History Rhymes; It Rarely Repeats  —  Beware of the Extrapolations…

News Flash: Social Media is a Noisy Place

By   Daniel Masters 3rd August 2018

Temperatures are high, opinions are strong and thousands of voices are shouting at the top of their lungs to be heard.

Add Bitcoin to this and you get the dual dynamite of hand-wavy confidence and statistical creativity prescribed in strong doses according to the general state of the market.

Sidebar: For more on the hand-wavy confidence phenomenon…our CSO Meltem Demirors recently talked with Jill Carlson and Laura Shin about this very thing — definitely worth a listen (link).

When the market is down the bears crawl out of their dens triumphantly roaring how they were right all along: ‘Bitcoin is a Ponzi,’ everyone’s a bunch of idiots and it’s going to zero by year end.

Then when the market swings, bulls come out in force, charging across the field, calling for moon, lambos and bitcoin to infinity. Their ideas are often “backed” by various instances of untested hypotheses and pseudo-science.

We’re not the only ones to have noticed the slew of somewhat low-effort graphs and curves permeating the space with dubious theoretical backing.

…you can almost make a fit for anything…

At CoinShares we focus on the facts, the hard numbers and the long term middle way. Before we go any further, let us be very clear:

In our view — bitcoin is not dying; but it is certainly not going to infinity (unless, of course, you believe the dollar will hyperinflate into infinity (different post/conversation).

To illustrate our point, we feel obliged to point out some recent Medium shenanigans by Pantera Capital.

Now I know Dan Morehead; he and his crew are a level-headed gang and we like and respect them, a lot. They too are pioneers in this space and that should not be forgotten.

We suspect they may be intentionally trying to spark some controversy here by the slight use of hyperbole and feeding into some primordial HODLer desire, but it’s still not an ideal thesis to place in front of the market.

As you can see from this chart, Pantera shows the bitcoin price over the last eight or so years on a logarithmic y-axis — which is, by the way, the only useful way to view that chart — and then adds a best fit linear trend line (on a log scale that is an exponential curve).

From here, they extrapolate a bitcoin price “guide” to the end of 2019. They do not list their R², but reverse-engineering the chart to the end of June 2018 gives us a figure >0.85. A pretty solid fit.

While their curve might end up being an accurate prediction (all predictions are obviously guesses, some more educated than others), we do suspect some eyebrows were still raised among the audience (ours included).

The issue is that extrapolating exponential trend lines — or even extrapolating at all in bitcoin — is a somewhat dangerous business. There are several reasons, here a few of the more critical:

  • Past performance does not imply future performance. Just because bitcoin has grown exponentially thus far does not mean it will continue to do so, this applies to all extrapolations
  • Extrapolating the curve a little further into the future reveals some pretty absurd results (see chart below), suggesting this thesis cannot possibly be correct long term, which in turn raises the question of short term validity
  • Trend lines are only valuable if there is some causal relationship which can be inferred — and preferably modeled — between the two variables. *While we appreciate that Metcalfe’s law could apply here, it cannot grow to infinity either
  • In both intellectual thought and quant/systematic trading models, trended data causes over-weight assumptions and skews logic

To illustrate, we have taken the liberty of extrapolating that “fit” trend-line 10 years forward to show where it would leave us at the end of 2028.

10-Year Bitcoin Price Trend Growth
According to the linear (exponential) fit, the price of a single bitcoin should, by 2028, approach $10bn. For reference, that puts the bitcoin market cap somewhere above $200qd (that’s quadrillion or 1,000 trillion… had to look it up myself).

Unless the Federal Reserve decides to take a page out of the book of the Central Bank of Zimbabwesuch an outcome seems rather absurd.

So what’s the way forward from here?

In my experience in markets and their accompanying cycles, there’s some established frameworks worth considering; these are not predictions, but simply suggestions for a framework of thought when considering extrapolating against bitcoin.

The Lindy Effect — A good example of decaying exponential growth…

Suppose that this curve is also plotted on a logarithmic y-axis. While it at first grows at a similar rate to an exponential curve (linear on log scale) it eventually tapers off towards an upper maximum.

On a linear scale it would translate into an S-curve….

In this model — bitcoin would at some point run into an asymptotic “ceiling” where its maximum attainable value has been, captured. Seems reasonable.

We obviously do not know what this “ceiling” would be, but for the sake of conversation/comparison — let’s use a couple existing “ceilings” we’ve seen postulated:

  1. M1 — also called narrow money; coins and notes in circulation and other money equivalents that are easily convertible into cash
  2. M2 — includes M1 + short-term time deposits in banks and 24-hour money market funds.
  3. Total Global Gold (above ground reserves, in central banks)
  4. Sum of M2 + Gold

Please note that this graph shows bitcoin market cap whereas the price trend graph shows bitcoin price. Sources: CoinShares Research, CIA Factbook, Bitinfocharts, World Gold Council

Which, on a linear scale, looks like this:

This is why linear scale adoption curves are silly. Same sources (and curves) as above.

What’s the point?

Nothing can grow exponentially into infinity. The long-term growth of an asset that has thus far grown at an approximately exponential rate is likely to dampen over time; especially an asset tied to tech adoption.

So how do we know when the approximate exponential growth will begin to abate?…After all — bitcoin has been defying gravity for a while.

As a parallel, one might consider that this decaying exponential curve looks like the top of the classic, technology adoption S-curve (decent explainer video from Andreessen Horowitz on that concept here).

Here is a classic example: the still ongoing global internet adoption curve.

Sources: CoinShares Research, Internet World Stats

Or on a log scale:

Sources: CoinShares Research, Internet World Stats

Look familiar?

If we examine for a moment the curve for internet adoption when it was roughly between 5 and 10% estimated global penetration — we began to see the adoption growth rate slow (visually — the curve flattens horizontally; and obviously 100% is the ceiling for internet adoption).

This is not to say we are at that point in bitcoin — but it is to say, when considering what growth models could be appropriate for bitcoin and which are not — there are prior signals, which from a historical perspective, seem to rhyme with the growth we have seen so far.

To use an overused aphorism…History rhymes, though rarely repeats.

For 15 years prior to launching the first regulated bitcoin fund, my partners and I ran a ran a multi-commodity quant driven trading strategy.

Wearing that hat — here is the real, blunt truth.

The systematic trader in me looks at all trended charts and thinks, quite literally, ‘this contains no information of value.’

I should say, no information of trade-able value; at least not from a systematic point of view. Let me unpack that a bit further:

Without getting technical — when building a systematic model, you train the model to look for causal relationships and react to their associated signal(s). When the signals show up — the model derives (and trades on) insights.

To train and test a model like this, one of the very first steps in the process is to literally, de-trend the data. Why?

Simple — a systematic model is only as good as the data it is trained on. If you train it to work on an ‘up and to the right’ trend — it will always work…on that trend. If you test it on the same trended data, again will always work.

But when the volatility shows up — most gains will be erased; the model is not used to that type of data.

Put more simply — building and training a trading model based on historically trended data, is akin to coaching a baseball team into believing that they are going to start every game with a 7 run lead.

In many ways, a trader (or investor’s) brain is a systematic trading model — so be careful what you train it with.

*Caveat: This is not a debate on the merits of systematic trading versus thematic investing — we’ve run both strategies — it’s simply to further highlight the dangers of relying on trended, historically weighted data for an indication of future events.

Final Takeaways:

  • Annual percentage gains going forward, past a certain point in adoption rate, are likely to slowly abate.
  • Watch your signals and as we progress further, consider the importance of the ‘efficiency’ of any vehicle you may choose when seeking exposure to bitcoin as signals change.
  • History Rhymes; Though Rarely Repeats. Learn from it, don’t count on it.
  • Beware of ‘trends’; for fear of biasing your own internal models.

Check out our website at for our tips on safe and convenient exposure.

Much credit to Christopher Bendiksen, Meltem Demirors and other members of the CoinShares team for the contributions, edits and comments — getting this right is always a team effort.


Please note that this Blog Post is provided on the basis that the recipient accepts the following conditions relating to the provision of the same (including on behalf of their respective organisation).

This Blog Post does not contain or purport to be, financial promotion(s) of any kind. This Blog Post does not contain reference to any of the investment products or services currently offered by members of the CoinShares Group.

Digital assets and related technologies can be extremely complicated. The digital sector has spawned concepts and nomenclature much of which is novel and can be difficult for even technically savvy individuals to thoroughly comprehend. The sector also evolves rapidly.

With increasing media attention on digital assets and related technologies, many of the concepts associated therewith (and the terms used to encapsulate them) are more likely to be encountered outside of the digital space. Although a term may become relatively well-known and in a relatively short timeframe, there is a danger that misunderstandings and misconceptions can take root relating to precisely what the concept behind the given term is.

The purpose of this Blog Post is to provide objective, educational and interesting commentary. This Blog Post is not directed at any particular person or group of persons. Although produced with reasonable care and skill, no representation should be taken as having been given that this Blog Post is an exhaustive analysis of all of the considerations which its subject matter may give rise to. This Blog Post fairly represents the opinions and sentiments of its author at the date of publishing but it should be noted that such opinions and sentiments may be revised from time to time, for example in light of experience and further developments, and the blog post may not necessarily be updated to reflect the same.

Nothing within this Blog Post constitutes investment, legal, tax or other advice. This Blog Post should not be used as the basis for any investment decision(s) which a reader thereof may be considering. Any potential investor in digital assets, even if experienced and affluent, is strongly recommended to seek independent financial advice upon the merits of the same in the context of their own unique circumstances.

This Blog Post is subject to copyright with all rights reserved.


Sign up for our monthly newsletter


Our latest insights & research. Never spam.