Measuring the predictability in coin price data

I love to dig deep to really understand the true nature of the crypto market. As such, I've created this heatmap showing you how predictable/unpredictable 100 coins I've randomly picked are (the x axis is kinda crowded so not all coin names are visible). Not just whether it's random or not—but how random it is.

The data is first collected from Coingecko (~10Y data) and then processed via the application of the wavelet transform. By doing so, we filter out the noise and random fluctuations. As a result, we get back finer information represented on 5 levels—level 1 contains the long-term trends, progressing to level 5 containing more and more short-term details.

Then, I apply a Shannon entropy calculation to determine how random/unpredictable each level is. The higher the entropy, the more random the data. The lower the entropy, the more predictable the data.

High entropy levels have darker colors, while low entropy levels have lighter colors.

Reminder that high unpredictability does NOT equal high volatility. Predictability here refers to how likely cycles repeat within the price data, while volatility refers to how big the price swings within/across cycles. A coin can be both unpredictable (no patterns) and involatile (no big price swings). On the other hand, a coin can be both predictable (recurring cycles) and volatile (big price swings).

https://preview.redd.it/nsyr9izu210d1.png?3422&format=png&auto=webp&s=1f92347f07b9b7a08721696f371b9061071bb9b0