A Study Of Bitcoin Price Crashes
“A picture is worth a thousand words” (English language-idiom)“History doesn’t repeat itself but it often rhymes” (Mark Twain)"Don't Panic" (The Hitchhiker's Guide to the Galaxy)
After reaching a high of almost $20,000 in Dec 2017, bitcoin crashed to a low of just under $6,000 in Feb 2018 (using intraday data from the Bitstamp exchange). That’s a drop of around 70%. Large enough to kill almost anything, and providing a good opportunity for bitcoin critics to declare that this time it’s really over for sure and bitcoin is finished.
However for anyone involved in the industry or following it closely, bitcoin appears far from finished. Bitcoin bulls are projecting a price of anywhere from $25,000 to $250,000 over the next 1 to 4 years, and the fundamentals have never looked better. Regulatory risk has been significantly reduced, the scaling debate has been put to rest, bitcoin survived a hard fork, a 2nd layer scaling and privacy solution called lightning network has been released in beta, infrastructure (exchanges, wallets, custody services) continue to grow and improve, and the “digital gold” narrative is spreading. Those in the space think that far from being finished, bitcoin is only just getting started.
In this article I want to take a look at the history of all past bitcoin crashes and see if they can provide perspective on the current one. For the purposes of this article I will be using the word “crash” to mean a drawdown of 25% or more.
The first step was to get a list of all significant crashes, with details such as high to low drawdown, start and end date, and how long it took for the price to eventualy recover. Instead of trying to do this manually or relying on lists compiled by others, I used python to do all the hard work and generate results from scratch. The iPython code is available on Github in the drawdowns.ipynb file. I used daily close prices going back to 2010 from Quandl. Bear in mind that some of the dates and prices would differ slightly by using data from one exchange versus another, or using intraday data versus daily close data. This won’t affect the overall picture.
The function that finds the drawdowns is shown below. It takes a list of prices and dates, and recursively finds all drawdowns starting with the largest and working its way down. I’m specifically mentioning it here since I was surprised by not being able to find something readily available elsewhere that does this (and also pleased to have finally found a good use for recursion).
In plain English it goes something like this. First find the largest drawdown in the entire price series. Then create two new smaller price series, one for the period before the drawdown just found and one for the period after. For each of those two new price series find the largest drawdowns, and then for each one create two new price series. Keep repeating this process until you can’t find anything more of significance. At each step the number of price series being searched for the next biggest drawdown doubles. From 1 to 2 to 4 to 8 to 16.
The table below shows all drawdowns of size 25% or more that were found using this method. The largest is a 93.5% drawdown in 2011 which lasted 164 days from high to low. It then took 469 days for the price to recover back to the same level as the high before the crash.
Looking at when each crash started, there are 3 in 2010, 3 in 2011, 2 in 2012, 2 in 2013, none in 2014 and 2015, 1 in 2016, and 3 in 2017.
In order to try understand the specific price patterns in each crash, I generated price charts for each crash and recovery. The chart below is for the same 93.5% crash in 2011. The first red cross shows where the high price occurs, the next cross shows where the low price occurs, and the final cross shows where the price recovery occurs.
Notice how the price literally takes off once the previous high was recovered. This pattern will repeat in almost every one of the crashes examined.
What follows is a copy of each price crash chart oldest to most recent. The story in this case is best told by the charts themselves (you recall the “a picture is worth 1000 words” part). It’s worth spending some time looking at each chart. After that I’ll discuss my own conclusions. If your eyes are like mine then the charts are best viewed on a larger screen (I’ll attempt to make them more mobile friendly soon).
The last chart shows the recent Dec 2017 crash. As mentioned above for this study I used data of daily close prices from Quandl which puts the low point on 6th April (just 19 days ago). Naturally there is no recovery date yet. If you use intraday prices instead then the low occurred earlier on 6th Feb which is 78 days ago.
Here then are my conclusions so far:
- In many of these large crashes it would have felt like this is the end. In each case the work being done to keep the bitcoin network secure and to grow the infrastructure and ecosystem continued. I recall in 2014 being amazed at how many new companies and initiatives were starting up daily. I couldn’t keep up with the news. If you were only looking at the price you would have thought it was all over.
- When price recovery is finally achieved this is usually followed by a large subsequent run up. Perhaps this is the point where people realise that bitcoin is not dead as they had previously thought. Headlines proclaiming bitcoin is at a new ATH (all time high) could also be responsible for an increase in speculative buying which then contributes to the next high and crash.
- Recovery from a crash can take time. It took 469 days after the 2011 crash, 772 days after the Dec 2013 crash, and 123 days after the Apr 2013 crash. The shortest recovery to a crash bigger than 50% was 37 days after the Sep 2010 crash of 65%. This crash happened over a period of just 2 days so was more of a “flash crash” and bear in mind it happened in 2010. Using intraday prices we are so far only 78 days from the low. Using daily close prices we are only 19 days from the low. History has shown price recovery could still take a lot longer.
- The low for the current crash might not even be in yet. There have been 3 crashes that were larger in magnitude than the current one. The Dec 2013 crash was 84% (based on closing prices), the 2011 crash 93%, and the April 2013 crash 71%. Based also on closing prices this crash has “only” been 66%. History has shown that we could still go lower before the current crash is over.
- Some of the relatively smaller crashes (25% — 35%) happened quickly and recovered quickly. I don’t think the current crash is one of these.
- As a next step I intend to “zoom in” on the days leading up to each crash and each bottom to see if any patterns recur that could be used to try identify these stages in the future.
- Bitcoin is often likened to the NASDAQ. As a comparison and to learn more about what happened with the NASDAQ I have repeated the same study using NASDAQ prices going back to the 70’s. The results of this and whether it sheds any light on bitcoin will be shared in a follow up article.
If you enjoyed this article or got any benefit please clap so that others can find it too.
Please note that nothing in this article should be taken as investment advice. If you got this far then you probably already realised that Bitcoin is extremely volatile. Before investing in Bitcoin or anything else please do your own research.
