Broker to Trading

The brokers in the comparison table all offer ECN trading and platforms which support the use of automated trading strategies and are accessible to traders. They also potentially allow the trader to scale up their algorithms, with accounts specifically designed for high volume trading. ECN brokers provide access to infrastructure which allows for high speed order execution with no dealing desk intervention, thus enabling the trader to use trading strategies which may not be possible elsewhere.

What algorithmic trading fundamentally is about can be expressed in the word 'algorithmic'. Algorithms are sets of rules. They are used to program computers, but can also be used to define any action to be executed whether human or machine. So an algorithmic approach is one which is based on rules. But there is a difference in the way a human trader can and may apply rules and the way a computer program will execute rules.

Arguably all trading is based on rules, unless the trader is randomly making trading decisions. Traders may follow an indicator, which is to follow a set of rules. Human traders though may have a heuristic approach to trading. That is they follow rules in a way such that they can change them or ignore them, in effect using rules of thumb, rather than robotically following rules. This can be a strength of human trading, using experience and intuition to help make decisions. However the sheer intensity of trading can expose such an approach to human weakness, tiredness, emotional responses, fear, greed and so on. Arguably a trading market is not a rule based system and is suited to a more creative approach. However the reality of trading may lead the trader into considering allowing a computer program to trade on their behalf. A computer program does not tire and will execute a set of rules (or algorithms) until told to stop which may bring advantages in terms of speed of execution and accuracy. This strength of a computer program is also a weakness as it does not allow for human discretion. It is possible to use artificial intelligence (AI) to try and let a program learn via machine learning, but currently there is (presumably) no computer program which can equal the human brain in its capacity to reason creatively, intuitively and make jumps to correct solutions. However there are some kinds of trading which only a computer can feasibly do and rely on algorithmic trading executed by computer programs (also known as algos). A big caveat is the emergence of LLMs, which have limitations but can provide a way for the human trader to interact with a computer program with access to a huge database and the ability to reason and sound human-like.

When trading one sometimes thinks that it would be nice to be able to be both intuitive and share the tireless precise capacity of a computer program. AI is not really about trying to produce a thinking machine per se (except in research), rather it is about doing things which a human mind cannot do (such as look for patterns in large data sets). It may be seen why AI then can be used in algorithmic trading, as an aim of this is to trade in ways a human cannot.

The sometimes harsh reality of trading can wear the trader down to the point where any intuitive abilities dissipate into near random actions governed by (presumably) older parts of the brain. One approach is to accept this and look out for warning signs. Another is to use algorithmic trading via computer programs as a tool, using strategies honed in and understood from discretionary trading, where the trader makes the decision to execute a trade, even if following a set of rules.

If a trader simply relies on computer programs they can find that they experience accumulating losses, known as drawdown. They may find that they need to adjust strategies and then still experience drawdown. Conversely, they may find that drawdown turns into a positive gain over time, but they have to be able to withstand it. Traders can backtest strategies on the trading platform, which is to say the algorithm can be applied to past market data, and optimise them. However, for a range of reasons, because a strategy performed well in the past does not mean it will do so when applied to the market in real time.

So one way to deal with all this is to step back and look at algorithms as simply that, rules and use them as either human trading rules (or guides) or as a computer program, letting the market and the trader determine when to use either approach.

The brokers in the table all support trading using computer programs. They are ECN brokers and do not restrict trading styles. As ECN brokers they may offer very rapid low latency order processing, making feasible algorithmic trading strategies which rely on speed and only a computer program can feasibly do.

What is feasible is an important consideration. Technically it may be possible for a trader to execute a strategy requiring rapid orders and make trading decisions at their own discretion. But practically this may be impossible because the trader cannot be at their computer or because the speed required rapidly exhausts the trader. It may simply not be possible for a human trader to apply some strategies (e.g. high frequency trading). But running through a strategy to the extent this is possible as a trader may potentially help them hone it.

As for the automated programs, the trading platforms in the broker comparison table support trading using a computer program executing rules (i.e. robots). MT4 and MT5 have Expert Advisors, which are automated trading strategies. cTrader allows cBots to be run. JForex lets the trader run scripts to execute algorithms.

As for discretionary trading, all these platforms support this style. JForex for example has a particularly large set of technical indicators inbuilt on the platform (though the other platforms also let the trader add more or develop them). So the trader can develop their own rules and apply them for use in different market conditions, or use rules developed by others and test them. Since ECN brokers have basic similarities, as they aim to let the trader have access to liquidity through the broker, as to which is the best broker for algorithmic trading may come down to such factors as the online trading platform or platforms offered.

So reasons for using algorithms are to let the program trade on behalf of the trader, to make use of the advantage a computer program has in terms of speed and accuracy and to use strategies which are native to modern high speed electronic markets and cannot be executed by a human trader.

It is important to note that because an algorithm worked on a past set of market data, does not mean it will work in the future. The human trader will understand this as they find their strategies stop working. They may wonder if this is because they are human and cannot trade like a machine. But when they use or code and then use their strategies, may discover that the computer program has other weaknesses as well as strengths.

One point to consider is the point made earlier in this article that the market is effectively not a rule based system, though it is composed of rule based strategies applied to it. The sheer range of inputs both congruent and conflicting (market data, strategies, computer programs and so on) creates a random market character, but perhaps with some residual order forming and reforming, which automated trading strategies may try and exploit (and some aim to do so specifically using the advantage of speed and accuracy). To deal with this may need the human brain, it is just that the human brain needs algorithms as well and sometimes has to rely on them via a computer program.

One problem to try and tackle with algorithms is making them too complicated, trying to capture seeming will-o'-the-wisps of order in the market (even though it may exist from time to indeterminate time) or to try and emulate the trader's own discretion. So the trader may need to exercise discretion with the rules themselves and perhaps apply ideas of simplicity, elegance and parsimony. Heavy rules sets become unwieldy and can have the opposite of the intended effect. This is true for human traders trading by rules and for computer programs, though arguably the human trader can handle more complex rules, especially if made flexible and adaptive by heuristics and creativity.

It is arguably better not to treat any rule based system (including an indicator) as a black box and blindly apply it, but to some extent that is what using a computer program can lead to, indicating why there are weaknesses in this approach to trading, even if smartened by backtesting and adjustment. So one way to tackle this is to understand as much as is possible what the computer program is doing and have a market based rationale of why and when to use it.

Beginner traders may wish to start with a broker offering an intuitive web trading platform and develop their strategies this way, before trying an advanced approach such as automated trading using computer programs. More advanced trader may wish to use either a discretionary or an automated approach or both. The human trader may find issues in scaling their discretionary approach. However algorithmic trading provides a route where the trader can apply a strategy at increased scale, that is to execute it at a frequency that they would find difficult or impossible to do, with the caveat that this can magnify losses that a discretionary approach (if it is possible) might be able to alleviate or step back from.

To trade strategies 24/7 and not face connectivity issues at their end, the trader may wish to rent a Virtual Private Server (VPS) which hosts the programs. Some of the brokers in the table offer accounts which may offer free VPS hosting for larger trading accounts. The brokers in the table may allow third party platform use as well as access to their FXI API (this allows the trader to directly access the market via the broker from their own front end).

A trader following rules determined by an indicator is executing an algorithm. But the distinction is that they do so at their own discretion. A computer program following this same set of rules is executing the algorithm exactly, though the trader may add their discretion by adjusting the program. A computer program which is using AI is also executing an algorithm. The use of AI and model based methods may seem different from using a set of rules based on an indicator, such as RSI. However, more complex approaches can be required by the kind of trading which is being applied to the market. ECN brokers are typically able to provide accounts which support high volume trading and algorithmic strategies which may rely on high speed order execution.

Best Broker For Algorithmic Trading

All of the brokers in the table offer infrastructure for rapid order processing with no dealing desk intervention. However since a pick needs to be made, the choice is IC Markets. This is because IC Markets focuses on automated trading, scalpers and high frequency traders (though other traders are also welcome).

IC Markets places its trading servers in data centers and connects with providers either there or through the data center. Since algorithmic traders will typically be using a VPS, IC Markets says its servers can connect with major VPS providers either in the center or through the center. IC Markets says that latency is under 1ms to VPS providers (the average order execution speed is under 40ms once the order is received by IC Markets).

Additionally IC Markets offers complimentary VPS hosting for traders who trade over 15 lots per month round trip for Forex and Metals, T&Cs apply. Spreads can be as low as 0 for EUR/USD with a competitive commission charge of $3.5 per lot per side. Thus given the argument in the article above, that algorithmic trading seeks to trade in ways the human cannot, then this broker can be seen as supporting (and indeed explicitly says it does) styles such as high frequency trading which rely on rapid order processing. IC Markets offers MT4, its successor MT5, TradingView and cTrader. The minimum deposit is $200 and the trader can add on different accounts once they have signed up.