It's a topic i’ve look into once before, and I believe will many times again - because the potential for AI and machine learning to impact markets shouldn't be downplayed. So if you plan on being around for awhile, now is probably a good time to start learning about it.
To gain a better understanding we spoke to Stephen Mathai-Davis the co-founder of Quantamize, a the firm that's applying this tech to the cryptocurrency markets, currently for the 25 top coins. Wall Street has been using similar tech for awhile now, his company being among those doing it. But stock or cryptocurrency, both share the same goal - better insight on when to buy or sale.
I should mention upfront - they don’t keep this data to himself, nor does he expect anyone to pay for it without proving it’s worth first. He’s confident enough to let people access the data for 3 weeks for free, and not some gimmick free trial they hope you forget to cancel - they don't even require a credit card.
I would suggest giving it a test run by doing some off-book theoretical trades, without really buying/selling any coins. First write down what you would have done without this new data. Then write down what the AI was predicting and follow the results for both. Do this for awhile and see the results,it will be obvious if you're making better trades with it or not. Personally i'm using it as a 'second opinion'. If both myself and the AI agree - it's probably the right move. If i'm about to sell a coin the AI is saying I should be buying, it's a sign I should take another look because perhaps it saw something I missed.
Like me you probably have a whole lot of questions, hopefully I covered them. Here's my conversation with Stephen Mathai-Davis from his New Jersey office..
Let's get to know who we're talking to before we get to the technicals, tell us about your background, what did you do before creating Quantamize?
Before co-founding Quantamize, I spent most of my career in institutional investment management as both a trader and research analyst. I was an analyst and junior portfolio manager with coverage responsibilities for global consumer, energy & power, and financials stocks. Over the course of my career, I have been lucky enough to have had the experience of working in areas related to trading (at an investment bank, hedge fund and a large asset management company), research, portfolio management and financial technology. They were all very cool experiences that really helped shape many of the skills I would need to help launch Quantamize.
How about the rest of the Quantamize team. What unique backgrounds or skill sets do they bring to the table?
I co-founded Quantamize with my two parents, Wallace and Prema Mathai-Davis, and we are joined by Rich Ackerman and Paul Baessler. The entire team has extensive backgrounds in finance ranging from basic asset management to venture capital and FinTech. Wallace and I are the two co-managing members of Quantamize. Wallace has had a distinguished career in institutional Wall Street, running several large asset management businesses as well as being a senior advisor and board member on several successful FinTech startups. The five of us are joined by Ed Boll, Bill Visconto, Jim Ryan and Arnim Holzer from the EAB Investment Group. EAB is an elite options investment/trading firm and has known the principals of Quantamize for several years.
So before cryptocurrencies you were using this technology on stocks, now you do both. There are some obvious similarities, and differences, how did you have to adjust to make that leap from stocks into crypto?
We actually decided to experiment first in the crypto market. From our vantage point, we thought it made more sense to experiment more with cryptocurrencies since the market is more welcoming to different types of approaches that might be looked at with more skeptical eyes in stock investing (and other traditional asset classes). Let me illustrate an example: we first implemented our approach of using several machine learning algorithms to cryptocurrencies. Since ML algorithms are notoriously unstable when applied to financial markets, we thought it made sense to use multiple algorithms simultaneously to increase the stability of the accuracy. Of course, the fact that crypto price data is non-stationary also weighed in heavily on this decision. The net result is that we use many types of machine learning algorithms at one time, both supervised and unsupervised, to predict each crypto we model. Just for reference, each individual crypto we model has its own unique basket of algorithms. However, all that aside, we still find it easier to predict stocks...the instability of crypto time series data makes it very difficult to model!
Why did you decide to expand into the crypto market?
As a firm, we have been drawn to the decentralization focus of the cryptocurrency, and blockchain movement at large. Personally, I have been involved with Bitcoin going back to early stages. Growing up, I was avid video gamer (RTS much more than FPS and MMORPG) and it felt natural for me to gravitate towards the cryptocurrency community given its similarities to the old online communities I used to participate in when gaming -- feels like many of the innovators in the space are some of those same gamers!
Between stocks and crypto, which is easier to predict? (Maybe not 'easier' - but which are you predicting more accurately?
Hands down, it is easier to model stocks. While we don’t have a blind faith in the brownian motion of stocks, it is certainly easier to build around a basic drift in stocks than it its to find anything remotely reliable in cryptocurrencies. Like I said earlier, the instability of crypto markets causes the time series data to become non-stationary which materially impacts the ability of traders to apply traditional quantitative techniques to these markets.
What factors are taken into account and analyzed by Quantamize?
We look at several factors and derivatives of factors. Without getting lost in the jargon, we focus on arbitrage across different exchanges (we look at north of 200 exchanges globally), volume trends, sentiment data, data from the blockchains themselves as well as the behavior of large players, or whales. Our models consider all these factors as well as embedded decay rates for their unique time series.
For a Quantamize user, how is that data then presented to them?
Results are presented in a very straightforward way -- we try to simplify the process of making an investment decision for our subscribers. Trading signals are either “Buy” or “Don’t Own”. Why did we choose “Don’t Own” over “Sell”? We just don’t think it is wise to recommend our subscribers try to short technology, which essentially is what a digital currency represents, in its nascent stage. To avoid that level of confusion, we chose to use the language “Don’t Own” over “Sell”.
How should a user use this data? People see a "BUY" or "SALE" next to a coin and think that means "dump it all!" or "put your life savings in!" - but really what kind of trading strategy is this intended for?
That is a great question. Our signals are based on Buy/Don’t Own. Now, in the real world, we wouldn’t expect users to put ALL of their life savings into a “buy signal” though we do feel that if “traders” see a “don’t own signal”, they should sell everything. Remember, these strategies are built for those subscribers looking to trade in and out of cryptocurrencies. For those subscribers simply look to create an “allocation” to cryptocurrencies, we have created our specialized portfolios. These portfolios are built around aggressive, low volatility and thematic strategies. We are especially proud of our low volatility strategy which seeks to manage overall volatility in the portfolio while also placing a special focus on left tail volatility. We believe left tail risk is the biggest risk facing an investor in the cryptocurrency market today. Our Fat Protocol “cryptofolio” is made up those digital currencies we consider viable utility tokens. We use unique form of AI to create a very balanced allocation for those of our subscribers who are interested in these types of cryptocurrencies.
How far out are you able to make predictions that are still at least 'usually' accurate?
Typically, for our trading signals, we only predict out 3 days. Why only 3 days? We find it difficult to predict further out effectively given how quickly the crypto markets change. Similarly, trying to predict short-term moves in cryptocurrencies is very difficult with any level of accuracy. We are very proud of the fact our machine learning models have high accuracy scores and are reticent about doing anything to compromise our ability to give our subscribers the best possible recommendation.
Care to share some current predictions? How about best/worst coin in the short term, and same for the long term.
Ha, that is a tough question. We never go that far out. However, if we you were to ask which coins we like in general, it is definitely “the fat protocol” coins. We are qualitatively biased to utility tokens since, essentially, the user is basically buying into a piece of technology. All that aside, if you were to ask us what coin your readers should consider RIGHT NOW, we would definitely say buy Bitcoin. As goes Bitcoin, so goes the rest of the crypto markets at the moment. Correlations are too high with dispersion trends too low to warrant buying smaller-cap coins in the altcoin space at this moment (assuming limited knowledge of the technology and/or blockchains behind these smaller, less liquid digital currencies). For those who are interested in investing in altcoins, we would recommend using our diversified low volatility strategy.
If anyone reading this would like to try Quantamize out for free, you let them. What do they need to do?
Totally! Quantamize offers a 21-day free trial without asking for a credit card. Sign up and take the Quantamize platform for a spin to see if you like it. Free trial members get full access to the entire platform for the entirety of their free trial.
Anything the horizon that we won't see yet when visiting the site?
We will be releasing an awesome asset allocation tool in the coming weeks which will let our subscribers create diversified allocation of different kinds of ETFs and Bitcoin. Just imagine a multi-asset allocation that also include Bitcoin based on your risk profile and risk biases. No robo-advisor is offering this type of combination. We will even let users backtest the results of their optimal allocation just to see how it would have done over the past few years. We are really excited to bring this awesome tool to our subscribers! Thanks to Stephen for taking the time to speak with us, and you can see it for yourself at https://www.quantamize.com
An Official Global Crypto Press Association Report - http://www.GlobalCryptoPress.com