# Last edited on 2014-10-23 23:57:54 by stolfilocal # NOT POSTED [quote author=rpietila link=topic=740394.msg8650477#msg8650477 date=1409729628] [quote author=JorgeStolfi link=topic=740394.msg8646696#msg8646696 date=1409702622] You saw the WorldCom plot. After five years of nearly perfect exponential growth, [b]by your reasoning [/b]...[/quote] Did you actually read the OP? :-[ Please revisit my reasoning. Hint: "pregnant with WorldCom" never existed. [/quote] Sorry, I did did not realize that the "pregnancy" discussion was meant to justify the choice of the model. This is the relevant part, I suppose: [quote] Meanwhile, the 2 years average delay from first hearing to first trial, is slowly becoming fulfilled in hundreds of millions of people. There were millions who heard about the bubble in 2011, and their "2 years" started counting. Not enough of them were ready to buy in 2012, which was still a year of slow price appreciation. But in 2013, the great advances were made by the people who had had 2 years to ensure themselves that it was not a fly-by-night. In 2012, the global media coverage about Bitcoin was tiny, but in 2013 all changed. In two years, we are talking about 100x more people in the pipeline of becoming Bitcoin owners. This gives rise to the exponential trendline. [ ... ] Hearing about Bitcoin is the start of your Bitcoin pregnancy. Of these pregnancies, not all perhaps result in a healthy baby (adopting Bitcoin as your home currency), but luckily here, the time is not limited to 9 months. The average with those who currently are power users may have been 3 years (or less, since the delay is shorter with people with the early adopter mindset), but for others it may be 5 years. Miscarriages happen often, but the reverse-miscarriage is also a very typical occurrence. We have a world-record number of Bitcoin pregnancies going on now, and hardly any births for several months already. I believe that the laws of nature and mathematics will take care that the babies will be seen in increasing numbers. [/quote] See if I understood it: "the growth of bitcoin happens in spurts (bubbles) separated by stangnation periods, because there is a delay of months or years between hearing about it and buying it; and each bubble is a new generation of adopters who have heard of it in a previous generation." [quote]That a professor is trying to label my investing over the past 21 years when I started using other than saving accounts for risk management, "luck", is lamentable. [/quote] I never said anything about your previous investment activity. Your methods may have been appropriate for other markets I don't know. But your "pregnancy and bubble" theory for bitcoin does not make much sense. First, I am not aware of any technology that has had such a stream-of-bubbles adoption curve, caused by a "pregnancy" delay. The theory does not make logical sense: if the delay is variable, as you assume, then each bubble would be substantially stretched out in time relative to the previous one; and pretty soon the bubbles would merge into a smooth continuous growth. Yet, in reality, the November bubble was just as sharp as the previous ones, or even sharper than them. [b]The prototypical bubble[/b] Usually there is no such "pregnancy period" in the spread of a new "hot" technology item. Usually there is a single bubble, which, in my understanding, follows the general pattern below: [img width=500]???[/img] In the initial "exponential growth" phase (E), as soon as people learn enough about the novelty to see its advantages, they buy it and start using it. As more people use it, it gets more exposure (peer-to-peer, in the news, in paid advertising); it becomes more useful and easy to get; and it gets improved and better packaged. All these developments in turn increase the number of users who find the item desirable and want to get it. Because of these feedback mechanisms, [b]utilitarian demand[/b] -- that is, people acquiring the item for some actual use -- usually grows roughly exponentially for a while. If the item is tradeable, this exponential growth is often enhanced by [b]speculative demand[/b] (people buying the product because they see that the price is rising, and hope to make a profit selling it later). However, the exponential growth eventually slows down and stops, somewhat abruptly, when the market is saturated (the "S" phase in the diagram). By that time the speculators often have overbought; as they realize that the utilitarian demand stopped growing, they sell their holdings -- as fast as they can, to get the best price possible. The surge in supply overwhelms th eutilitarian deman and causes the price to crash (phase "C"). Once speculators sense that the item is undervalued, they rush to buy it back, Thus the speculative demand oscillates for a while (phase "T") until settling at some steady-state, nerly constant demand (phase "Q"). The latter is mostly utilitarian, but includes some speculative demand by day traders who exploit random variations in price. The demand then may be stationary or grow slowly, e.g. because of "vegetative" market growth; or may decay, e.g. because of competition by other items, government bans, etc.. This generic single-bubble pattern seems to describe the evolution of demand for most new technology items items, and fits each ONE of the past bitcoin price bubbles. [b]Causes of the bitcoin bubbles[/b] So, why did bitcoin's history have several well-separate bubbles, rather than a single bubble? My explanation is that each bubble was caused by bitcoin entering a new market, and saturating it. Each market was a different population with different goals and needs, isolated from the other markets by differences in purpose, culture, geography, etc.. Therefore, the enthusiastic adoption in one market was not even noticed by the others, until they found that bitcoin was useful for their own needs to. My guesses as for the markets responsible for the past bitcoin bubbles, perhaps not in order, are * Computer nerds and programmers who worked in security, distributed systems, cryptography, payment methods, etc., mostly in the US and Europe. They adopted bitcoin because they understood the problem, understood the technical details of Satoshi's solution, and liked it as a cool piece of technology that they enjoyed hacking around; * Libertarians in the broad sense (anarcho-capitalists?). They adopted bitcoin because they thought that it could be the currency of a "libertopia", a community of libertarians outside the control of govenrments and banks. * Users of illegal drugs, attracted by the opening of bitcoin-based internet drug bazaars. They adopted bitcoin when they realized (erroneously) that its payments could not be traced. * Con artists. They adopted bitcoin when they realized that, if they started a bitcoin-related enterprise and asked for investments in bitcoin rather than dollars, they could avoid the checks and barriers of national financial watchdogs, and then steal all the money without fear of going to jail. * Chinese computer nerds. They adopted bitcoin for the same reasons as the Western colleagues, but discovered it much later. * Chinese amateur commodity traders. They were used to trade all sorts of items, and adopted bitcoin because it had three advantages over fermented tea bricks: (1) its price was a lot more volatile, (2) they could trade it even without understanding what it was, and (3) all bitcoins are exactly alike. I am not claiming that all these communities generated visible bubbles, nor that they "bubbled" in that order. But these communities are clearly isolated from each other, socially and even lingusitically, so that a bitcoin craze among libertarians (say) would barely attract the attention of the bulk of the drug users (say), until these realized that it could be good for their purposes, too. There may have been other communities responsible for smaller bubbles that are not easy to distinguish among the larger ones. The rise from 120$ to 200$ during Oct/2013 may have been a separate mini-bubble, due to adoption and saturation within some community distinct from the one responsible for the Nov/2013 bubble. Each of these episodes of adoption followed the standard adoption curve: roughly exponential demand (and therefore price) increase until saturation, overshooting by speculators, a few oscillations, and finally a steady state regime of nearly constant price. However, since each bubble was [b]added[/b] to the stady state demand of the previous bubbles, the rapid growth phase does not follow a pure exponential B**(t - t0), but rather a [b]shifted exponential[/b] A + B**(t-t0). (While a pure exponential looks like a straight line in a log-scale plot, a shifted exponential will still look like a shifted exponential at first, which gradually straightens out.) Within and across these bubbles there was also a demand by long-term investors (like you perhaps) who bought because they believed in substantial price rise over the long term. [b]The June 2011 bubble[/b] The Jun/2011 bubble was atypical because its steady state was not a contant plateau but a relatively fast, roughly exponential decay -- as if that particular community gradually abandoned bitcoin. The decay was interrupted at ~2$ in Nov/2011 by the start of another (smaller) bubble, that lifted the price to ~5$ and kept it there until Jun/2012. It seems possible that, without this second bubble, the price after the Jun/2011 bubble would have decayed to the pre-bubble level of ~1$. [b]The 2013 Chinese bubbles[/b] I hope it is beyond question now that the Nov/2013 bubble was due to adoption by the Chinese amateur and semi-professional traders, following the opening of exchanges in the mainland (such as OKCoin and Huobi, that opened in Beijin on 2013-06-12 and 2013-09-02, respectively). I am almost convinced that the Apr/2013 bubble too was due to adoption by some segment of the Chinese population. It started only a couple of months after Bobby Lee took over the management of BTC-China in early 2013. He attracted some investors and, having worked for large companies, perhaps he had the marketing intuition that the previous owners lacked. The community in question could have been Chinese computer nerds in Shanghai, and their recruitment may have been connected to the appearance of Litecoin -- created by Charlie Lee, a former Google employee and Bobby Lee's brother. The Apr/2013 bubble started at ~14$ and, by early Aug/2013, appears to have settled to about ~100$. It seems that another small bubble started at that time and settled to ~140$ by mid September. Yet another bubble started in early October and was about to settle at ~200$, when it was overrun by the Nov/2013 one. Apparently it consisted almost entirely of short-term speculative demand, with little utilitarian demand. According to media reports from late 2013, China had a large population of amateur and semi-professional traders who, lacking access to the stock market, were used to day-trade on bizarre commodities like fermented tea bricks. When the mainland Chinese exchanges opened, those traders switched enthusiastically to bicoin, because of its fungibility and high volatility, without caring for its potential as payment system etc. In any case, this utilitarian potential was squashed by the government decrees of mid-December, which banned bitcoin from internet payments and the financial sector. Perhaps for being mainly speculation-based, the Nov/2013 bubble did not reach a constant-price steady state, but instead ended with a roughly exponential decay (like the Jun/2011 one), until it was interrupted by the May 2014 mini-bubble. This decay was broken by a few sudden jumps and drops, by 100--200$, which were clearly corrleated with the MtGOX events and various news and rumors about decisions of the Chinese government about bitcoin. The steady price of ~400$ during most of May/2014 may have been the result of a joint pledge by the Chinese exchanges to make the market more friendly to amateur traders. The last "bubble" started at the end of May/2014, and did not follow the general bubble pattern. Instead of an exponential onset, the price suddenly started to rise from ~400$ on 2014-05-20, in a sequence of 4--5 intense buying spurts, until it reached ~650$ by 2014-06-06. After some oscillations, by July it seemed to have stabilized at ~620$; but then it started an accelerating drop, to the current ~350$ level. Considering of its anomalous start, this bubble may not have been the opening of another broad market. Perhaps it was only a handful of individuals who, for some reason, believed that another bubble was imminent. Or perhaps some investors with high stakes, worried by the price decline, tried to "jump-start" the next bubble with some large buys. [b]Separating the bubbles[/b] The plot below is a crude attempt at separating the bubbles seen so far. Each bubble is represented by a separate price chart B(1,t), B(2,t), ..., B(n,t), such that the historical price chart P(t) the sum of those n individual charts. The split was generated by the following procedure. It worked on a residual price chart Q(t), initially equal to the historical chart P(t). The first bubble was identified in Q(t) by the roughly exponential increase from near 0$. Each bubble was assumed to follow the general pattern above, until reaching a steady-state phase where the price was either constant or (in some cases, such as Jun/2011) a slowly decaying exponential. This steady-state phase was usually interrrupted by the next buble. The chart B(k,t) was then set to a copy of the the price chart Q(t) until that time, followed by the extrapolated steady-state price, up to the present time. This price chart B(k,t) was then subtracted from Q(t), and the procedure was repeated. [img width=600]??[/img] This decomposition implicitly assumes that demand is linearly related to price, so that a bubble B(k,t) that resulted in a sustained increase of 100$ in the price would have increased the price by 100$, no matter what was the starting price. This assumption is probably incorrect, but it is harmless for this purpose, since the only observable quantity is the sum P(t) of all bubbles. In any case, a non-linear relation between demand and price would only change the amplitude of the bubbles, with little effect on their shape. [b]Long-term evolution[/b] A key point of this theory is that each of those markets is now saturated, so further substantial growth in demand depends on the opening of new markets. The error in the prediction was chosing a straight linein the log-scale plot. That choice made the conclusion of continuing growth inevitable. It is important to realize that this conclusion is not a consequence of the data, but was predetermined by the choice of the model. See the example I posted earlier, where the choice of a particular mathematical formula -- a quadratic function -- was guaranteed to predict a crash, even before being fitted to the data. For the claimed long-term exponential trend to continue, a new market must open every six to twelve months, each an order of magnitude larger than the one before. Considering the disparate nature of the markets, that expectation has no rational basis. It has been conjectured that eventual approval of the COIN ETF could open a huge market in the US. Others hope that alarge demand may come from Latin America or Africa. However, the fact that the Chinese market was large enough to lift the price from ~200$ to ~800 does not make those future events more likely, and does not imply that they will be even bigger than the Chinese market. Indeed, the hypothesis that each bubble is much larger than the one before may not have been true in the past either. Since each bubble adds to the demand of the previous ones, if two small bubbles of the same magnitude are separated by a large one, the second is much harder to notice than the first. For example, the bubble that lifted the price from ~0.07$ in Nob/2010 to ~0.25$ by Jan/2011 is quite conspicuous in the log-scale chart, because it occurred when there was no demand. If a market with the same total demand had opened in Apr/2012, when the price was ~5$, it would perhaps have lifted the price by less than 1$, which would have been indistingushable from a random fluctuation. Thus, it seems more reasonable to assume that bubble come in various sizes, mixed large and small; but we only [i]notice[/i] another bubble when it is much bigger than the previous one [i]that we noticed[/i]. In summary, the fitting of an exponential over all bubbles is unwarranted, because it [i]assumes[/i] from the start that markets will continue to open at a regular pace, each much larger than the previous one. But there is no economic or social force that will ensure that. What seems to be happening is that the Nov/2013 bubble, like the Jun/2011 bubble, had a short steady state (Jan/2014, 800$) followed by a general decaying trend. My guess is that the Chiense speculators are geting discouraged and, since Feb/2014, have been gradually dropping off. By this theory, if nothing changes in China, then the demand from the Chinese market will continue to decay exponentially. If only the Nov/2013 market is affected, the limiting price should be 200$ or so. In any case, the decay rate seems to be 50% drop every 6 months or so; at this rate, 200$ would be reached only in 2015. If the decay affects also the Oct/2013 market, the limiting price may be around 140$. If it also affects the Apr/2013 market, the limiting price may be below 20$. By this theory, if the causes of the Chinese market decay are not changed, a reversal of this trend is possible only if some [i]other[/i] substantial market opens up. If the COIN ETF gets approved, it may open a new large market in the US, perhaps.