) so the aim is improvements in materials, algorithms and technology to reduce latency up to this speed limit as well as reducing errors in transmission.

While the built in limit on signal transmission makes it unlikely that advanced methods such as quantum networks which aim to leverage the phenomenon of entanglement would result in speed improvements beyond this limit, there may potentially be applications in enhancing network security (which is also a factor in order transmission) as well as improving the computational efficiency underlying this technology.

VPS - Virtual Private Servers

While using a broker with rapid order transmission can result in speed on the order of tens of milliseconds, to reduce this even further, the trader may wish to consider a VPS (a Virtual Private Server). A VPS is a type of software which allows for separate programs to be run on shared hardware resources. So instead of having one server for one user, multiple users can run their own programs on one server, which is to say they utilise economies of scale.

In the use case of trading, traders can rent a VPS and have their trading algorithms execute on a server. Thus the bottleneck of sending an order from the trader's terminal to a broker and back again can be ameliorated (as if the trader were executing orders in the data center, assuming this is the model used by the broker). So with a VPS, potentially greater speeds can be achieved (the actual speed depends on a range of factors). The cost of renting a VPS varies, but sometimes brokers will offers a free VPS for higher volume traders (T&Cs apply).

Dedicated servers

As we have seen, a VPS uses shared resources on a hardware server. However some traders may want the additional security and potential improvements in latency from using a server which does not share resources with other traders' programs. Dedicated servers are normally significantly more expensive to rent than a VPS, however if the trader is interested they can find such an offering at some VPS providers. Generally traders who apply advanced and resource intensive trading strategies such as high frequency trading may find a need for this type of individualised co-location.

Algorithms in routing and AI

As well as hardware, algorithms (specific computer programs) are utilised in rapid order transmission for example in routing, order processing and risk management. There is also the potential to use machine learning, AI and game theory to predict order related outcomes in the market, therefore moving away from a reactive model.

Generative AI, the technology used for smart chat bots, is based on machine learning. Underlying this technology is a range of algorithms, including neural nets. Neural nets can be trained via machine learning on large data sets. These computer programs have the capacity to then make predictions based on their training data. The data is key in this type of AI and using AI to predict in a use case such as order routing would require that the data has to be updated regularly as well as feedback and error correction in addition to supervision by data scientists and those who are expert in the domain knowledge. Game theory is a well developed framework used to also predict outcomes, so it has potential applications in routing.

Bear in mind, the reason for using these kinds of predictive technologies is to attack the speed bottleneck. The core users want speed (as well as accuracy) but there are limitations on the technology, even with co-location and advanced cabling. So to predict rather than react is a way around this bottleneck, but it has its own demands in terms of accuracy and adaptability.

ECN vs market makers and order processing speed

Some brokers classify themselves as ECN brokers. This is contrasted with market marker brokers. Market makers can set bid/ask prices and act as the counterparty to the trade. ECN brokers will go out into the wider ecosystem on the ECN (which connects brokers and other market participants) to find prices, though they may also match orders internally. Some brokers might offer different accounts, some of which might be ECN and others market maker.

The potential advantage of ECN brokers is that there may be improvement in prices available (from a wide range of liquidity providers) and indeed some of the lowest spreads (and low average spreads) can be seen at such brokers. However there is a trade off as this has a greater lag (latency), in principle, than matching internally. Thus such brokers may (but not necessarily will) utilise technology such as co-location in data centers to improve latency in the order process.

What this means is that higher frequency traders are not simply looking for speed improvements, they are more particularly looking for price improvements. As trading frequency goes up, then the cost of the trade can go up as the spread cost is added more frequently.

It should be noted that while ECN brokers may offer very low spreads for liquid markets, there is normally an additional commission charge added to the cost of the trade. But this charge is normally fixed, although it may be volume dependant. So the trader can calculate this as a fixed cost to be added to a normally variable spread, but a spread which may be relatively low compared with other types of brokers. This said, spreads without a commission charge whether at a market maker on an ECN broker have been trending down over the years (for some markets).

Brief summary

Traders who want fast order execution may be those who trade on higher frequencies, either on their own behalf or using robots. But these traders also want (or may primarily want) the lowest cost for their trades, particularly trades of larger size. So there is a trade off, as brokers which have deep liquidity may use processes which are not as fast as matching or filling orders internally. So those brokers which aim to offer fast order processing and low spreads, may turn to technology to increase the speed at which orders are processed. Currently there are various ways to optimise speed of order routing and processing, and there are also various other technologies being explored.

Trade on a demo account

A trader may already know that speed of order execution matters to them. But even if they do and particularly if they do not, testing out brokers and trading platforms is a way to get to grips with this advanced factor in trading, without risking real money.