The Cryptocurrency Market is No Exception — There is NO Best Rebalancing Period
We backtested 10,000 cryptocurrency portfolios to find out.
At Hodlbot, we automatically rebalance people’s cryptocurrency portfolios. So inevitably, one of the most common questions I get is, “what is the best rebalancing period?”
I attempted to answer this in my last article by simulating 100,000 portfolios using the Monte Carlo method and found more frequent rebalancing did not have any effect.
But, many of my readers & followers were not satisfied. A lot of investors are under the impression that there is something special about the cryptocurrency market — something that gives an edge to more frequent rebalancing.
I’ve never been one to shy away from data. So in this article, I will be investigating the impact of rebalancing frequency on portfolio returns by running backtests on historical cryptocurrency market data.
Case Study Methodology
For this case study, we’re going to generate 10,000 random, evenly-distributed, portfolios and backtest each portfolio across a large range of different rebalancing frequencies (1 to 90 days).
After each backtest, we’ll record the rebalancing bonus, which we’ll define as the excess return the rebalanced portfolio produces over the drifting one.
We’ll only include coins that have:
- Price data for the last year (December 1, 2017 — November 30, 2018)
- Ranked within the top 200 by market cap on November 30, 2018
Backtest Assumptions
- Use BTC as the main trading pair
- Apply a 0.1% trading fee
- Apply a 0.001 BTC minimum trade limit (common on most exchanges)
*Code and dataset can be found at the bottom of the article
10,000 Simulations Later….
Monthly, Weekly, Daily… Does it Matter?
Each dot represents a single simulation. Outliers > 3 standard deviations removed
If having shorter rebalancing periods improved the rebalancing bonus, we’d be able to fit a trend line from the top-left to the bottom-right of our graph.
But that’s clearly not what we see. In aggregate, we observed no relationship between rebalancing frequency and rebalancing bonus.
That is not to say, that changing the rebalancing frequency on a specific portfolio has no effect. Quite the contrary. When we take a portfolio and change the rebalancing frequency, we end up with a different rebalancing bonus for each period.
Rebalancing Bonus for the 5 different portfolios across different rebalancing periods. Each portfolio is denoted by a different colour.
The problem is that there is no general rule for all portfolios.
Every portfolio has its own ideal rebalancing period that is based on the price movements of its underlying assets. One portfolio’s ideal period could be 78 days, while another is 3 days.
If we take a look at our simulations, we’ll see that the highest rebalancing bonus values come from a mixed bag of different rebalancing periods.
Hindsight is 20/20
While we can evaluate what the best rebalancing frequency would have been, but there’s no guarantee that will continue the best one going into the future. We don’t have the ability to predict the ideal rebalancing period ahead of time.
To do so would require an uncanny amount of insight regarding the volatility, correlation, and price behaviour of the underlying assets. If any trader could predict the future with such a high level of precision & accuracy, there are better ways to make money, compared to trying to choose the best rebalancing period.
Fortunately, there are still some general heuristics we can apply.
1. It’s generally better to rebalance your portfolio
A hexagonal binned plot, further shows that the majority of rebalancing bonuses fall between 0 and 20% across all rebalancing periods.
Outliers > 3 standard deviations removed
In this context, rebalancing is generally better than no rebalancing. But that does not mean it is always better.
Summary statistics of rebalancing bonuses after 10,000 simulations
2. Shorter Rebalancing Frequencies Increases the Likelihood of Failed Trades
Rebalancing periods shorter than 1 day are mostly infeasible for everyday investors.
Unless you have a large portfolio or are only holding a couple of coins, most of your trades will fail due to minimum trading limits imposed by exchanges.
Here’s what the % of successful trades look like, assuming a $1,000 portfolio dispersed evenly across 10 coins. When rebalancing hourly, less than 4% of trades go through.
Average % of successful trades based on 1,000 simulations of a random 10-coin portfolio
Even with a $10,000 portfolio, hourly rebalancing doesn’t look great. Only ~60% of trades go through.
3. Shorter Rebalancing Frequencies Increase Trading Fees
Assuming a 0.1% fee across all successful trades, here’s what the fees look like for a $10,000 portfolio.
Daily rebalancing racks up ~4.5% in fees, hourly a whopping ~20%.
To me, hourly and daily rebalancing frequencies are prohibitively expensive so I would avoid rebalancing at those intervals.
Wrap-Up & Takeaways
In summary, shorter rebalancing frequencies do not improve the rebalancing bonus. But, shorter rebalancing frequencies do increase trading fees and the likelihood that trades will fail.
For a cryptocurrency investor with a portfolio under $10,000, any rebalancing period anytime between 7–90 days is appropriate.
Furthermore, as my previous blog showed, you can try to maximize your rebalancing by:
- choosing assets that are volatile
- choosing assets that have a similar mean rate of return
- choosing assets that have a low or negative correlation with one another
- choosing mean-regressing assets
- having an evenly distributed portfolio
- having a large # of assets in your total portfolio
Why Do I Care so Much About Rebalancing?
I’m the founder of HodlBot.
We automatically diversify and rebalance your cryptocurrency portfolio across market indices. We also enable our users to automatically create and rebalance their own portfolios.
Join the Conversation
If you want to talk about this article, you should join our Telegram group of 300+ members.
Code
<a href="https://medium.com/media/a03cd93b81914b6d90e007748cdfb945/href">https://medium.com/media/a03cd93b81914b6d90e007748cdfb945/href</a>
Price History Data