On second thought I believe that my conclusions about low correlation between SPY buy&hold and Sector Rotator are premature. In my previous post I have been comparing the Sector Rotator realized profits with the SPY buy&hold unrealized profit. This is apples and oranges. A better comparison is when combining the Sector Rotator's realized profit with its portfolio value, and comparing that against the SPY buy&hold portfolio value. This gives a different picture: both lines (blue for SPY, red for Sector Rotator) now have a much more similar trajectory, indicating a higher correlation. Especially during 2020, from April onward, are both lines pretty similar. 2021H1 shows some larger differences between the two.
Not so. Your journal is very relevant to anyone who is on the Systematic Trading journey and has a smaller account size (i.e. me!). After finding Rob's (@globalarbtrader ) journal some months ago, I have just found yours. The important things for me here are your reasons for making various changes to the system to adapt for a smaller account size. I was very interested to see: how high (relative to account value) you are running your margin. My system currently runs at about 12% margin to account size, but my system is currently neutral gold, oil and most of the grains, so I expect that this percentage will increase, maybe to about 20%; and a parallel system of ETFs using what I would call a Relative Strength system. I have something similar, but it isn't live as yet. It will be improved after reading of some of your experiences. Thanks for the journal. KH
Thank you for your kind words @KevinBB . I don't regularly update this journal with new information about the futures trading system. But I do follow Rob's journal (and his blog) and if I can contribute to the conversation I'll write something there. About margin usage: today my system has a 33% maintenance margin, and 40% initial margin requirement. With 33% I mean that if the account size were 100 k USD, the margin requirement is 33 k USD. This Sector Rotator ETF system is based on the futures system: it uses the exact same set of formula's to calculate the forecast for the ETFs.
This reminds of a strategy from the "stocks on the move" book. Maybe it should work better with individual stocks.. also probably better to use "total return" S&P which includes dividends, but then you didn't include dividends into the strategy curve as well, so I guess it evens out.. With single stocks dividends and splits become a headache though..
I had the same, intuitive, thought. Within each industry sector some companies' stock perform better than others. However, if I were to change from ETFs to individual companies the system would become more complex: more tickers to monitor (e.g. for each ETF its top 5 constituents, resulting in 80 companies), possibly more trades to execute, and measures are needed that the system does not converge to only companies from one industry sector. I didn't want those complications, which is why I decided to stay with ETFs.
Hi HobbyTrading. Just letting you know I have read this entire thread and find it endlessly interesting - particularly as I am following the same path as you (but a long way behind). So thanks for keeping it going! In fact my aim is to have a diversified trend-following strategy embracing (in theory) all liquid markets, and a separate ETF rotation strategy- so it is good for me that you have written about that one in greater detail. Re the comment above on the momentum strategy and individual stocks, personally I couldn't get that to work profitably (based on Stocks on the Move). Also that strategy has a monthly rebalance, and I believe you are doing daily rebalance - so you had more agility to respond to March 2020. Have you tested a lower frequency rebalance BTW? Another question if I may - if I understand things right, for your futures strategy you deal with smaller capital size by picking the best markets in each cluster (I forget the measure you use - maybe liquidity). So by necessity you are missing out on lots of markets. Have you missed out on any big trends due to this? I am not sure if this is a pure trend following strategy or not, but I believe a principle of that is to be exposed to as many markets as possible to maximise your chances of catching the big trends.
Thank you for your compliment. I do not have a system which uses a monthly rebalance. I do not know about "Stocks on the Move". The sector ETF system uses a daily review and selects the top 5 of the 15 ETFs. But once an ETF is in that top 5 its position size (number of shares) will not be modified until the moment it drops out of the top 5 and gets closed. The markets are ranked by value volatility per contract: the market with the smallest volatility gets the highest rank. In formula (instrument's currency is ABC): (value volatility [USD]) = (price volatility [ABC])*(multiplier)*(exchange rate[USD/ABC]) The price volatility is based on a 25 day look back period of daily close prices. I'm pretty sure that because of limited funds (and thus limited number of markets) I have missed out on some other markets. But I'm not rally interested in doing a "what-if" analysis, simulating the case when I would have had a 100x the account value. I don't have that capital, so it is an unreachable situation.
You don't need to know about Stocks on the Move because you have strategies working just fine! But FYI it's by Andreas Clenow who previously wrote a popular trend-following book and it's an approach to capturing momentum in S&P companies in order to beat that index. I'm working through his newer one Trading Evolved as a way into learning python and zipline (previously everything I built was in Talend (java-based ETL tool) which I think you might have seen in my journal). I'm also reading Robert Carver's Leveraged Trading as a way into his earlier Systematic Trading. I think you have written your code in java? How has that decision gone in retrospect? Personally I still expect to write my automation in Talend/java, but I have learned python for the backtesting.
I have only read "Systematic Trading" and used its framework to build the futures trading system. Later on I made some modifications, but in its core it is still as is described in that book. And I have copied the way how a "forecast" is calculated for a futures instrument into the ETF and stock trading systems I use. Yes, I only code in Java. I don't have a background in software programming. When I started using Interactive Brokers (2014) Java was one of the few languages it supported. At that time Java seemed to be the preferred language. I wanted to write trading software, so I had a clear goal to learn programming. My programming skills (if any) have developed along with my various experiments with automated trading. As Python is nowadays the preferred language I sometimes try to understand code presented by others (e.g. on YouTube channels). But somehow I can't wrap my head around it and get lost after only a few lines of code. I can do what I need to do in Java, and there is no pressing need for me to learn any other language.
Intermezzo: The engineer in me likes these numbers: The round number of posts (3,000). And the "funny looking" 1234 likes. What a coincidence to have this combined today. Back to the regular show.