Bitcoin’s price history is known for being especially volatile and uncertain. However, one finance professor believes that someday AI could be used to create a predictive model of the world’s first cryptocurrency.
As much as TA experts like to write otherwise, Bitcoin is a fundamentally unpredictable asset. It has puzzled traditional investors for years, broken all expectations, and continues to confuse critics who previously claimed it was yet another “tulip bubble.” Yet, as Bitcoin remains strong, we’re still no closer to understanding its price movements. One finance professor, however, believes that more accurate estimations will ultimately need to come from AI.
What If You Had a Crystal Ball?
Finance Professor Nikola Gradojevic and his colleagues at the Lang School of Business and Economics at the University of Guelph have created an artificial neural network (ANN) model to try to predict Bitcoin’s price movements. Set for publication in late 2019, the model pulls daily data from 2011 to 2018 with three predictors: 200-day buy-sell signals, 50-day buy-sell signals, and returns. They also included stock market volatility into their equations to see if it had any impact on Bitcoin’s price.
How the ANN works is similar to the neural functioning of the human brain. In short, the end goal is the discovery of patterns. Overall, the research team says that the ANN model was successful in reducing prediction error by about five and 10 percent. Being an AI, however, it is still in the process of learning and becoming more accurate. Yet, as volatility increased, the ANN struggled to keep up.
The ANN was able to map the entire 2011 to 2018 period with some 63% accuracy. As Gradojevic explains, this means that the ANN has proven to be more profitable than pure random buy and sell orders.
Certainty is Impossible
Although the findings of ANN seem promising, they nonetheless end up underscoring Bitcoin’s inherent unpredictability. The fact that only 63% accuracy is considered an impressive feat just speaks to how little we still know about what the future brings for Bitcoin.
For Gradojevic, he is not interested in the theoretical and economic debates surrounding Bitcoin. Instead, he sees it as the perfect means to test the limits of our predictive AI models. With ANN, the research team hopes to at least sketch out some fundamentals on how Bitcoin’s price tends to move. Although we may never have certainty, the ANN at least provides us with hints on the future—pushing AI to its limit while doing so.
Do you believe artificial intelligence could ever learn to predict Bitcoin’s price movements? Let us know your thoughts in the comments below.