We believe that today’s investment analytics products are missing a key element: while they provide raw data, they leave the interpretation entirely up to the user.
However, analyzing thousands of predictive factors without a dedicated quantitative analyst team is a complex and challenging task.
Unravel is an investment research platform designed to address this gap by quantifying the predictive strength of each metric and ranking the most important predictive factors.
Lets take an example predictive factor: Exchange Outflows. High number of Bitcoins leaving exchanges (to cold wallets) can be a great sign that investor confidence is increasing, and historically, on average, it lead to price appreciation.
There are two ways of looking at Exchange Outflows: its absolute value ($245.34 Million) and its relative value (Very High - measured against its historical average).
In practice, its really hard to tell if $245.34 Million is actually a Very High value or just a normal week.
The meaning of $245.34 Million also changes over time - at 2015, it could have 10% of the whole crypot market cap, but now it's just 0.1%.
By normalizing a metric (compared to its historical average), we make it comparable across time. This enables predictive analytics.
We analyze thousands of scenarios, and tell you how BTC performed when Exchange Outflow was Very High or Very Low.
We turn this into a forecast - if Exchange Outflows are very high right now, we display what happened in the past when that was the case.
We quantify predictive strength with industry-standard best practices: first, using linear correlation.
Simple correlation is used to express how assets are moving together. Predictive Correlation refers to how a metric leads an asset and can forecast future price movements.
When a hypothetical “perfect predictive factor” is high, the asset should rally in the next days, when it’s low, the asset price should fall.
Predictive Strength
We also created Predictive Strength, an easy-to-digest version of Predictive Correlation (with 90 Day forward returns), with the following buckets:
Predictive Strength
Predictive Strength