A Stroll Down Crypto Trading

A good trader, before anything, is a good risk manager. This was one of the first things my boss at SocGen on the prop trading desk told me.

The more complex the assets you trade, the truer that statement is.

Cryptocurrencies are one asset class that is super interesting, increasingly more relevant, and that lends itself well to illustrating this adage:

• They are very volatile, about 5x more than stocks - They have strong trends, about 6x more than stocks
• They experience much faster drawdowns than stocks, of otherwise similar magnitude compared to their level of risk
• They are challenging to short

As part of my illustration, I will put together a simple trading strategy that only makes risk management considerations over a static long position. Two simple principals in risk management underpin that strategy:

• You need to use your risk budget wisely with changing market conditions?
• Markets can be wrong longer than you can stay solvent?

I’ll rely on a few metrics to quantify outcomes from risk management decisions on the stream of returns generated by my trading strategy: - sharpe ratio - max drawdown in proportion to my risk budget - average drawdown in proportion to my risk budget?

This section won’t go in the technical details of how I implement risk management decisions, categorize and detect market regimes, or determine the pace at which information gets incorporated. These matters are IP-heavy, and therefore are a subset of the expertise my LPs pay for.??

1.??? BTC and ETH Risk / Reward Analysis

First things first, let’s look at how BTC and ETH behave across market regimes:

• Strong Bear / Bull
• Moderate Bear / Bull
• Range Bound

A few observations seem worth flagging:

• When not much is going on, you are better off being long these assets, simply due to fundamental considerations over their long-term adoption / utility / value
• These are very non-linear assets. Crypto Summers generate a lot more value than what the Crypto Winters destroy
• BTC and ETH behave much more differently in bear markets than what a lot of market participants think. In particular: a) They each appeared to have dropped by the same magnitude whether in a strong or moderate bear regime. b) ETH losses have been a lot more clustered in time than BTC losses

A similar analysis on volatility, rather than performance, reveals a few complementary findings:

• Volatility macro regimes are quite clear and don’t seem to have a dependency on price levels
• BTC and ETH behave quite differently in risk terms. Notably ETH has a more elevated but more stable risk

?

Now, let’s look at how quickly new information gets absorbed by the markets:

• New information related to ETH is incorporated much faster than for BTC. a) It is reasonable to consider that the type and volume of information about either asset are similar. b) Therefore, the main explanation for why information is incorporated much faster with ETH than with BTC has to do with the differences in market participants (proportions of retail/wealth/institutional, proportions of active/passive investors, proportions of short-term/long-term traders)
• Markets in those assets have become more efficient

Where does that leave us?

• Sizing of positions needs to be fairly actively managed
• Drawdowns need to be actively managed
• BTC and ETH positions should be treated separately, even if they are part of the same portfolio

2.??? Dynamic Portfolio Management

I am using monthly returns on BTC and ETH, and any risk management decision is based on measures lagged by a month. This provides a very conservative evaluation of the benefits of those decisions. In practice, an investor is a lot nimbler to implement decisions, and a much more accurate analysis should therefore rely on daily returns. However, this also comes with other constraints on representing the impact of portfolio rebalancing. Note as well that given that markets have become more efficient and more sophisticated in the recent period, working with more granular data and more accurate representations of the impact of portfolio rebalancing decisions is a necessity in any case for institutional managers even though the framework I use here remains valid:

• Stabilize the volatility produced by the strategy over time. This makes portfolio returns more predictable so easier to manage
• Don’t remain passively long BTC or ETH when there appears to be a meaningful shift in regime. This makes portfolio returns more stable so easier to scale
• Exit when cumulative losses are getting too acute. This aims at not getting fired / shut down so that you can continue deploying

?

Synthesizing with a few useful portfolio metrics: Sharpe Ratio, Worst Drawdown relative to the strategy Risk Budget, Typical Drawdown relative to the strategy Risk Budget

• The dynamic portfolio management tools behave as one would expect. More consistent generation of positive returns, generally milder losses
• Improvements would be magnified when operating more realistically over daily timescales. Instead of reacting with a month lag, one would react with a few days to a week lag, accounting for the time needed to rebalance the portfolio with reasonably low trading costs