Based on Watson (won on Jeopardy game show) supplied structured and unstructured data
Data is fed into Equbot, some form of a big-data driven deep learning exercise
Equbot developed by academics (big name schools)
Analyze 6000 stocks daily so holdings will span small-to-large capitalization US stocks, best reference index could be SP500
Long only, generally individual stock exposure could be 4-5% at the higher levels (10% max); carrying 70 stocks right now
Expense ratio of 0.75%
Here is our experienced take on it, some obvious and other's perhaps important to potential investors:
No live track record (yes, obvious on new issue ETF) for the creators in managing other people's money
Relatively high expense ratio given the lack of active (rebalancing) necessary to generate net of fee alpha
It is dealing with stochastic information (stock prices) vs that of processing spoken words into high speed data access and learned responses (Watson Jeopardy), which is not stochastic in nature
High active risk relative to its benchmark, good and bad depending on your goals
Selecting from the basket of stocks that best fit its learned, optimization criteria can lead to "cherry picking" or sub-optimal real results
AI and big data seem be leading a great Smart marketing story ala "Smart" Beta so let's cut through the marketing hype.
What we see now, especially in the systematic, data driven product development world is products that come to market backed (not with their own money of course) by either academic research or academics themselves. Invariably, most of these products are hard-pressed to deliver what their creators thought their research or academic backgrounds were supposed to convey in terms of relative real world out-performance. We track as much as possible in this space (no doubt we miss many), but, to date, most of these well-meaning ideas have been trounced by the Efficient Market Hypothesis, meaning you are better off with a cheap passive product.
Here's why: For real AI in stock investing, it has to attempt to leapfrog "expert" systems to that of "cognitive", which allows it to deal with stochastic information and not get thoroughly lost. However, in our thinking, it has to be relatively simplistic in what data it sees and the learning framework (vs that of Watson) to make that jump a clean, effective one.
That said, congratulations to the Equbot and ETF Managers team on garnering $70M in new assets in the ETF! Clearly, they have identified strong appetite for AI-oriented products and we hope AIEQ is as intelligent at stock picking as its marketing.