A few days back one of our readers suggested that a useful comparison would be to compare the performance of our AI models to the performance of just holding Bitcoin (Thanks for the idea Slayer). The suggestion, quite rightly, was that with cryptocurrencies on a roar a user could get good gains just from sitting in Bitcoin so it would be good to see the added value from the models. I liked the suggestion immediately because, as was suggested by the poster, it provides a comparison against simply buying and holding Bitcoin in it’s own right – largely seen as the Gold Standard for Cryptocurrencies.
It also allows us to split our performance into Beta (how well we track the return on the designated market) and Alpha (how well we outperform the return on the designated market). So before publishing the regular results plus the bitcoin comparison I should clarify what this is and what this isn’t.
First what it is. We have taken the bitcoin price at start of the experiment (our first post) and assumed that the comparison is to fully invest in Bitcoin at that point in time. We then hold onto the bitcoins for the full duration of the experiment. This model incurs no commission (we assumed zero commission on the initial trade). This forms the basis of our comparison.
Now what it isn’t. This is not a comparison of how the models would have done if trading against bitcoin (flattening back into bitcoin rather than USD). To do that would require a retraining of the model as the relative strengths and relationships would be very different.
So with that said, current overall performance has continued to be way over long term expectations, we are now showing a gain of 54.61% since inception which is far from expected returns.
|Currency||Starting Balance||USD||Crypto||Rate||USD Equiv||TOTAL||Gain/loss|
As can be seen buying and holding also returned exceptional performance, 50.53%, giving us a modest outperformance of 4.08% in little over a month. The closeness of the two models shouldn’t be any surprise as it has been a period where the models have been long for much of the month. The other factor which results in similar performance, at least at first, is the fact that 70% of our starting funds were committed into bitcoin. In fact, the surprising thing is that we have been able to outperform at all. The real test of the value of the models will come when we eventually get a downturn and, although it feels like this will never happen at the moment, it will and we should watch the relative performances as the model actually needs to start actively trading the markets.
Thanks for reading,
— Wintermute —