Disclaimer: This is NOT investment advice. This research is for informational purposes only. We do not offer investment advice and nothing we say should be construed as investment advice.
Over the past year, the crypto market has been riddled with colossal price swings – and with BTC and ETH’s most recent nosedive, we’re likely not out of the woods yet.
These enormous ebbs and flows have prompted market participants to explore new ways of protecting themselves from downside risk. Inspired by the old ‘eggs in different baskets’ investment motto, we decided to try and minimize the crypto market drawbacks by building a traditional, high-diversity/low-correlation portfolio.
Known as the “holy grail strategy”, the idea is simple: balance your portfolio’s performance by offsetting negative returns of some investments by positive returns on other, largely-unrelated ventures.
Though it’s never that simple, the legacy financial markets at least have a working blueprint for this strategy: spread out your investments across different asset classes, include a healthy mix of domestic and international stocks, short-term investment and low-volatility bonds, real estate and commodities etc.
In crypto, however – where price changes across projects are so insanely correlated – building a low-correlation portfolio can feel near impossible. But even if you were to build one – would it be any more stable than the market?
To find out, we set out to identify and backtest a portfolio of projects that are least correlated to the two biggest market movers – Bitcoin and Ethereum.
Why those two? Well, despite the massive dips, Bitcoin still accounts for more than 50% of the total crypto market cap so can serve as a good proxy for the general market.
We also included ETH since a large part of the crypto assets currently in our database (around 700) are ERC-20 tokens. These are, to varying degrees, affected by the ETH price and vice versa, so it makes sense to include ‘low correlation to ETH’ as a bonus criteria.
Our methodology was as follows:
- We downloaded prices, volume, and market cap for all 1100+ cryptocurrencies/projects in our database.
- For each of the last 12 months (our selected time frame), we constructed a correlation matrix for the previous month to avoid the look-ahead bias.
- We excluded projects with low volume and low market cap to assure some quality in the projects. We also excluded stablecoins since they might bias our results.
- We selected top 20 projects with the lowest summed correlation to BTC and ETH and invested in them for the given month.
It’s worth noting that we were also always invested in both Bitcoin and Ethereum, since it obviously makes little sense to pick projects uncorrelated to something NOT in your portfolio.
Why only pick projects based on their (low) correlation to Bitcoin and Ethereum? Well, because we assume that the projects uncorrelated to the two biggest market movers are also likely to be uncorrelated to each other.
Alternately, we could also have looked at each project’s correlation to all other projects individually, rather than to the entire market (or its proxy in BTC and ETH). However, that approach would have likely increased the randomness of the end results, so we opted for a more reliable method.
The graph below shows the performance of our low correlation portfolio, compared to HODLing both Bitcoin and Ethereum.
The blue line is our portfolio, green is HODLing ETH, and orange is HODLing BTC. The y-axis shows the cumulative percent change, meaning that a value of 5 on the graph would mean the portfolio turned $1 into $5.
Here are the results:
As you can see, for the selected time frame (November 2017 – November 2018) our strategy would have resulted in a slightly negative return, while still beating out the HODLers.
Moreover, our low-correlation portfolio performed best during the end-of-2017 bull market, recording significantly stronger tops than both benchmarks. After the bull run, the portfolio experienced a similarly extreme decline, ending up at very similar levels to BTC and ETH HODLers today.
The main insight here, however, is that the low-correlation portfolio has proven even more volatile than our benchmarks, which is the exact opposite of what we expected to see!
For additional testing, let’s compare our low-correlation portfolio to one made of random crypto projects. The following graph shows an average of 20 such random portfolios, plotted against the same two benchmarks:
Here we see that our random portfolio(s) perform pretty similar to both HODLing Ethereum and Bitcoin. The tops are much lower than with our low-correlation portfolio, while the rest of the graph is at times indistinguishable from the ETH benchmark.
Now let’s look at how the exact opposite – a high correlation portfolio – would perform. We’ve picked and invested in 20 projects with the highest correlation to our benchmarks (BTC and ETH). Here are the results:
As expected, the performance is quite similar to our random portfolio, this time almost entirely matching Ethereum’s long-term volatility. These tests give us even more reason to believe that the effects of the low-correlation portfolio are not random.
Well, there you have it. On the one hand, looks like putting your eggs in low-correlated crypto baskets is indeed a viable investment strategy, at least in the sense that it outperformed both our benchmarks (alhtough not by much).
That said, we were NOT able to confirm our original assumption – that low correlation assets actually make a more stable, less volatile crypto portfolio. In fact, we found the exact opposite to be true; our strategy performed stronger in the observed time period, recording much higher highs, and analogously bigger dips. Myth – busted.
On an interesting sidenote, the results also seem to indicate that a low-correlation portfolio could work very well during a bull market, where it smoked both BTC and ETH HODLers – and by a wide margin. It might prove an interesting strategy to try and invest in a low-correlation portfolio at the start of the next bull market, and then switch over to stablecoins – or other less volatile projects – when the market turns ugly.
How can we explain our findings?
While the results were not what we expected, they (somewhat) make sense. On the whole, the crypto market is still way too correlated at the moment, so being relatively immune to big market movers could mean that the project is seen as either very bad or very good, contributing to its volatility either way.
If and when the crypto market develops clear separation between different sectors and asset classes (like the stock market), we may expect low-correlation portfolios to become more stable.
Until then, it doesn’t matter how many egg baskets you have in crypto – they will all still wobble like crazy.