1/10 @btconometrics published a new critique on @100trillionUSD’s Stock-to-Flow (S2F) & S2F Cross Asset (S2FX) models today, underlining the statistical flaws in both models

Time for a (hopefully simplifying) summary 🧵

btconometrics-84377.medium.com

2/10 He sees three primary components to the underlying theory:

1) S2F is a surrogate of privacy -> undisputed

2) S2F causes the price rise -> disputed

3) S2F ratio can be used to estimate estimate the future price [in a statistically valid way] -> disputed

3/10 To prove point 2, @btconometrics recreates the S2F model & shows that:

1) The S2F model parameters look great at first

2) ... but after controlling for autoregression (the fact that price on any date is related to price on the day(s) before), S2F no longer predicts price

4/10 He then introduces the “whitepaper to first block mined ratio” (WP2Rel) model, that does a better job at explaining the historical price, even after controlling for autoregression

The model is completely meaningless though and was essentially introduced as a parody

5/10 @btconometrics goes on to discuss the famous cointegration analysis (that he introduced himself in ‘19)

These analyses were originally thought to prove that S2F WAS non-spuriously related to price, which was invalidated by the fact that S2F ratio is deterministic

6/10 Next target: the S2FX model, that was designed to overcome these limitations

His arguments:

1) The choice of creating 4 clusters was arbitrary & prone to bias

2) The fact that the phases monotonically increase over time would mean it is still a time series by definition

7/10 @btconometrics then suggests that instead of attempting to forecast price action, first principle reasoning on why is (increasingly) valuable may provide a useful fundamental framework

8/10 @btconometrics then gives another example of a nonsensical model, illustrating that 93.5% of the price can be explained simply by the supply curve

This relationship is theoretically nonsensical and simply an artefact of both time series trending up

Follow

9/10 Finally, @btconometrics performs a different statistical technique (Fourier Analysis) to show that the price does in fact move in 4-year cycles thanks to its halvings - which is part of why S2F *seems* to work

10/10 The article is concluded by stating that:

1) Bitcoin is designed to become more expensive - forever

2) No tops or rubber bands holding the price to any level

3) Given the obvious increasing demand, at some point demand increase > supply increase, which means 🚀

Sign in to participate in the conversation
Bitcoin Mastodon

Bitcoin Maston Instance