Hi HobbyTrading. If you don't mind, could you share how you deal with small capital? When I try to run a backtest, futures that are small enough to trade now may not have long enough history (e.g. Korean bonds). Using a bigger universe often lead to many positions with less than one contract, while having one futures per sector seem insufficiently diversified. I am curious whether you use some heuristics to get a dynamical universe that changes slowly compared to position turnover. Another question is, what are your typical and maximum leverages for your futures portfolio?
Thank you for your interest. I used this blogpost as starting point: https://qoppac.blogspot.my/2016/03/diversification-and-small-account-size.html This is Robert Carver's blog. The same person who wrote the book which formed the basis of my trading system. Now that I have my system running am I not doing backtesting any longer. And I don't worry about instruments which have only a few years of price history available. As that blogpost also concludes, is there an optimum between diversification and accuracy. If you can handle only one contract in an instrument your system cannot respond accurately enough to price changes. If you reduce the number of instruments and allow larger position sizes in each instrument you become less diversified. The basis of my approach is that I have defined certain asset classes (which I call clusters). Within each asset class do I allow a position in the instrument with the lowest value volatility. As example: cluster Grains contains Corn, Wheat and Soybean. Corn has the lowest value volatility so I do not allow a position in Wheat and Soybean. However, if my account has grown to a size where the position in Corn can reach four contracts I start to allow a second position in Grains (either Wheat or Soybean, depending on the value volatility). I use the same approach for asset classes such as Bonds, Equity, Forex, Meats, Metals. I review once per week how many positions per cluster I allow and what percentage of the total risk capital may be used by each of these. As a guideline do I use an (1/n) approach: if n is the number of clusters in which I allow a position, then each cluster gets (1/n)*100% of the available risk capital assigned. If a cluster is allowed to have two positions I assign half of that cluster percentage to each position.
Thank you for your detailed reply. I have been following your thread since early this year, and GAT's thread and his blog from a few months ago. I have been experimenting with optimization with constraints and turnover penalty (because bootstrapping leads to numerous zero positions after rounding to integers), but they generally lead to high turnover. I'll just adopt your approach which is more measured and will allow me to sleep better when I eventual start to trade.
I found the bootstrapping method too complicated and too labour-intensive. Comparing that to the handicraft method which he describes in his book I didn't see the need for more accurate simulations. My attitude is that I rather be approximately right than precisely wrong. So I have been using handicraft weights since the start in 2016.
OK, time has come to face the fact: it is highly unlikely that I will be able to achieve the next milestone. For the account to grow back to the previously achieved high water mark by August would a run be necessary similar to that of the period Oct. '17 ~ Jan. '18 (see graph). I don't see any sign of that happening right now. I don't see multiple markets all showing a clear trend, and having the low volatility as was present in the markets last year. In order for this futures trading system to be sustainable would it need an account size of approximately five times the annual amount I want to withdraw from it. Why a factor five? Backtests by others on a similar system (e.g. @globalarbtrader , and others reporting in his journal) have shown an average annual return of around 20%. My mistake was, when I set this year's targets, that I took the return of the last 12 months (104% *) and assumed that I would be able to enjoy a return in the same ballpark also this year. However, 2018 until now doesn't even come close. So, instead of making further withdrawals am I changing the approach for the remainder of this year and invest more money to the account, to achieve the desired factor five. I don't have enough money available to achieve this on short notice. I can put in what I have and the rest has to come from organic growth of the account. So it might take some time before I get there. This is the current graph of the account (net liquidation value and high water mark, both on a normalised scale): The futures system prodded along during April, nothing of relevance to mention. Until late in the month, when I injected all money which I have available for it. This resulted in the step function near the end of the month. The ETF system is another story: of the 40+ ETFs which I track were during most of this month only two identified which gave a signal strong enough to invest in them. I had not expected that I would end up in such a situation and had developed the software from a standpoint that there would always be enough to choose from (at least four candidates to choose from). I modified the software this month to not invest all money in only two ETFs, but keep a portion in cash. Only when there are three or more positions is the entire amount of available money used for ETF positions. I ended the month with only one ETF position, plus a cash position. * See here a simple overview of the growth of my trading account over the last twelve months: (all values are normalised to the starting value, which is set at 100%) Start value (6 April 2017): 100% Deposit 1 (25 May): +18% Deposit 2 (10 Nov.): +16% Withdrawal (2 Feb. '18): -71% Profit: +104% End value (5 April 2018): 168% Positions by the end of April (IB symbol codes): futures, long: CL, GBS, NIFTY futures, short: AUD, CHF, FLKTB, GE, PL, V2TX, XINA50, ZB, ZC, ZN, ZT ETFs, long: USO
After increasing the account value one month ago has the situation improved. More instruments are within reach and more diversification is possible. More positions were entered early in the month. Their forecast signals got stronger during the month, resulting in expanding position sizes. Until I reached a point where IB blocked further order lines as the account was not able to fulfil the margin requirements. This made me (a) throttle down the system a bit and (b) reduce the percentage of the account value to be used for the ETF system, resulting in a larger cash position. Nevertheless, the account reached new high water marks several times during these weeks. And then everything changed and nosedived, starting on May 19th. The account fell from the latest high water mark to a drawdown of -28%. This happened pretty fast, within one week. All position sizes got reduced and some positions got closed altogether. Overall the account value decreased by 14% this month. See the normalised graph of the net liquidation value and high water mark in the upper chart and drawdown in the lower chart: The drawdown chart shows a lower minimum (-28%) than I experienced in 2017. In 2017 I experienced drawdowns up to -20%, as can be seen from the graph. (Ignore the period Feb. ~ Apr. 2018, where I first withdrew money from and later added money to the account). Recently I increased the volatility setting of the account, allowing for larger upswings, and obviously, downswings. In hindsight I could have throttled down the system a bit more when IB started blocking order lines due to margin requirement. That might have lessened the drawdown which I experience now, at the end of the month. I was a bit too greedy, perhaps? The ETF system returned to a more normal situation, compared to my remarks last month. Enough ETFs got a positive forecast signal and there was enough to choose from. The full amount of money which is made available to this system was indeed used to buy positions in four ETFs. Open positions by the end of the month (IB symbol codes): futures, long: DJ600, GBM, GBS, NIFTY futures, short: CHF, GE, M6E, MXP, PL, V2TX, XINA50, ZT ETFs, long: IWM, QQQ, XLK, XLY
To what factor(s) do you attribute the drawdown and do you have any desire to avoid it in the future?
I believe that a new high water mark is always followed by a drawdown. It can't be an endless party. However, I use diversification of instruments in order to avoid very strong changes in drawdown (i.e. rate of change per day). Even though I select instruments which are not too strongly correlated with each other over the longer term, they are sometimes strongly correlated on a short term. In such a case is the daily rate of change larger than intended. On May 17th suddenly the US treasury bond (short term 2Y, as well as long term 10Y) changed its trend and went up in large steps. Unsurprisingly did EuroDollar (GE) do the same. What emphasised the situation was that a few days later suddenly crude oil changed trend and started to fall. And at the same time German bonds shot up (e.g. BOBL, BUND). You see: several items which are normally not correlated but this time all decided to move in the same (short) time frame. And the positions which I had built up started to move against me. Do I want to do something against it? I would if I know a failsafe way of doing this. One thing I did consider but did not implement: if the account value shows a rising trend allow for larger position sizes and thus more risk. However, when the account value then changes trend and starts to drop, reduce the risk settings for all instruments and thus reduce open positions sooner. I did not implement this as others, who have backtested this, stated that it did not help them. Another possibility could be to search for more instruments which are not linked to the US dollar. I have found a few, such as the German government bonds and the NIKKEI equity index, but the list of instruments is still very much USD-based. A limiting factor here is that market data subscriptions in other markets are rather expensive when compared to my account size.