Quantitative analysis is a technique that sets out to better understand behavior. To help accomplish this, the method utilizes mathematical and statistical modeling, measurement, and research. The goal of quantitative analysts is to properly represent a given reality in regards to a numerical value. Quantitative analysis is useful for an array of tasks, including measurement, performance evaluation or valuation of a financial instrument. Moreover, it helps in predicting real-world events. Today, we take a look specifically at quantitative finance strategies that may be helpful in the cryptocurrency realm.
Drawing from this type of analysis is an extremely complex area of quant finance, which is ‘quantitative trading’. This form of trading consists of trading strategies that derive from quantitative analysis. Two of the most common data inputs in this particular type of trading are price and volume. They are the primary inputs when it comes to mathematical models.
Financial institutions and hedge funds frequently utilize quantitative trading. Because of this, the transactions are typically quite large. What’s more, they sometimes involve the purchase and sale of an abundance of shares and other securities. Be that as it may, quantitative trading is gradually becoming a more prevalent technique that individual investors use.
Over the years, financial securities are gradually becoming more complex. In accordance with this change, there is a notable demand that is steadily growing. There is a dire need for people who fully understand the complex mathematical models that price these securities. On top of that, people who have the ability to enhance them in order to generate profits and diminish risk. These highly sought-after individuals are ‘quantitative analysts’, or simply ‘quants’.
Working in quantitative analysis and trading requires managing a blend of mathematics, finance, and computer skills. Due to the obvious challenging nature, it is no wonder that there is a great demand for quants. Moreover, it is unsurprising that they have remarkably high salaries.
There are two types of quants: those who are ‘front office’ and those who are ‘back office’. Quants that work directly with traders, providing them with pricing or trading tools, are often of the ‘front office’ variety. In the ‘back office’, however, quants will validate the models and conduct research. Moreover, they are the ones responsible for creating new strategies.
For banks and insurance companies, the focus of the work is mostly on risk management rather than trading strategies. In terms of stress and demand, front office positions are typically more under pressure. However, they do receive comparatively better compensation.
What exactly drives this high demand for quant trading? Well, the answer derives from the following trends:
- The rapid growth of hedge funds, as well as trading systems that are automated.
- There is an increasing sense of complexity pertaining to both liquid and illiquid securities.
- A prominent need to give traders, accountants, and sales reps easy access to both pricing and risk models.
- The continuous search for market-neutral investment strategies.
Quantitative finance strategies: Breaking it down
Quantitative traders take full advantage of a lot of factors for the sake of their trade. These include mathematics, modern technology, and the basic availability of comprehensive databases. With these components, they are capable of making logical trading decisions.
Quantitative traders take a trading technique and construct a model of it by employing the use of mathematics. Afterward, they develop a computer program that implements the model to historical market data. The model then undergoes backtesting before inevitable optimization. Should it successfully achieve favorable results, then the system is then put into action in real-time markets with real capital.
The best way to describe how quantitative trading models function is by using a basic analogy. Imagine a weather report in which the meteorologist forecasts an 80% chance of rain. The catch here is that the sun is shining as they make this prediction. The meteorologist acquires this seemingly nonsensical conclusion by collecting and analyzing climate data from various sensors all throughout the area.
When it comes to quantitative trading systems, each one consists of four vital components to make it work:
- Strategy Identification – This pertains to finding a strategy and then exploiting an edge before ultimately deciding on the trading frequency.
- Strategy Backtesting – The focus of this is on obtaining data. Moreover, it analyzes the overall performance of the strategy and then later removes any potential biases.
- Execution System – This involves linking to a brokerage and automating the trading before eventually minimizing the costs of the transaction.
- Risk Management – The focus of this is mainly on optimal capital allocation, “bet size”/Kelly criterion, and the psychology of trading.
Identifying a strategy
All quantitative trading processes start with an initial period focusing entirely on research. This research process embodies the task of finding a strategy. Furthermore, it determines whether the strategy has a place in a portfolio of other strategies you may be running. During this period, you obtain any data that will be necessary for testing the strategy. It will allow you to try and optimize the strategy for higher returns and/or lower risk. For this, you will need to take into account your own capital requirements. This is especially crucial if running the strategy as a “retail” trader, as well as any transaction costs, will affect the strategy.
Contrary to popular belief, finding lucrative strategies by way of various public sources is actually quite straightforward. Academics frequently publish theoretical trading results, though they are mostly the gross of the cost of transactions. Quantitative finance blogs typically talk about strategies in detail. Trade journals often outline some of the strategies that fund regularly employ for their operations.
Many might inquire about why individuals and firms are adamant about discussing their profitable strategies. Especially when it pertains to them knowing that others “crowding the trade” may halt the strategy from working in the long-term. The reason for this comes from the fact that they don’t typically discuss the parameters and tuning methods they use. These optimizations are a key part of transforming an otherwise unremarkable strategy into one that is highly lucrative.
The main intent of backtesting is to supply evidence that the strategy that the above process identifies is profitable. Specifically, when it is applied to both historical and out-of-sample data. This establishes the expectation concerning how the strategy will perform when it is working in the real world.
It is important to note that backtesting does not necessarily guarantee success. There are a variety of reasons for this, but the bottom line is it doesn’t automatically equate to favorable results. It is arguably the most subtle area within the field of quantitative trading. This is because it encompasses an array of biases. All of these prejudices need to be taken into consideration and should undergo removal as much as possible.
As soon as a strategy is identified, it is essential to acquire historical data. Through this data, you can execute testing and – if the situation calls for it – refinement. There are a substantial number of available data vendors that exist across all asset classes. The general costs of using them typically scale depending on the quality, depth, and opportune of the data.
The conventional starting point for novice quant traders (at least at the retail level) is to use the free data set from Yahoo Finance. Whoever the providers do not matter for the purposes of this article. Instead, we will go over some general issues that come with handling sets of historical data.
Historical data concerns
The leading concerns with historical data include accuracy, survivorship bias, and modification for corporate actions. Such actions include the likes of dividends and stock splits.
Accuracy relates to the general quality of the data. Overall, it’s about checking for any notable errors that could be present. Oftentimes, errors can be simple to identify. This is especially true when using a spike filter, which pinpoints incorrect ‘spikes’ within time series data and corrects them. Other times, however, they can be incredibly tricky to spot. More often than not, it’s necessary to have two or more providers before checking their data against each other.
Time and again, survivorship bias is a noteworthy feature of datasets that are either free or cheap. A dataset retaining survivorship bias essentially means that it does not consist of any assets that are no longer trading. In the context of equities, this basically means stocks that are delisted/bankrupt. This bias signifies that any stock trading strategy that was subject to testing on such a dataset will probably perform better. Especially when you compare it to the “real world” since there was already a preselection of historical “winners.”
Corporate actions typically include ‘logistical’ activities. These activities are executed by the company that usually causes a step-function change in the raw price. This is something that need not be included in the calculation of returns of the price. Stock split and dividend adjustments are both recurring culprits. A process going by the name of ‘back adjustment’ is absolutely necessary for each one of these actions. It is imperative for one to be careful so they don’t confuse a stock split with a true returns adjustment. There are a lot of traders that wind up getting caught out by corporate action.
An ‘execution’ system is the means by which the broker sends and executes the trades list that the strategy generates. Regardless of the trade generation being semi or fully automated, it is entirely plausible for the execution mechanism to be manual. Likewise, it could potentially be semi-manual (i.e. one click) or fully automated.
For low-frequency trading strategies (those that hold assets longer than a trading day), manual and semi-manual techniques are quite common. For high-frequency trading strategies (those that hold assets intraday), it is essential to create a fully automated execution mechanism. This one often will undergo tight coupling with the trade generator due to the prominent linkage of strategy and technology.
The key considerations to keep in mind when constructing an execution system are the following:
- The interface pertaining to the brokerage
- Minimization of transaction costs, which includes commission, slippage, and the spread
- Difference(s) between the performance of the live system and performances that have been backtested
A significant issue for execution systems centers on the notable divergence between strategy performance and backtested performance. Something like this could occur for a wide variety of reasons. Right off the bat, one can easily point the finger at bias and optimization bias, especially when considering backtests. However, there are some strategies that fail to make it easy to test for these biases prior to deployment.
Another component that aids in building quantitative trading is the process pertaining to risk management. Generally speaking, ‘risk’ consists of an array of biases, some of which we already covered. One of the primary ones is technology risk. Such an incident includes servers sharing a location at the exchange experiencing a hard disk malfunction out of the blue. An additional risk is brokerage risk. An example of this would be the broker becoming bankrupt. This is not entirely outlandish, especially when you take the MF Global scare into account.
Put simply, risk management covers almost everything that could potentially cause interference with the trading implementation. When it comes to this, there are an array of sources. There a plentiful amount of books that devote themselves to risk management for quantitative strategies. To explain them in great detail would take a lot of time, so doing your own research will prove to be beneficial.
Bringing crypto into the mix
When it comes to risk management, ratios will often correlate volatility with risk. Such a mindset is not necessarily unreasonable for investments such as bonds. However, it may prove to be too narrow when you apply it to alternative assets such as cryptocurrencies. In lieu of this, a considerably more useful measure of volatility tends to be the correlation between multiple cryptocurrencies. This can often be between Bitcoin and Ethereum, Bitcoin and altcoins, and Ethereum and altcoins.
Now, in the context of a bull market, everything suddenly goes up. When Bitcoin is starting to rise extensively, altcoins will usually outperform Bitcoin in terms of percentage. During exceptionally shorter time horizons – sometimes in minutes – altcoins historically have more volatility and greater betas.
In a bear market, altcoins will still continue to trade at a higher beta, but to the downside. Back in 2018, we saw Bitcoin drop by roughly 80%, whereas altcoins’ drop was by 95%. Despite cryptocurrency market movements as a whole, the parallels between Bitcoin/Ethereum and altcoins continue to be incredibly strong. Be that as it may, altcoins proceed to possess a much higher beta. This is something that is completely different from what you would anticipate in the financial markets.
Knowledge about these parallels is generally transparent. What’s more, it is possible to algorithmically and programmatically trade them for steady and consistent returns.
In keeping with tying the topics of quantitative finance strategies and cryptocurrency together, a study was conducted by Nicolas Rabener to connect the two. His analysis covers five strategies: size, momentum, low volatility, mean-reversion, and short-term momentum. According to him:
“The quantitative strategies are created by constructing long-short portfolios of the top and bottom 30% of the cryptocurrency universe.”
Strategy #1 – Size factor
The first quantitative strategy Rabener put to the test was the ‘size factor’. This is a well-established factor in the world of equity markets. The long portfolio contains cryptocurrencies that have small market capitalization and the short portfolio cryptocurrencies have large market capitalization. Drawing from his findings, the observation proves that the returns were positive overall. However, there was also a substantial amount of volatility.
Strategy #2 – Momentum factor
The next strategy that was analyzed was the ‘momentum factor’. This purchases the winning cryptocurrencies and shorts the losing cryptocurrencies. The measurement covers the previous year and excludes the previous month. The outcome of the factor was the generation of highly unattractive results, some of which includes two drawdowns.
A crash in January of 2017 was the result of a short position in Litecoin. It had the lowest performance over the previous year at that point in time. However, in two weeks’ time, it came close to tripling. Quantitative finance strategies help in the avoidance of such single stock or currency risks by way of applying diversification across numerous assets. Though, the limit as to how many cryptocurrencies are currently available is hindering this.
Strategy #3 – Low volatility factor
The creation of the ‘low volatility factor’ is from purchasing cryptocurrencies that have low volatility. Moreover, it comes from shorting cryptocurrencies that have high volatility. The fundamental theory is that low-risk assets typically surpass high-risk assets on a risk-adjusted basis. This is mainly because of behavioral biases stemming from those who are participating in the market.
Strategy #4 – Mean-reversion
‘Mean-reversion’ is a short-term trading strategy that comes from buying the losers from the previous week. Likewise, it comes from shorting the winners of the previous week. The basic theory behind this is that investors often overreact and traders receive compensation by providing liquidity in these situations. Therefore, mean-reversion benefits from high volatility as investors make more mistakes.
However, based on Rabener’s findings, this strategy does not exactly perform well. This is regardless of the high volatility of cryptocurrencies, which, in hindsight, is quite perplexing.
Strategy #5 – Short-term momentum
Seeing as how the returns for mean-reversion were persistently negative, the opposite strategy may appear to be more appealing. This opposite strategy is ‘short-term momentum’, which is, of course, the opposite of mean-reversion. The creation comes from buying last week’s winners and shorting last week’s losers.
This quantitative finance strategy would have likely led to the generation of strong returns during the previous years. There is something important to keep in mind, though. The S&P 500 was showing attractive returns for a short-term momentum strategy for decades up until the early 1990s. However, there were negative returns from that time forward.
Short-term momentum appears to be enticing prospect within the crypto space. Be that as it may, mean-reversion dominates a majority of financial markets on a short-term horizon.