Up to you, and I don't trade Corn like this, but personally I'd do "Use the value of the last correct carry forecast (on 4th June) from that day on until I roll to the Nov25 contract and have a nearer contract available again" Rob
Thanks for your answer, Rob. You mean that you don't trade Corn at the MATIF at all, right? I understand that you only trade the big contract (ZC, CORN in pysystemtrade) and the mini contract (YC, CORN_mini), both at the CBOT. For the big contract, you only hold the December contract, but for the mini you hold all of them (Mar, May, July, Sep, Dec). If you traded Corn at the MATIF (EMA, CORN-EURO), you would also hold all of them (Mar, Jun, Aug, Nov) At least that's my understanding after reading the rollconfig.csv file in pysystemtrade. Am I correct?
Yes my preference would be to to hold only December, but last time I checked mini CORN Dec wasn't liquid enough far enough forward to make that possible. Looking at the MATIF, same situation. So yes I'd hold all of them. Rob
@globalarbtrader Hi Rob, In AFTS Strat 9, How would you calculate IDM since for every instrument we have multiple spans' substrategy returns? Do you take a weighted average of substrategy returns per instrument (using forecast weights)? Or any other method? Thanks and Regards, cryptocaptainx3.
IDM is based on the correlation matrix of substrategy returns and instrument weights so not sure why you are talking about forecast weights. When you say 'multiple spans' are you referring to the fact that different instruments trade for different times? If so something like this: restimate IDM every year, at the start of each year if any new instruments are coming in I assume their correlation with others is 0.99 so the IDM will be conservative, then once there is enough data I properly measure the correlation. Rob
No no, what I mean is that every symbol is trading multiple EWMAC lookback windows, so there is no single substrategy returns history for each symbol. So to arrive at a single substrategy return for each symbol, we have to take an average using forecast weights?
Not to be that guy selling books but this sort of detail is covered more in Systematic Trading or you can wade through the blog. Rob
@globalarbtrader Hi Rob, I hope you're doing well. I had a question regarding Strategy 9 from AFTS and would appreciate your guidance. I've been implementing backtests using dynamically selected look-back windows (According to Footnote 135), where the selection is constrained by the trading speed-limit rule. To estimate turnover, I'm using a rolling 6-month window (annualised). However, for many instruments, none of the candidate look-back windows pass the speed-limit test under this methodology. I have just 20 instruments in my universe. Interestingly, when I ignore the speed-limit constraint and simply use an equal-weighted average of all look-back windows, I end up with a Sharpe ratio(net of all costs) close to 0.45 (0.15 is 1/3 in costs) — similar to the assumed average costs Sharpe ratio mentioned in the strategy. Given these observations, I’m unsure how best to proceed. Would you recommend relaxing the speed-limit rule in some cases, or perhaps adjusting assumptions such as expected Sharpe? Thank you in advance for your time and insights.