Hi George, nice to have you. You’ve had quite a long and illustrious career from high level positions at banks, to starting companies, to running hedge funds. Can you give us an intro of yourself and what you do?
Great to be here together with other social traders on a platform that I believe is a glimpse of the future of asset management. I am an electrical engineer by training and I moved to a sell side investment bank just after university, trading derivatives in London & NYC. 25 years later, mostly spent in finance, some running a CTA, some in tech, and I am still trading primarily liquid equities, futures, & FX. Currently, I also run an AI consulting company and I am involved in some deep tech ventures.
I read the Nakamoto paper very early and I have to say I underestimated the potential implications initially. I experimented early on with crypto but only really started trading round 2016. Crypto is an amazing risk asset, nicely volatile but with liquidity challenges. Since 2016 I gradually increased my trading in digital assets from initially 10% of my accounts to now almost 50%. Running money is primarily an exercise in risk management but also understanding and developing a certain system of investment “faith” that fits one’s personality and capital base.
Can you go into more detail about your Set, its strategy, and any backtest data?
The ETH Smart Beta set is a rebalancing process that aims to profit from most of the ETH upside while diminishing downside risk. It is an ideal vehicle for investors that want to have exposure to Ethereum, but avoid long periods of drawdowns. It is a way of gaining ETH beta with downside protection, or in other words, optimizing for Sortino. I am also preparing a BTC set with the same approach.
My system is a quantmental approach (quant+fundamental) where I look at the statistical properties of price data and alternative data from both the crypto universe, and the wider investable universe of fiat priced assets. Despite earlier uncorrelated behavior, currently both BTC and ETH trade as risk assets with a high cointegration to equities. That means my toolkit of macro indicators is becoming more relevant than ever to trading the top crypto assets. The quant price signals are risk weighted inversely by a proprietary entropy based risk engine. At that point I use discretion to look for potential gap risk or other fundamental factors that may affect downside performance and decide on overall system exposure from 0% to 100%. So the quant model may indicate 80% long but I may trim that to 40% based on my assessment of market risk.
As such the backtests are not fully relevant to the set but I may actually publish them at a later stage to engage in discussion with the community and possibly create a fully automated set.
How has working at the bleeding edge of big data and AI affected your trading strategy?
I have become much much more aware of the dangers of overfitting. As we like to say in the quant space “if you beat the data enough it will confess to anything.” Machine learning in finance is fundamentally a different proposition than in tech. Non-stationarity and low signal to noise ratio are key issues that need to be acknowledged for good long term performance. That, along with a need for some interpretability that stems from a (however transient) true alpha source, makes the use of cookie cutter AI approaches in finance rather useless. Custom AI approaches are needed to extract alpha.
I have found AI approaches to be super useful at a higher level than model creation, primarily in detecting model failure, regime identification, and the creation of meta-strategies (strategies of strategies).
What has been the hardest part about transitioning between managing a traditional fund and trading crypto?
The hardest part has been getting to grips with liquidity gaps and also the 24/7 nature of trading. The dispersion of trading venues and the potential risks from hacking are also very new and unusual for someone trading listed futures with almost zero counterparty risk or potential for loss due to hacking. That is why I like the DeFi decentralized, non-custodial approach to trading crypto, albeit at a cost of some speed loss and slippage.
Where can people find you online?
Twitter : @gcotsikis
Telegram : @moonshot
LinkedIn : https://www.linkedin.com/in/gcotsikis/
TradingView : https://www.tradingview.com/u/entropycapital/
Blog : https://entropy.substack.com/
Anything else you’d like to tell us or your potential followers?
I would tell them that if they are reading this they are at the cutting edge of finance, way ahead than most. However, passive long only positions can be painful in any asset and they should use the flexibility of managed strategies offered by TokenSets to safely participate in the upside of the ecosystem. Crypto does not magically exist in a bubble where different laws of finance apply. The elegant algorithmic central bank of Bitcoin or the global Ethereum computer are amazing innovations, but the price of those assets against fiat currencies is not immune from demand/supply imbalances that in the short term can be very dramatic.
Another point to make is that backtests are often flawed and almost never as good as the live trading. Without expanding on this, I would urge investors to be rather cautious when looking at nice backtested charts and evaluate the strategy thesis at a deeper level if possible.
It is still early days for digital assets; take the opportunity to be invested wisely via managed strategies!