Different data, different results

For over 50 years I have been doing mathematical analysis on a variety of subjects including sports statistics, stock investments, large mainframe computer performance, and more lately cryptocurrencies. I’ve learned that mostly the results are surprising and don’t support your thesis going in. That is the case in my latest analysis.

In my previous post I reviewed an analysis I did last summer on two simple cryptocurrency trading strategies. One was a buy and hold. The other was to sell a portion of the currency with the best performance for the day and use the funds to buy the one with the worst. Effectively somewhat of a re-balancing daily. There were 3 currencies involved BTC, ETH, and LTC. The buy and hold outperformed the re-balancing approach. That was disappointing but not a major surprise.

I’ve now rerun the analysis with more and largely different data. I’m using the last 4 quarters, from April 1^st, 2017 to March 31^st, 2018. There are 4 months in both sets of data. I started with $1,000 split equally between the coins and iterated daily computing the new balance. The buy and hold was simply a price change. The rebalancing required computing the changes and doing a simulated sell and buy, then computing the new balance. The resulting balance for the buy and hold was $9,532, while the re-balancing ended with $10,423. In this case the re-balancing outperformed the buy and hold.

The two tracked more closely than with the previous study. The next step is to dive a little deeper in the results to see if anything else can be learned. Follow me on Twitter, @ole44bill, to know when the next post is made.