There's a paradigm shift happening in trading.
Today's investors have access to data and information that would have been unheard of 10 years ago … and unfathomable 20 years ago. In the past, investors relied on information and experience from their real lives, from counterparties, and from fastidious attention to CNBC and stock tickers.
While the games, the rules, and the players have all changed, the goal hasn't … more alpha … more money … more reliably.
Algorithmic trading isn't new, but there is a shift in who's making the algorithms. For example, you can crowdsource development through Quantopian … or let machines do the heavy lifting through A.I.-based firms like Sentient.
Some argue that artificial intelligence is unable to generate significantly different results because "analyzing more and more data results in increasingly similar strategies".
But I'd argue that's only true if you look at the same data, the same way.
The Future of Trading
One of the reasons A.I. is a great option for trading is that it takes away the human element of fear, greed, and discretionary mistakes.
Sentient's founder says:
"For me, it's scarier to be relying on those human-based intuitions and justifications than relying on purely what the data and statistics are telling you." – Babak Hodjat
In addition, people tend to get similar results because they do things similarly. As A.I. matures (and more researchers become better versed in what's possible) solutions will evolve.
It won't be a Ph.d. writing an algorithm … it will be machines and code trying unthinkable combinations and finding edges that otherwise would remain invisible and unused.
Currently, most people train their algorithms on markets, or with human intervention, but there are more data sets that can be used to build more robust models.
Alternative data, to most, means tracking Twitter and Facebook sentiment, but confining your definition to that limits potential alpha.
New sources of data are being mined everywhere, and are letting investors understand trends "before they happen".
For example, mobile devices, low-cost sensors, and a host of new technologies have led to an explosion of new potential data sources to use directly for predictive insight or indirectly to help improve models.
In addition, private company performance, logistics data, and satellite imagery are becoming popular data sets in a data scientist's alpha creation toolbox.
There are often concerns about the cost and completeness of these datasets, but as we get better at creating and using them, both will improve.
Here is a chart of alternative data sources.
Finding more ways to train algorithms on new data can help traders once again find an edge on their competition.
The thing about "sustainable alpha" is that while one might be able to achieve it, you can't expect to have it doing the same thing everyone else, or that you've always done.
Markets change, and what worked yesterday won't necessarily work today or tomorrow. Trading is a zero-sum game, and as we move toward the future, this only gets more apparent.
Behavioral Game Theory shows that human choices don't necessarily reflect the benefits they expect to receive. That's no longer the case with algorithms.
For more on Big Data and its potential, here's access to the full panel discussion I participated in recently at The Trading Show in New York.
Let me know if you have questions or comments. Thanks.