Using alternative indicators to confirm sentiment change and price breakouts

Alternative data, the term broadly applied to datasets outside of the traditional financial reports and economic indicators, can be seen as a subset of big data used in machine learning and improving algorithms.

Alternative data is also increasingly used in the traditional markets by hedge funds and traders to find an edge in a market saturated in strategies and players. It can be found in many places – data scraped or bought from geotagging, social media, divorce rates, retail spending, and in one famous example satellite images of unsold Tesla cars in its factory car park.

Perhaps the most widely cited example of "alternative data" in cryptocurrency is Google Trends and social media sentiment, although whether it is a leading or lagging indicator is debatable.

When the news broke of JP Morgan coin (around Feb 14) there was no pick up in public interest for bitcoin on Google Trends until late on Feb 17 (NZT).

Around 11:00am on Feb 18, on the 30m BLX (the global price for bitcoin), BTC price broke through resistance around $3,620, at which time Google Trends was still ranging and took a dip. Google Trends didn’t break the 75 ‘resistance’ level until the early hours of February 19, around 1:30am. In this case price led and the interest on Google Trends subsequently followed. The Google indicator showed a topping out and reversal just before price reversed – seemingly leading it down.

In this recent example of fitting narrative to price, the mainstream media attributed bitcoin’s price rally to the JP Morgan news in spite of the 3-day delay in movement for an asset that trades 24/7. If this was new ‘mainstream’ money coming into the system it would most likely be picked up on Google Trends immediately after the news broke as entrants search the news for ways of onboarding to a crypto exchange etc – and price would subsequently follow.

In this case price it appears price moved by endogenous forces, though it is an example of how we could use Google Trends to confirm if there is a change in public sentiment or narrative underway by checking:

1. The speed at which interest follows a significant event
2. Whether that interest leads or lags the price movement (whether price force was exogenous or endogenous)
3. The relative strengths of each breakout in sentiment and price
4. Crosscheck whether the mainstream media narrative is fitting the price curve or vice versa

Alternative data for specific crypto projects

Given that crypto networks behave fundamentally like social networks they can also be valued like them (see NVT and PMR ratios) as their prices are a reflection of investor perceptions and misperceptions about user growth. It would therefore make more sense to look at the secular trends in socio-behavioral data of crowds and society in general.

This is an exploratory phase for crypto valuations and a good opportunity to be creative with data to find correlations. The following data sources may be a good place to start.

Github

Github can be a useful tool for a granular look at a project’s health but with the caveat that it can often represent quantity over quality. A recently published correlation chart between the price of EOS, for example, and the commits of the project’s lead developer Dan Larimer shows a tandem rise and fall leading up to and just after its launch in May 2018. However, it is difficult to say which was the leading indicator – price or commits.

This shows the frequency and number of Github commits to the EOS project by creator Dan Larimer ramping up and falling with the price of EOS. Source: Reddit

Gauging social and news sentiment

Reddit, Twitter and other forum sentiment is the most common form of alternative data in crypto and there are several tools which parse the content through natural language processing (NLP) algorithms to gauge sentiment and semantics. There are several crypto vendors selling such tools though, with some coding proficiency, one could build their own using the free Python NLP libraries and toolkits available online and feed it data from the Twitter or Reddit API.

When assessing sentiment on crypto forums however one should be aware of an echo chamber effect and the extreme partisanship that followers of projects are prone to which can generate a lot of noise and false signals. If trying to pick a cyclical change in direction for the market one should be focused on new money coming in and therefore monitoring mainstream news and official announcements (government, central banks etc) for a change in sentiment.

News sentiment, filtered by positive or negative headlines, can be easily accessed by plugging into a crypto news aggregating API like Cryptocurrency Newsfeed and incorporated into a trading strategy such that certain keywords and events carry a higher probability of affecting the price positively or negatively.

The dashboard of Cryptocurrency Newsfeed shows the sentiment behind crypto headlines and events and its API can be fed into natural language processing models.

With NLP and machine learning the impact of recurrent crypto events and official announcements will be predicted with accuracy similar to the traditional currency markets where scheduled economic announcements, such as FOMC meetings and non-farm payroll data, are rated by low, medium or high impacts.

The economic announcement calendar on Forex Factory which rates the impact of events from low to high by color, grey to red.

Measuring the impact of events on prices is still a work in progress for crypto sentiment tools to assign probability to repeat events having an impact on prices.

Event analysis on Coinmarketcal shows the incidence of airdrops, announcements and other significant events for projects.

Coinmarketcal relies on discretionary input and upvotes from users to attest to the veracity and impact of certain events taking place so is susceptible to human error and bias.

Meetup groups

The number of Meetup groups and members for a project can be used as another gauge of community interest but, again, sorting for group and member activity is difficult and can end up being a case of quantity over quality. For example, Ethereum has different levels of meetups: General meetups – a broad intro from beginners to investors; ETHDev – meetups for Ethereum developers that involve solidity coding and hackathons; ETH niche – for specialized Ethereum developers.

The number of Ethereum Meetup groups globally from Feb 2016 – November 2018. Data reference

Ethereum has a vibrant community of developers, enthusiasts and curiosos. The strength of interest in it as a technology and not just a speculative asset can be seen in the increase of Ethereum meetup groups which have continued to grow steadily – even after the price bubble burst in January 2018 and the subsequent year-long bear market.

The number of Ethereum Meetup group members globally February 2016 – November 2018. Data reference

Today there are 1,197,658 members across 3,872 groups globally. In Australia alone – not exactly renowned as a technology hub – there are 114 Meetup groups and over 39,000 members.

Source: Meetup.com

Looking for the mainstream change of narrative

Stories or narratives are used as frameworks to help us process the information that flows from an account of events, experiences, or phenomena, whether true or fictitious.

Investment narratives create feedback loops that over inflate and deflate bubbles, a byproduct of the inefficiency in markets. A prominent example of asset prices overreacting to new information is Robert Shiller’s "irrational exuberance" in real estate which drives a fear of missing out and pushes prices higher.

The crypto industry has been described as a perfect environment for narrative-fuelled price bubbles with a:

1. Lack of enough relevant historical data to form valuations
2. Conditions that attract retail investors
3. Oftentimes poor regulation
4. Relative strength of narratives to grab attention in an opportunity rich investment environment.

Cryptocurrencies have been through several narrative stages with sub-narratives among competing projects within each stage: 1) the first was bitcoin’s and the early crypto-anarchist tenets of private autonomous currencies in opposition to central banks and the financial status quo 2) the ICO and utility token craze and ‘decentralization’ of everything; 3) the ‘tokenization’ of everything from traditional equities to real estate, which has been much hyped in the accredited investor corner of the industry but has yet to deliver.

The last four crypto cycles have been started by retail investors and the next one will also likely be started by same. But we are still waiting for the next narrative that will draw in the retail public and convince them of the benefits of alternative currencies over national currencies.

As industry analyst Tony Sheng notes: "When people are ripe for a mass movement, they are usually ripe for any effective movement, and not solely with a particular doctrine or program."

So what are the some potential narratives that could trigger the next change in public sentiment in favour of alternative currencies? The following all have potential:

• A currency war between China and the US
• The disintegration of the European Union and its currency the Euro
• A global debt blow up
• A global recession
• Civil unrest over inequality (are France’s Yellow Jackets the start of a global movement?)

Conclusion: Thinking outside the box

Alternative data isn’t as illusive as it sounds and the ‘alternative’ tag will likely be dropped in due course as the use of geolocation tagging, social media sentiment and satellite images become more commonplace.

Nor is alternative or big data limited to purely quantitative strategies or algorithms but can complement fundamental research and analysis of companies and assets – anything that will give an edge to discretionary decision-making.

The crypto markets rely heavily on technical analysis to pick bottoms and reversals but there is a growing interest in fundamental valuation metrics and sentiment analysis. While these metrics are still being refined to create accurate overvalued/undervalued signals for individual crypto-assets, as a broader market turning point we should keep an eye outside the crypto-sphere on the macro changes in sentiment: social, political and monetary.

Charting prices to pick bottoms and tops has its place in trading but can also be complemented with broader sentiment indicators to assess a change in human behavior and market environment.

Follow @AndrewBNC

Subscribe to BNC’s newsletters for insights and forecasts direct to your inbox