Algorithmic execution is prevalent in almost every instance of institutional risk-taking. More quantitatively-focused investors such as systematic hedge funds, market makers, and high-frequency traders develop proprietary execution algorithms explicitly designed to serve their investment strategies. Other investors, such as pension funds, mutual funds, and discretionary hedge funds utilize trade execution algorithms made available, for example, through investment bank trading desks.
The reasons algorithmic execution has become best practice in traditional markets are many, starting with the aim of entering and exiting a risk position without moving the market against you. In trading terms, you want to minimize market impact and slippage. Furthermore, you want to enter or exit your position without letting the market know what you are up to.
By splitting orders and trading them in a random pattern, you can trade ‘undetected’ and avoid getting penalised by your own ‘wake’. You want to achieve this as efficiently as possible, meaning that you want to trade more when liquidity allows for it. This sums up to the goal of wanting to ensure the best possible total price for your fills, with minimal market impact, over the life of your trade.
A well-calibrated algorithm will always produce superior results when optimising across these dimensions than a human, saving both time and money for the investor.
All trades are not equal
It is important to recognise that a trade can have different paths depending on the investor’s objective, making the calibration of trading algorithms paramount. For example, executing a trade as fast as possible does not necessarily provide the best outcome when it comes to investing. Admittedly, if you are looking to capitalise on short-term price movements, then aggressive execution might be needed, and you will pay appropriately for that (i.e., allow for more market impact). However, if you believe that your profits will develop over time, or if you are looking to enter or exit a longer-term position, it might make sense to take a less aggressive trading approach.
One way of illustrating this is that throughout a day or a week it is somewhat random as to whether the price of an asset will move up or down. You can use the current variance, or volatility of the asset to approximate the level of price change, but when it comes to direction, your predictions are usually not much better than a coin flip. The only thing you can be sure of is that market impact and slippage will adversely affect your fill price, so it makes sense to minimise these over the life of the trade.
One peculiar trait of human traders is that we tend to execute fast if the price goes in our favour, but if the price goes against us, we tend to execute slower as we hope that the price will come back to our target entry or exit point. Even the most disciplined traders can fall victim to this bias (our beloved crypto HODLers come to mind). Over time this creates the undesired trading profile of letting losses run and taking profits early. This investor behaviour is relatively easy to mitigate with algorithmic execution, while also fully incorporating the current and historical state of the order book and fill prices.
These are all sound reasons to use an algorithmic approach to trade execution for traditional markets, as we will see in the remainder of this article there are even more reasons to use it for digital assets.
Algorithmic Trading is uniquely suited for the Digital Asset landscape
Liquidity in Digital Assets is Highly Fragmented — in using only one crypto exchange, it is not possible to access even half of the market’s depth. When executing one relatively small order, this might not translate into many dollars lost, yet over time, or in executing a larger order, the costs paid to the market can add up to large amounts.
Market Depth is Opaque — the unfortunate reality in the digital asset markets is that the majority of volumes reported are fake. Despite seeing that a given exchange has $20mm of reported volume per day for a particular altcoin, time and time again we have found that the real volumes traded are sub $1mm, for example. Unfortunately, reported volume figures have become a form of PR, as having more volume not only implies that your exchange is more popular, but also that it offers better liquidity than its competitors. The obvious solution might be to stay away from the exchanges with fake volumes, but this is complicated as some of these exchanges still have some of the best authentic liquidity available. Algorithmic trading improves visibility on real volumes by, on one hand, deriving knowledge from the actual order book and not from reported volume and, on the other hand, controlling for the level of wash trading observed in given liquidity pools. In some situations it’s possible to interact with the wash traders and trade against them, as their algorithms can be very predictable and unsophisticated.
Dealing with Low-Quality Exchanges — even for semi-institutional traders that opt to be present on multiple exchanges to amplify their liquidity, there are a variety of risks to face. By leaving crypto or fiat on exchanges, there is a substantial risk that funds will be stuck or stolen. Even the highest quality exchanges do not like to see dollars leave, so it is common to experience delays and other frictions with withdrawals. These episodes are especially scary when it takes weeks to get a response from customer service. These exchange-related risks are minimized when executing programmatically through a market maker.
Algorithmic Trading versus OTC brokered trades
One way to mitigate exchange risks and trade large volumes is to find a broker to execute your trade directly with another counterpart. While this is common practice, it usually involves paying significant fees to all parties involved, at times more than 5%, and it often takes a lot of effort and time to source the other side. For example, consider trying to sell $50mm BTC — even if a broker can find a buyer and close the deal in three days, that leaves plenty of time for the market to drop significantly. With programmatic execution, however, it is far more cost efficient — there is no reason to pay several percentage points away from the market, and execution begins immediately. There is no waiting, which significantly reduces unnecessary market risk.
A more customized and dynamic service model
Algorithmic trading is an excellent fit for a large holder of a digital asset who wants to minimize the impact on the price of the asset when liquidating part of their position. For example, it is possible to build in parameters to automatically sell less when there is not a robust bid on a particular day. It is also possible to only sell during a specific window each day or around market-specific events. Any time there is a predetermined set of instructions, it can be programmed ahead of time, providing a variety of efficiency gains. The same applies for an asset manager or fund manager attempting to enter into a large crypto position; they can lower their average cost of crypto by executing algorithmically, thus maximizing potential gains.
More anonymity — if you want to keep a flow quiet, going through a reputable market maker minimises the risk of market participants finding out than if you contact a broker. Brokers, after all, are interested in talking to everyone and carry a set of their own motivations to the table. Working directly with a market maker, the KYC process only has to be done once, and the flow can be dispersed across dozens of exchanges, making it very difficult to detect the origination.
Reduced fees — by trading with a reputable market maker, there are fewer exchange fees paid. Because traders can negotiate volume discounts or rebates with trading venues, market makers end up paying much lower fees than regular market participants. This means that more often than not, you can get high-touch customized liquidy from a market maker at a lower fee than you would for executing the trade yourself.
Transparency — trading directly with a reputable market maker makes it possible to see exactly where all trades are occurring and at what levels. This fair and transparent model is very different from brokered trades or even on exchanges, where unknown or hidden fees are pervasive.
Reduce the possibility of human error — another benefit of algorithmic trading is that it eradicates tendency for human error. No more ‘fat fingers’ or accidents which could have a devastating impact on returns in a thin order book. Additionally, it removes any emotion from trading — a critical characteristic of successful traders — and leads to less stress and more focus on your business’s core competitive edge. When executing via algorithms, instructions are pre-programmed, creating far more predictable and stable results.