Why Use Algorithmic Trading?

Author: Senior Consultant, Dominic Madden.

Trading can generate vast rewards – which is why it’s such a competitive space. It’s often seen as an institutional activity, and mainly, it still is. Other players, such as Hedge Funds and other funds tend to trade on behalf of clients, whether these are retail or high net worth individuals.

However, the trading activities of these entities are often different. Algorithmic trading can hugely benefit trading boutiques and day traders, who usually trade based on arbitrage opportunities.

Of course, the barriers to entry are still large. It requires lots of capital to open a leveraged trading account with direct market access, but these organisations have the means to participate. Often these traders have the knowledge and ability to recognise the opportunities as well as a method to be in the markets. What they lack is the knowledge and skill to implement an automated strategy.

Algorithmic trading can automatically spot the points in the market to trade and can do this much quicker than a human ever could. It doesn’t make mistakes and use the wrong size or price. This will enable us to ‘beat the market’ and get in and out of trades at the optimum time.

Building the System

If you’re building an Algorithmic trading system for the first time, you’re going to have a lot of questions:

Where to start? What do I need to have in place to be able to create an automated system? What about my risk, and how do I manage unexpected movements in the market? What do I do when the market starts to move against me?

All of these questions can be answered, but a partnership with an experienced organisation who have experience of building these systems is required.

Strategy First

The starting point is to have a strategy in mind. One that can consistently produce desired returns.

If we look at the European Bond markets, the futures markets are highly correlated with predictable behaviours. These are usually quiet until close to the rollovers, where institutional participants need to look at their positions and holdings. These times can provide arbitrage opportunities, which can be exploited quickly and efficiently.

Using European Bond markets as a case study to demonstrate the power of algorithmic trading can show us both the huge potential.

As we look at the markets then there are certain things that we need:

1. A leveraged trading account
2. Market access
3. Live market data

We must assume that these are in place and any missing pieces are the ones where help is needed.

To implement the strategy we need to have processes running. These can be on prem, in the cloud, or even, for simple solutions, running on a laptop.

The first piece of the puzzle is understanding the data needed. We need to look at the market and consider the price, size, side, and the matched amount. With this information we can start to monitor markets and look for signals to indicate there is a potential to trade. So we need to build a component to do this. We’ll call this the Watcher.

Once we have this, we’ll need a component able to trade. We’ll call this the Trader.

Finally, we need to manage the book, watch positions, and raise signals if we need to scratch trades, get another leg on for hedging and then to exit the trade.

The Watcher

In the European Bond Markets example, the Watcher will be looking at the market information coming in from the exchanges. This will be monitored, and opportunities searched for.

The BOBL, BUND, and SCHATZ markets are very well understood and heavily traded. When market movements move away from the expected parameters, then a trade can be placed. These three markets have a very high correlation. For much of the time Eurex Euro-Bund / Euro-Bobl / Euro-Schatz Butterfly is 93% positively correlated with the generic German government bond 2 year/5 year/10 year cash butterfly.

In the example of the European Bond Markets, we’d monitor short term price relationships between the different instruments by consuming the streaming market data. It can monitor the last 30 seconds of data and from this we can calculate the fair value of the instrument.

Then we look at the price movements in the correlated markets. If one moves up a tick and the other moves 4 ticks, we may expect a movement of ½ a tick in a third market. If this isn’t the case, then we can signal that we have a trading opportunity.

Once we have this signal, we can raise an event to place the order.

The Watcher will usually be run as a service, communicating with the other services in the solution.

The Trader

The Trader will place and work the orders. It will scratch the orders when there is no favourable follow on. When there is follow on, a hedge trade would be placed. This forms the second part of the strategy (the entry being first). Finally, the third part of the strategy is to exit the trades. This should return a profit, assuming there was a good edge in the initial entry.

The Trader will be running as a service. In larger enterprise scale solutions this will typically be in the cloud, with all the associated instrumentation and robustness.

User Interface

In addition, a UI can be built to show the positions and give an indication of the state of the markets and orders, as well as having a ‘Kill Switch’ to pull all orders.

Communication

How these components communicate depends on the requirements.

Events raised can generate a simple data object which can be placed on a Message Bus. This message bus is usually part of a larger enterprise infrastructure. A light touch technology like Mass Transit, which is an enterprise scale message bus, can integrate with messaging systems like RabbitMQ. This can be as effective as using AWS or another large cloud solution. Ultimately, clients’ needs will be the driver for these decisions.

What about the future?

Algorithmic trading offers a huge advantage. It’s faster, more accurate, more efficient, decreased costs, and offers back testing capabilities. But that’s just the start. As more of these systems get implemented then vast amounts of data will be accumulated from recording market information.

However, we must balance out the positive message with a word of caution. Using automated systems to enter trades can be fraught with danger. Missing opportunities, the market moving against you, or connection issues can all incur losses. So any strategy must take these into account and have ways of flattening positions.

All this information can then be used to build AI systems, which can learn to identify trading opportunities. Recording as much market data as possible will result in vast datasets of market information, full of rich data. Machine Learning tools could mine this data and be able to identify new strategies. A powerful thought!

For more details on how we can help you with your automated trading, or help leverage what you have, please get in touch.