The power law model as a function of time

Yesterday I posted that the power law model was never more accurate than this cycle. Today I fitted the power law model once per day using the whole dataset since the beginning until that day, and I compared the forecast of each model to the actual data "from the future" that we now know.

The first graph compares the forecast of 9 power law models evenly chosen between January 2016 and March 2025 with the actual data.

The second graph shows the prediction of the power law model for today (25-Jan-2026) using models from every single day in the past.

Interestingly, if we look at the predictions for today, since 2016 they were always in the range 90-200 kUSD/BTC. Not bad taking into account that 1) back then 1 BTC was around 400 USD, 2) BTC was only 6 years old and the predictions were up to 10 years in the future and 3) USD/BTC changed about 2 orders of magnitude since then.

Whether or not this model will still hold in the future, is of course unknown. But those who trusted it in the past, were not that off from what is actually happening, as we can see.

See ya in 10 years.