As I mentioned to you earlier, I have 10% of my networth invested in bitcoin (bought at the 45,000 level). I am still waiting for your feedback about PlanB : Stock-to-Flow model and Ben Cowen : Algorithmic Regression Rainbow (both well known and respected in the crypto space) which I am following right now.
For members here, BITO and GBTC are now covered by P123 although there is little history to backtest them. I hope P123 can upload the history of GBTC and ETHE in the near future.
Scarcity. I first became aware of this in this (P123) forum from a post by MisterChang who invests in BitCoin. I agree this seem legitimate. I have done nothing with this on my own.
Non-linear regression. I have done a few things with non-linear regression methods but nothing with BitCoin. I agree it can work well for certain problems. But again I have done nothing with BitCoin using this method.
While it does seem like this is a legitimate idea worthy of investigation, my limited experience does not give me enough knowledge to add anything to the links you provide. BitCoin is not an area of interest for me at this time. Generally, I think BitCoin is a difficult problem and I certainly cannot help anyone as far as BitCoin strategies at this time.
I read over Plan B’s stock to flow model, his idea’s on scarcity and EMH. They’re reasonable assuming the demand for Bitcoin is constant, which is something that he does not address. This could be a critical flaw in his model.
I couldn’t find much written on Ben Cowen : Algorithmic Regression Rainbow without watching the videos. Do you have a link to text?
GBTC can be used in books but not simulations. BITO has less than a month of history. I asked P123 about expanding the access to Bitcoin for simulation back testing and was told that it will be done but it’s not clear when. PortfolioVisualizer.com has gbtc in its database so can be used in asset allocation. Have you found any sites where you can back test Bitcoin (I’ll probably code it eventually when I have the time if I can’t find one)?
That’s very brave putting 10% in Bitcoin as it’s 10X more volatile than stocks
You can check out this link for a comparison between PlanB vs Ben Cowen. Algorithmic Regression Rainbow differs a from Stock-to- flow and actually predicts a decreasing return as the bitcoin cycle gets longer and it will take increasing more capital to push bitcoin higher comparing to a few years ago.
You may also want to take a look at this paper about technical analysis on cryptocurrencies. This paper suggests that we use trending following indicators to trade bitcoin which greatly reduces max drawdown and improves the Sharpe ratio (which is already high for buy-and-hold). The authors also employs 4 major methods to prevent overfitting in achieving the backtested results for using these trend following indicators.
Based on the above, I don’t use a buy and hold strategy for bitcoin, I am now using weekly Parabolic SAR (0.018,0.09) + 7 week DMI (above 35) to confirm the trend. (which is not mentioned in the paper but the concept is similar). I am also watching the Algorithmic Regression Rainbow and will reduce half of my bitcoin holdings if it reaches overbought signal.
For your info, Jim (who has an active subscription to Portfolio Visualizer) has kindly helped to run a simulation and the best combination to maximize sharpe is to hold about 20% GBTC and 80% TLT (if only these two assets are in the portfolio) since GBTC inception. The max drawdown is about 20% (about the same as SPY)
I guess this is a low risk strategy to invest in bitcoin but the potential return cannot be compared to having a directional risk-on/risk-off view with trend following indicators
Thank you for that link on Cowen and the paper on crypto technical trading. It’s interesting that Cowen used a logarithmic regression model as that is usually used for binary variables where a linear regression model is more commonly used with price data. It looks like he fit a line to estimate where the price of Bitcoin should go. However the disclaimer on the site you linked was that this model will be true until it’s true no more. I view this line more similarly to technical trading and the stock to flow model as more of fundamental investing as it tries to value Bitcoin. Both of them could be self reinforcing if enough people believe in them. The weakness of the first model is what is fundamentally making Bitcoin follow this line rather than deviating some time in the future? The weakness in the second model is as I mentioned before it assumes constant demand.
I skimmed the trading paper but will read ii in more detail later. However this line was very informative, " Finally, we show that technical trading rules cannot generate positive returns in the out-of-sample period for Bitcoin, but can for other cryptocurrencies.
Therefore our results demonstrate that technical trading rules have significant predictive power in cryptocurrency markets even after accounting for multiple hypothesis testing, but Bitcoin does not offer any predictability in the out-of-sample period. " So maybe the asset allocation model that you proposed using tlt and gbtc on portfolio visualizer (thank you for sharing) is the optimal model
I agree with you that both stock-to-flow and algorithmic regression can be self reinforcing as more and more people, especially social media (sentiment) and more retail/institutional buying towards those price targets.
Regarding the out-of-sample issue, I have gone through at least 3-4 more academic papers which confirms the results using trend following indicators for bitcoin. I guess maybe the time frame that is used in this particular paper does not favor bitcoin and only for other cryptocurrencies. I have attached 2 more papers below for your reference (with more out of sample data).
For Quantconnect, as far as I understand it is kind of similar to Quantopian and you have to do the coding yourself to build a model. (unlike Portfolio Visualizer).
Thank you for those additional papers and information. What was informative was that one of the papers decreased the max drawdown from 89 to 64 % using a 20 day ma. Using the 80 % tlt 20 % gbtc with frequent rebalancing (1 wk - 4 wks ) decreases the drawdown to 30-35 % however also decreases the returns. One probably could decrease this DD further by exploring other technical indicators however testing many indicators over a small sample size with optimization increases the risk of curve fitting.
The max drawdown for bitcoin (GBTC) will always be more significant than stocks no matter how you try to manage it with technical indicators or more frequent re-balancing between GBTC/TLT. That is the reason why there is higher return and as long as there is a higher sharpe/sortino which is further improved by investing with a trend following indicator. There really is no point to continue look for ways to reduce the drawdown which increasingly lower the potential return.
My view is that one should not be investing in bitcoin unless he/she can stomach the volatility that is a little bit higher than stocks. As I have mentioned earlier, I use another trend following indicator, weekly Parabolic SAR (0.018,0.09) +7 week DMI (ADX above 35) to confirm the trend instead of simple moving average that is mentioned in the paper.
I believe in the saying “Fortune Favours the Brave” (to a certain extent) in taking calculated risk.
Actually there is another way to play BITO and GBTC which is to long GBTC (basically trading at a 17% discount to NAV) and short BITO for an arbitrage play (to pocketing the contango roll BITO bleeds about 13% per year due to the futures roll ).
Jim,
There is also a tax benefit which I am not too familiar with since I don’t live in US but you can check out this reddit link.
Short Bito in a taxed portfolio and long GBtc in Roth IRA giving a risk neutral 11% yield. If Bitcoin Runs you get a massive tax write off, you move value into your IRA, and you don’t have to pay taxes on the risk Neutral yield. And as a bonus you get an extra 17% or so when SEC allows GBTC to convert to an ETF in the next couple years.
I really should do this or something similar. I think I could probably hedge occasionally in a taxable account and write off the losses for stocks and ETFs (including any short positions). Not limited to GBTC, perhaps.
My present accountant is still trying to figure out why we had to pay taxes on profits from a MLP in my SEP-IRA But I think this is a fairly standard thing in the US that she could do, and that I should do. Excellent point, I believe.
The amount you can write off each year may not be huge: other members will know more than I do. The losses can be extended into future years however and is not trivial when considering this (again, if I understand this at all).
Thank you for the arbitrage idea. I have been looking at options to test securities with some correlation to crypto on P123 until gbtc is enabled in sims. The closest, which aren’t that close (similar to gold vs gold miners), are the ETF Blok and this list of stocks (coin,si,mstr,sq,pypl,riot,mara,ostk).
I agree with you regarding the volatility of Bitcoin works both ways (higher drawdowns & higher returns). It’s like a stock ETF leveraged at 10x
As mentioned in my email, FEAR is back in the market due to the new COVID mutant and VIX is currently up 40% in pre-market. I am not sure it is going to be a happy Thanksgiving for most investors.
Bitcoin is falling together with the overall market, I will take profit on my position if it drops below 52,000 (based on weekly parabolic SAR).
I just bought some VIXY at market opening with this simple strategy based on VIX with just 2 buy rules daily rebalancing. (just buy VIXY when VIX rise more than 7.5% from previous close) Rather than losing money, I hope everyone can make some money out of the FEAR factor.
But I think this is a fairly standard thing in the US that she could do, and that I should do. Excellent point, I believe.