| Index | Recent Threads | Who's Online | Search |
|
New Thread |
|
ustonapc
|
Jim, 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. https://twitter.com/100trillionUSD https://stats.buybitcoinworldwide.com/stock-to-flow/ https://twitter.com/intocryptoverse https://www.blockchaincenter.net/bitcoin-rainbow-chart/ 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. Regards James |
||
|
|
ustonapc
|
Jim, Many thanks for your feedback. Regards James |
||
|
|
sgmd01
|
Hi James, Thank you for the links 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 Scott |
||
|
|
ustonapc
|
Scott, 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. https://beincrypto.com/planb-and-benjamin-cowen-discuss-the-future-of-bitcoin/ 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. Regards James 
Technical Analysis and Cryptocurrencies.pdf
(522079 bytes)
(Download count: 9)
|
||
|
|
ustonapc
|
Scott, 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 Regards James
GBTC+TLT.png
(205492 bytes)
(Download count: 227)
|
||
Edit 3 times,
last edit by
ustonapc
at Nov 15, 2021 8:04:51 PM
|
|
sgmd01
|
Hi James, 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 was looking for another site to test trading rules on Bitcoin and found: https://www.quantconnect.com/ It looks like it will take a bit of effort. Have you tried this site? Scott |
||
|
|
ustonapc
|
Scott, 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). Regards James
Bitcoin Predictability and Profitability via Technical Analysis.pdf
(1647750 bytes)
(Download count: 17)
Optimizing Algorithmic Strategies for Trading Bitcoin.pdf
(720538 bytes)
(Download count: 10)
|
||
Edit 1 times,
last edit by
ustonapc
at Nov 16, 2021 2:56:18 AM
|
|
ustonapc
|
Jim & Scott, This link on call options expiry for bitcoin looks interesting : The Bitcoin bulls are betting big, with large open interest clusters at strike prices of: - $100k (OI = $500M) - $120k (OI = $420M) - $200k (OI = $380M) https://twitter.com/glassnode/status/1460901197905227783 Regards James |
||
Edit 1 times,
last edit by
ustonapc
at Nov 17, 2021 8:36:30 AM
|
|
sgmd01
|
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. |
||
|
|