Thread by @dilutionproof

1/n Yesterday, @100trillionUSD published his third Medium article, called 'Bitcoin Stock-to-Flow Cross Asset Model', in which he introduces a new iteration of the S2F model that utilizes a new approach to modeling assets based on scarcity:

medium.com/@100trillionUS…

A thread. 👇

2/n The first S/F model (March 2019) innovated by creating a power-law model, regressing #Bitcoin's price by its Stock-to-Flow ratio (S2F); it's total supply divided by the new issuance. The underlying idea was that scarcity is driving Bitcoin's price.

medium.com/@100trillionUS…

3/n The first model used monthly #Bitcoin data and used silver and gold as benchmarks. Both aligned well, albeit suggesting that the modelled price was on the low-end. As @RaoulGMI pointed out; this already indicated that the method might hold merit for cross-asset valuation.

4/n In the original model, @100trillionUSD conservatively rounded down the numbers for the coefficients to tame down the expectations a bit, ending up with a model that predicted a price of around $55k per #Bitcoin after the upcoming May 2020 halvening.

5/n @BurgerCryptoAM & @btconometrics then attempted to falsify the model but ended up supporting it. I won't go into detail here; these articles are mandatory reading if you want an in-depth look:

1) medium.com/burgercrypto-c…

2) medium.com/@btconometrics…

3) medium.com/burgercrypto-c…

6/n After multiple iterations using different data, there appeared to be consensus favoring the version of the model that predicted a $100k #Bitcoin price after the May 2020 halvening (or people just like round numbers 😅).

Chart by @hamal03 - available: s2f.hamal.nl/s2fcharts.html

7/n The high predictions stimulated the development of more conservative models like the time-based model. Unlike the S2F model, it was statistically invalidated.

@BurgerCryptoAM, Nov 2019: medium.com/burgercrypto-c…

@btconometrics, Jan 2020: medium.com/@btconometrics…

8/n More recently, @btconometrics suggested that the statistics used in prior iterations were sub-optimal. Nonetheless, using a new approach, he came to the same conclusion that there's a non-spurious relationship between #Bitcoin's S2F ratio & price.

https://twitter.com/dilutionproof/status/1246731121150263296

9/n Interestingly, @btconometrics' latest ARDL-based S2F model is actually much more conservative, suggesting a #Bitcoin price of ~$30k after the May 2020 halvening.

https://twitter.com/btconometrics/status/1246066744382246916

10/n In yesterday's piece, @100trillionUSD introduces the concept of phase transitions, and gives the transition of ice cubes to liquid, to gas and eventually ionized water as temperatures increase as an example.

medium.com/@100trillionUS…

11/n @100trillionUSD then suggests that fiat currencies also underwent phase-transitions; from gold coins to gold-backed paper to unbacked paper.

12/n Finally, he cites a July 2018 article by @nic__carter & @hasufl about how #Bitcoin narratives changed (seemingly continuous) over time to come to the hypothesis that Bitcoin might be undergoing one or more phase transitions as well.

medium.com/@nic__carter/v…

13/n Based on that article, @100trillionUSD identifies four overarching phases: (1) Proof of concept, (2) Payments, (3) E-Gold and (4) Financial asset.

14/n @100trillionUSD then used a genetic algorithm that minimizes the distance to identify four clusters of #Bitcoin price that fit the four phases.

15/n In his article, @100trillionUSD suggested further research could focus on different clustering algorithms like K-neirest neighbors. @btconometrics quickly implemented this suggestion to come up with four similar clusters.

https://twitter.com/btconometrics/status/1254998197186879491

16/n Using the four clusters along with the S/F ratio's and prices of gold & silver, @100trillionUSD created a new model: the 'Stock-to-Flow Cross Asset Model' (S2FX). This model iterates by using the S/F values of multiple assets to model price, leaving time out of the equation.

17/n Compared to its predecessors, this model is actually much more optimistic about #Bitcoin's future price, and suggests that #Bitcoin's price will oscillate around $288k per BTC during its fourth 'reward-era' (2020-2024).

18/n @100trillionUSD finishes with 2 notes. The 1st is that the high statistical significance, R-squared (the amount of variance that is explained by the model; 99.7%) and the previously shown cointegration give him confidence in the model despite only 6 time points being used.

19/n The 2nd note describes that the model can be used to interpolate prices within the existing range; up to gold's S2F ratio & price. Under that assumption, the model is particularly useful during this upcoming period (2020-2024), where Bitcoin's S2F will approach that of gold.

20/n From here on some of my own thoughts. First, I am very intrigued by the idea of a cross-asset S2F model and applaud @100trillionUSD for once again coming up with such a novel asset-valuation approach. 👏

However, I do have some critical remarks.

21/n I am personally not surprised that the R^2 value of this model is higher than the previous iterations. By using a classification algorithm to cluster many monthly datapoints inoto four aggregated ones, the actual variance in the used data is more or less explained a priori.

22/n Also, as convincing as the narrative regarding the four suggested phase-transitions is, the choice to actually cluster the data in four datapoints also introduces possible bias, as choosing a different number here may change the results. IMO, more research is needed here.

23/n The model uses four #Bitcoin-based datapoints and just two other sources; gold and silver. Since two thirds of the data are Bitcoin-based, the current model is relatively 'Bitcoin-biased'. Like @100trillionUSD also pointed out, more research with additional assets is needed.

24/n Finally, I find it interesting that @100trillionUSD chooses not to follow @btconometrics' approach at statistical modeling and is sticking to the Ordinary Least Squares (OLS) method. Will the underlying OLS assumptions to this latest model hold up statistically?

25/n The addition of this latest model means that we now have versions of the S2F model that predict #Bitcoin prices in a range of $30k-$288k in the next reward era (2020-2024). Since understanding the complex nuances of the underlying fundamental and statistical (...)

26/n (...) assumptions are very difficult to grasp for the average Joe, I think it is important to emphasize that these model's should not be used as absolute price targets. The way I see it, these S2F models are part of a growing body of knowledge. However, most of the (...)

27/n (...) current evidence seems to be supportive of the hypothesis that #Bitcoin's scarcity, measured via S2F, has been an important driver of its price until now. I personally cannot wait to see what the upcoming out-of-sample tests over the next few years will look like!

Argh, 27 tweets and I still managed to forget something. A live version of the #Bitcoin S2FX model is already available on @digitalikNet's website. Very nice!

digitalik.net/btc/s2fx

@threadreaderapp unroll