Quantitative Investing
Quantitative Investing is a rapidly rising investment strategy. Quantitative investing revolves around using systematic and mathematical methods to determine whether to buy or sell products. The advantages quantitative methods bring in terms of efficiencies and tapping into unique opportunities and data, are heavily impactful in making money in financial markets. Hence, we have seen the prominent rise of Quants in Asset Managers and Hedge funds.
Data Science plays a big part in Quantitative roles, coding using R and python (Programming Languages) is largely used to create quantitative models. Datacamp and Udemy are great resources to learn the basics and then go on to apply your data science skills. The beauty of data science is the liberty of its applications.
Data Sets
Funds and AM’s often pay large sums for datasets not easily accessible to the public. This may range from live global weather data to traffic patterns. The rise of alternative data has long been a buzzword but is more prominent now than ever before. Especially given the current rise of robin hood traders, such traders leave data in public spaces whether it be Reddit, YouTube, or other forums. Hence, such information can be very valuable and publicly available. The real skill comes in deciding how to utilise such data.
An example of a common freely available and easily accessible data set is Google Trends. Google Trends counts the number of searches for a word or topic, then displays the counts over time in a graph. Search demand can be used as a loose metric for demand level for products, companies or even industries.
In the chart above you can see the search demand for the word “Tiles” in the UK over a 5-year period. The trend line I added shows that over the past 5 years there has been a gradual increase in the search terms for “Tiles”. However, it is evident to see a clear breach of the trend line between May and July 2020. Suggesting higher than expected demand for tiles in the UK between those months.
Topps Tiles is the UK’s leading tiles specialist, a retailer with multiple stores around the UK. Topps Tiles Q2 2020 (April – June) earnings were up 13% YOY, beating estimates and pushing the share price up 50% on an exceptional earnings catalyst (2). Hence, this displays how using Google trends as a leading indicator, can inform you of exceptional or underwhelming company performances.
Why Location?
When looking at big data, formatting the data to which it represents the sample set you need is one of the largest parts of the process. The location is important as Topps Tiles only operates in the UK. If we were to include the global populous, then the data would not properly reflect Topps Tiles potential consumer base demands.
Why Time?
I chose the 5-year time span to get rid of seasonality biases. You can see every December (Winter), searches for “Tiles” drastically drops. Thus, if you only looked at the yearly time span, such a decline in searches in December may make you negative on earnings, pushing you to short the stock. However, seasonality is expected and thus largely priced in. During Winter, falls in tile sales is expected. Hence, we want to find deviations from the normal seasonality patterns.
Model
Quantitative strategies can allow you to find opportunities that fundamental analysis cannot. I created a simple equity model over the quarantine period looking to take advantage of lockdown trends. Thus, I layout an overview of the steps from inception to profit.
Thesis
To start with you need an idea, hypothesis, or thesis you want to test. An idea I had over the Quarantine period was around customer bought product stocks. I noticed that the isolation conditions created by quarantine made certain products in more demand – I needed a method to find which products and companies were benefitting.
I saw a pattern in SMG a stock that I was researching. On their YouTube page they had tutorial gardening videos. Focusing on videos that were a few years old I saw a pattern in their comment section dates. When these videos were first released years ago there was a lot of comments, and then only recently a resurgence of comments occurred, with a big time gap of no comments or very few. This suggests new customers had started purchasing their products and thus interacting in the comment sections when watching their tutorial videos. Hence, I looked to roll this out with a focus on UK Consumer bought product stocks.
Dataset
A suitable dataset to test your hypothesis must be chosen, or even multiple data sets can be used. The dataset I was using was YouTube, focusing on YouTube’s comment section. To access data from YouTube you can use their API, create a bot or web scrape the data.
An API in generic terms is an interface in which you interact with. Companies like YouTube make their API data accessible to the public, allowing partial access to the information and activities that occur on the website.
A bot is a software application that runs automated tasks over the Internet. Typically, bots perform tasks that are simple and repetitive, much faster than a person could.
Web scraping is a bit messier, in which you have to extract data from a website manually or using web scrapers, generally no publicly available interface is made accessible.
YouTube has a feature in which you can sort comments by date and time. Hence, by looking at the density of comments at different time periods, I could create a model that alerted for patterns in which comment sections had substantial time gaps.
Back-Test
In time series data you back test the strategy on past data, from the results you then can conclude whether the strategy is worth employing going forwards. I ran the back test on a random sample of public listed stock YouTube pages – ones in which customers would interact with.
The results were good, with stocks such as Gears4music, Pets@home, Ultabeauty, Camping world and Scotts Miracle-Gro all alerting. When researching these companies, they all experienced surges in demand, seen through either their earnings or their share price. The model would thus allow me to pick out stocks that were thriving under the quarantine period restrictions.
Run Model
The first company to alert was Angling Direct on the 8th June 2020. Angling Direct is engaged in the retail sale of predominantly fishing gear and camping goods.
Quantamentals is the combination of both quantitative and fundamental principles. It allows you to run models, whilst also fundamentally exploring further the model’s outputs. Employing such a strategy I phone multiple Angling stores and visited the store nearest to me. The main information I was given supported the model’s suggestion:
- Fishing License sales up 40%.
- People who have not fished for 20 years had returned to the industry.
- Family and Kids entering the market – Staycation trend.
- Environmental Agency sold 250,000 more tickets during lockdown.
- Demand in Stores the highest it has ever been.
Furthermore, for confirmation, using other available Big data sources can add value to the process. The main sector Angling Direct is involved in is the UK Fishing business. As the graph clearly shows searches for “Fishing” has enormously surpassed the trend line, hence acting as a confirmation indicator for my Quantitative and fundamental research.
Results
I bought Angling Direct on the 5th July at £49. Angling Direct released a report on the 18th August, stating that they will benefit somewhat from the staycation trend. This caused a 20% push upwards from £50 to £60 in their share price. However, in no way did it fully convey the heavy influx of demand into the industry, evident from my primary fundamental research. This suggested further upside in Angling’s earnings report out on the 14th October (4).
Most of the mediums we use to interact in some frame creates data. So much data is now freely and widely accessible. It is no longer solely about having the money to buy data sets, but encouraging the creativity to find and access alternative data sets. More importantly spurring the ingenuity and creativity to think of how to use data in inventive ways. Such first mover advantages to create models and algorithms using uncommon data, can give immense competitive advantages. Now more than ever it not only pays to be technically skilled, but heavily rewarding to those who are innovative and adroit in how they utilise their skills!
Reference List:
(1) https://trends.google.co.uk/trends/explore?date=today%205-y&geo=GB&q=Tiles
(2) https://uk.finance.yahoo.com/news/coronavirus-stocks-topps-tiles-tui-gvc-europe-ftse-090338350.html
(3) https://trends.google.co.uk/trends/explore?date=today%205-y&geo=GB&q=Fishing