The Signal and the Noise – Nate Silver
Market Meditations | April 7, 2021
There is no certainty in the crypto markets. The potential rewards are too high to entertain certainty. We are in the business of making predictions. Calculated predictions cap our downside and create opportunities for exponential returns.
That’s why, today, we will take a close look at the art of making predictions. Specifically, we will summarise Nate Silver’s The Signal and the Noise.
First things first, what is a signal and what is noise. The presence of media and then social media has dramatically increased the amount of information in circulation. Now that would be a good thing, if the amount of useful information was increasing at the same rate. Which it isn’t. That means traders and investors are left with the challenge of trying to distinguish between useful information (signal) and misleading information (noise).
The goal of a prediction model should be to capture as much of the signal whilst drowning out the noise. As always, we come back to how biologically unsuited we are to trade the markets. Evolutionary instincts sometimes lead us to see patterns where there are none and as such, a prediction model is even more important.
1️⃣Correlation does not imply Causation
When creating a prediction model, it is important to recognise that correlation doesn’t imply causation. If we plot two coins next to each other on a 1 month time frame, it might look like they have a perfect negative correlation. That is, when one rallies, the other sells off. This correlation is not enough evidence to suggest that one event is causing another. And so often traders and investors jump to this conclusion. It is imperative to find more evidence before drawing causality. For instance, a technical pattern ought to be matched with a fundamental observation. Or, a strong narrative, should be underpinned by what you are seeing on the charts. Do not overestimate the predictive power of variables in isolation.
2️⃣Binary to Probabilistic Models
Next, we need to move from binary models to probabilistic models. If we asked you what will happen to Ethereum prices over the next few months, a good answer is not ‘up’ or ‘down’ and it’s not ‘up 40%’ or ‘down 20%’ either. A good answer is a spectrum of outcomes with probabilities attached to them. That is the way to view reality.
Once we have established a method to view reality in this way, we need to create a pragmatic model. One that is susceptible to adapting to new information. Or robust. You might be short term hyper bitcoin bullish and have a prediction model which depicts it as such, but, if, for example, we close below the 0.382 level, we should be able to adjust our predictions to factor for this new signal.
From one day to the next, we might be looking at very different predictions. When new information presents itself, we must always update our hypothesis. Only through this process of refinement will we move further away from the noise and closer to the signal.