Thank you for your interest. I stopped the regular updates as it was only useful for myself, and not for others. The graphs and numbers became confusing and difficult to understand if you don't know the nitty-gritty details, which I was not willing to provide in this public forum. I have continued to run this futures trading system ever since I started it in Q4 2016. So I am now running it for four years. Although I must admit in hindsight that for a couple of years it was underfunded and thus limited in its potential. This year there has been a major change: in May I decided to fund it more properly. Since then is it able to have about 20 ~ 25 open positions at any one time (instead of only a handful). And, more importantly, it shows a nice low correlation with a regular 60/40 equity/bonds investment portfolio. Note to self: this reminds me: it might be appropriate to repeat the Efficient Frontier simulations (which I did in May), using the actual daily returns since that moment. To verify whether my choices were indeed the correct ones.
Thanks @HobbyTrading Great to see you continuing the journey. I myself are running 2 systematic strategies through risk parity & volatility trading and am keen to add trend-following in futures to my portfolio. So your thread is a good read and reference. Just curious, what's the % of portfolio (or % of risk budget in futures context) you dedicate to this strategy?
The target annual volatility is 25% of the account value. That risk budget is evenly spread over six asset classes. Within each class is it subdivided equally over several instruments.
The end of the year is a good moment to look back and see what has changed. This log has been about my futures trading system, and analyzed this system in isolation. This changed fundamentally for me during 2020. My main investment is since a number of years in a basket of ETFs, which I use as my "retirement fund" and from which I withdraw some funds every now and then to pay the bills (I am happily unemployed). The futures trading system was more like a hobby project on the side, meant as a learning vehicle. It has served me very well as such during the previous years. In March we experienced the strong dip in stock prices worldwide due to the Covid-19 pandemic. My ETF investments were not immune to this. This made me investigate how I can use the futures trading system to my advantage, and how it can complement my ETF investments. So, instead of seeing the futures trading system as a side project it now became an integral part of my total investments. The ETF basket and the futures trading system have a low correlation. I determined the historical daily return (in %) for each of them and ran some Monte Carlo simulations which gave me the "efficient frontier". Such a graph looks like this: The horizontal axis is the expected volatility of the account value (%), and the vertical axis is the expected relative profit (%). The actual values for the volatility and expected returns depend on many parameters. As those are not relevant for the remainder of this post have I omitted these values on the horizontal and vertical axes. Point A is where all available capital is invested in the ETF basket and none in the futures system. Point B is where all available capital is invested in the futures system and none in the ETF basket. Point B gives a much higher expected yield, but this comes at a much higher volatility (point B is much further to the right than A). As an indication: in my case is the volatility in point B about five times larger than that of point A. Over the last couple of years have I gotten used to the volatility of the ETF basket and feel comfortable with this. However, as this investment forms my "retirement fund" do I not want to increase the volatility. I plan to continue living for several more decades and thus need to use a fairly conservative risk profile. This lead me to the conclusion that I want to be in point C: split my total capital in such a way over the ETF basket and futures system that the resulting combination has the same volatility as the ETF basket, while at the same time profiting from the extra expected yield (point C is higher than point A). In my case this meant that about 25% ~ 30% of the total capital should be put towards the futures trading system, with the remainder in the ETF basket. This was quite a large increase for the futures trading account so I decided to execute this in a few steps and let the system gradually adapt to this new situation. The futures system started to open many more positions than it did until then, as the available capital had increased so substantially. Futures which previously were too large in size, or had too much volatility, suddenly came within reach. Early in the year I had about 9 positions at any one time, this has now increased to about 25~30 open positions. Interestingly I noticed that this extra diversification initially led to a lower volatility of the futures account. I modified some parameter settings to get it back to its desired level of volatility. These days my main activities in managing my investments have changed due to the change in approach. The main parameter which I now pay attention to is comparing the volatility of the combined account with that of the ETF basket, in order to remain more or less at point C of the graph and not gradually drift away from it. In fact, the futures system is lately moving up a bit too fast, resulting in higher volatility of the overall account. Regularly I take some of the profits from the futures account and move it into the ETF account (thus rebalancing between futures and ETFs). Another regular activity is to ensure that the ETF basket is to my liking and that the futures trading system behaves as it is supposed to do. After five years of working with this software are there hardly any cases where I need to either manually interfere, or make software changes. It only takes a couple of minutes per day to verify that everything is hunky-dory. I am very happy with the results. Comparing the end of 2020 with the account value of January 1st I gained a lot. In fact, this has become one of my best years ever.
Very interesting post, especially about the rebalancing. I do things a little differently as you might remember, I keep the same amount of notional capital at risk (adjusted for drawdowns), and periodically sweep any excess cash out into my long only portfolio. It amounts to the same thing, but the mechanics are slightly different; my futures system won't try and take on more risk if it has made a lot of profit, but I will end up with excess cash until the point when I sweep it out. One interesting thing about using futures as a diversification in this way is that you can actually set the vol target, whereas with the long only stuff you're limited to setting it within a fairly narrow range unless you want your expected return to take a hammering. In practice increasing the risk target on your trading account is equivalent to upping the risk allocation and keeping the risk target the same; it's just that with the former you don't have to put more cash into the account, wheras for the latter you do (well you don't have to, but I'd personally feel irresponsible not doing so). For me personally, I'd also be better off in theory putting more of my capital into futures trading. However I am a bit scared of the absolute daily £ swings that this would result in, which is why I originally cut my risk target from 50% to 25% annualised. This has given me a nice idea for a blog post about portfolio optimisting for a 3 asset toy model; bonds, equities and CTAs. Will hopefully get round to that next week... GAT
Interesting response, thank you. You are precisely pointing at the topic which is on my mind since some time. I have no theoretical background in finance and am confronted with these topics as I go along. I was indeed thinking about which method I should use to rebalance from futures into my long-only ETF basket. And I haven't made up my mind yet which method I would prefer best. These are the methods that I could think of: (*) Build empirical experience on when the futures account has increased too much in value. Then transfer some cash from the futures account to the ETF account. The account value goes down, and thus the available risk value goes down as the risk percentage is fixed (I'm using 25%). This will prevent positions from getting larger in size, thus preventing larger volatility. This is what I am doing (or try to do). (*) I could limit the maximum risk capital by putting a cap on the account value: if the account value exceeds a certain limit then the risk value will be limited at the risk percentage multiplied by the capped value. You described this a long time ago (either here or on your blog), but I wasn't sure whether you are still using this. Your reply seems to imply that you do. Also in this case would I need to build up empirical experience as to what the capped value would need to be. I would therefore have to go through the same learning curve as the previous bullet point. Which make both options seem rather similar to me. (*) I could modify the risk percentage: if the account value gets too large I could limit the available risk capital by reducing the risk percentage. I don't think this has any benefits over changing the account value, so I'm not pursuing this. As I said: I don't know whether one method is preferable over the others. Or whether there are even more methods, which are better than the ones I mention. Related to this, but slightly different, is a second consideration. Which is the question on how to use the available capital in the optimum way. The futures system and the ETF basket are in two separate accounts. I have thus capital tied up in the futures account. Not only because this capital defines how much risk capital I make available. But also to handle the margin requirements. I notice however that the margin requirements are substantially lower than the capital in the account. Wouldn't it be better to transfer more capital to the ETF account and run the futures system with a higher risk percentage? A higher risk percentage, coupled to a lower account value, would result in the same position sizes and the same absolute volatility (not same relative volatility of course). The capital thus freed up and transferred to the ETFs would create additional returns without increasing the volatility of the combined account. Ever since reading your book (Systematic Trading) and creating my own software am I running the system at 25%. It feels like a major decision if I were to increase this to, let's say, 40% or so.
In June 2020 I decided that my futures trading system, fully automated, would no longer be a small-sized stand-alone side project, but form an integral portion of my overall investments. This meant that I had to add more capital to it in order to give it the desired weight in the overall portfolio. I wrote about this in my post on January 1st, 2021. Now, one year after making this change is the time right to look at what this has brought me. IB presents me with a nice graph when I log on to my account. See it reproduced (I've edited account number and USD value) below. Until June 2020 was the performance not so impressive: it used to "bob along": no value lost, but also no value gained. Once the account value got increased, and the portfolio got expanded to more open positions, and thus more diversification, did it start to take off: Notice that IB gives SPX as reference. This is not really appropriate as this futures system covers more asset classes than just equity. However, as SPX is a well-known reference to many, does it make for easy comparison. Since adding capital to it in June 2020 has the system created profit. Since then have I withdrawn money multiple times, to rebalance with my other investments and keeping the size of this futures trading system at the desired weight. Earlier this year I did make a few changes though: (1) I increased the value at risk from 25% to 40% and at the same time reduced the account value. The value at risk expressed in USD remained unchanged. The capital which came available because of this was moved back to my ETF basket. (2) I put a limit to the value at risk. If the account value reaches a predefined limit, is the value at risk no longer increased. I can then transfer the excess money to my ETF investments. Although these are changes, the overall risk profile of this futures system is unchanged.
In November 2019 I started a new automated trading experiment. I call it the Sector Rotator. It reviews a list of US industry sector ETFs and invests in the top candidates. These are selected as follows: for each ETF it calculates a forecast, based on historical daily close prices. The forecast is a combination of trend following and momentum rules, each with multiple lookback periods. Then the ETFs are ranked based on their forecast: the one with the highest forecast is #1 and the lowest forecast is #16 (I use 16 sector ETFs). I invest money in the top 5. So if an ETF enters the top 5 (and its forecast is positive) I open a position with a value of 10 k USD. This position size remains unchanged as long as the ETF remains in the top 5. Once it drops out of the top 5 (or its forecast becomes negative) the position is closed. This is a long-only strategy, I don't open short positions for the worst performing sectors. This system has a daily review, done shortly after the market opens, and uses the daily close prices, up to and including that of the previous trading day. The intraday price at the time of review is not included in the analysis. I had let this strategy run since the start in November 2019 without giving it much attention or detailed review. I did notice that in general it was profitable, but I didn't know exactly by how much. This week I decided to collect all data and make a better analysis of the results. The two charts show the following: the left chart shows the invested capital (red) and the value of the ETF investments (blue). The horizontal axis represents time, starting in November 2019 and going up to today. The investment (red) is always close to 50 k USD (5 times 10 k USD). The value of the open positions (blue) fluctuates a bit, between 50 ~ 57 k USD. Initially, Nov ~ Dec 2019, I used a different position sizing strategy, which led to higher investments. But it also led to much more trading. I didn't like that and simplified it to what is described above. On 11~13 March 2020 all positions got closed and no new ones opened as the forecasts plummeted to negative territory, as a result of the Covid-19 pandemic spread. Positions started to get opened on April 9th. By April 16th I had again 5 open positions. The chart on the right shows the realized profit over time (the purple line). As profits are not reinvested (I only use a fixed capital of 50 k USD) this is cash coming out of the strategy. Closing all positions in March 2020 caused a loss, but that has since then been recovered. The total profit until now is almost 20 k USD. Is this good, or bad? To answer this I calculated a SPY buy&hold strategy as comparison: invest 50 k USD in SPY at the same start date, and see how the profit (i.e. value - 50 k investment) of this position develops. That is shown by the green line, which is the unrealized profit of this SPY buy&hold strategy. I am surprised by the result: I had not expected that Sector Rotator would perform equally well. I had expected that SPY would outperform Sector Rotator. The second interesting observation is that, although the end results are very similar, the way how we got there is very different: the purple and green lines follow a very different trajectory. This shows that there is a low correlation between SPY and Sector Rotator. Which means that Sector Rotator can be used as a diversifying instrument, reducing the overall volatility if I combine it with an investment in SPY (or something similar). Sector Rotator executed 304 trades since inception, which cost a total of 343 USD in commission (broker is IBKR). I received dividends from Sector Rotator: the cumulative net total is 638 USD (gross dividend minus US tax). Both values are small compared to the overall profitability, and the dividends compensate for the trading costs.