Year end, almost 2 months running the system. Christmas had a bit of drawdown with Cocoa and some trends pulling back, then a somewhat happy new year recovery back to positive territory. As Cocoa turned down, I noted the strong backwardation in Cocoa and how much cheaper May was. I took the opportunity to credit $thousdands in roll yield by rolling to May early. $867 per contract price difference, which ends up being quite a lot as it's 10 tonnes per contract. The backtest only includes the fact that there is roll yield (or cost), but on a pre-programmed contract roll design. This was a little discretionary override that happened to work out, maybe due to my previous experience spread trading. I saw the chance of getting stopped out, and wanted to cash in the roll yield before it happened. Some luck that it happened to be a wide part of the spread too. Fortunately the new contract had slightly less volatility and my short term system just survived without getting stopped. This screenshot isn't the exact stop from the day, it does show where the stop is moved to for next market open but it was real close. Also found a bug in the playback of the live trades CSV (how I generate my reports) - it wasn't adjusting the prices that fell on prior contracts. Therefore it wasn't reporting roll costs or roll yield properly. The result is that some of the P&L on positions will now be different to previous screenshots since this is fixed. Also the roll yield > roll costs so far, trend following tends to be good at sniffing out cases of backwardation when long. There's a paper that explains this, page 7 and 8 on this one: https://quantica-capital.com/en/publication/rolling-down-the-curve Final market stats for Dec and 2014: Interesting that it started out very long side biased, with shorts losing a bit. Now shorts have helped handle the drawdown and are catching up. Also decided to breakdown timeframe. 2 months is way too short to know anything yet, but it's still worth tracking and keeping an eye on things - there's been plenty of bugs I've detected and squashed along the way so far. Dealing with holiday mismatches between countries is the latest thing, some missing bars for some symbols but not others. RealTest is adding a feature to handle this that I've been testing historically, and now live testing as well for new years.
Based on the way the sizing works, before running out of margin the system would hit a situation where the backtest is showing over 60% drawdown and huge daily volatility ie I'd have to be risking too much. I would never go forward to trade something like that. You just reduce risk% per trade until volatility and drawdown gets to where you need, keeping in mind that the largest drawdown is in the future not in the backtest and could be double that of the backtest drawdown. Currently the number of positions is pretty high vs the backtest, and there's plenty of maint and initial left even for new positions. It helps that unrealised gains are not included for sizing purposes. If doing it with a smaller account, I'd do just one long and one short system, but multiple entry signals like Tom Basso talks about. I believe Tom checks Donchian, Bollinger, and Keltner then just enters whichever happens first. With micros and carefully selected markets that could be run with 150-200k account. Less systems/markets but higher risk per trade.
Your profits are unrealized whereas losses are realized? If it were me I would rather it be the other way around?
That means you're suited to a much shorter term system, or a negative skew system that makes profits every day but makes the losses in big lumps. This is the opposite, it cuts losses quickly and regularly (hence the realized), and lets profits run (hence profits are unrealised, for now). Check back to page1, at WinLen, the average WinLen is 66 to 125 market days. That's 13 weeks to 6 months long. Currently I'm at week 8, so I'm not even at the average length of a win for the short term system yet. Until I get there, the profits will keep running up and stay open. Psychologically it can be hard to run for some, as it's tempting to just close positions and take profits. From testing we know that's the wrong move, on average, over time.
Daily closing price. It might be possible to calculate the worst case scenario by looking at all positions going max against me at once, but there are other exit rules besides stops so it's not very useful to know. To prevent a rare case like that causing catastrophe, the system does subtract all open P&L from total equity before sizing new positions. Also note while it's called open P&L, technically some of it is realised in broker terms because most of these having rolled contracts already, e.g. Cocoa March contract was sold for approx $75k profit as May was opened. But for the purpose of trade tracking I still consider it open as one long trade, matches the backtest better and easier to track.
We are so back? After getting screwed by Santa, January started off slow but stuff happened quickly over following week. Nobody expected rates to move like this, and to be frank I have no idea either when it comes to bonds. But trend following was already short a whole bunch of currencies (benefiting from higher US rates than expected), as well as short Long Gilt, the US 5 year and 2 year. Soon after last post it got short the Euro Bund and Japanese bonds. Then I knew things were really serious when it went short the US Tbond, the 10 year, and the Euro OAT for 9 Fixed interest markets total or 13 strategy positions. Stocks got hammered a bit, as did other longs, but this is why we have shorts. Even though the shorts contribute a very tiny amount in the backtest, they really carry the portfolio forward through tough weeks. Even if shorts would lose money overall, I'd still keep them in as they keep me alive. Very interesting to start trading this at a time when shorts make more than longs, this should be rare. Some signs of life are there for longs though - Live Cattle moved quick, Corn has kicked off, and Gold looks like it's coming back to life. I'm still a bit under the high watermark from mid-Dec that Cocoa and Coffee pumped me too, but on the way there. One Cocoa actually exited, on a Bollinger band rule (short term strategy) which was good to have locked in as I'm aware I never caught the Cocoa trend properly - I just got to ride the tip of it where it's super bumpy. I changed my CSV reporting a bit to now track the portfolio the symbol comes from, it was so damn confusing with them just in a long list like that I couldn't keep track of what I was in or not in. Pretty colours etc, easier to read IMO. Pleased to see I'm starting to see some exits that are actually in profit and one category went from red to black (softs). Realised loss is slowly starting to decline, I'm hopeful I can be in gain territory by month 6 but really I have no idea. As long as it gets there eventually. Since I have a weekly context requirement for entering positions (weekly bars trend check), I tend to get more orders on the weekend for the Monday. Monday coming is even MORE fixed income shorts, definitely got me nervous about bottom ticking that. I have sizing much smaller for large universes like fixed income, as well as some position limits if things get too hectic but I have no issue holding a broad coverage of many fixed income markets as long as volatility sizing is not too large. Monday new long trades coming: -Crude Oil Monday new short trades coming: -Another Korean Wong (3rd of 3) -Another US 5 year (2nd of 3) -Canadian 10 year bond -Euro Bobl 5 year bond
1/ What was the starting AUM? 2/ What is the maximum Margin:Equity ratio at any time? 3/ Do you adhere to an upper limit of Margin:Equity?