Thee above image is the median sharpe ratio (pre and post cost and the ratio between post-cost sharpe and pre-cost sharpe) for strat 8 for all instruments. I've backtested selected trading rule variations dynamically — meaning, I choose (choose According to the speed limit rule for trading rule variations) which look-back windows to trade for each symbol during the backtest using only backward-looking statistics like turnover, as suggested in footnote 135. This decision is made monthly or quarterly within the backtest, based on a rolling 6-month annualised turnover. To avoid pre-selecting look-back windows based on overall backtest results for every instrument , I tried to backtest by dynamically selecting look-back windows based on speed-limit rule. However, when I apply this to many symbols, I often find that no look-back window is selected for several periods. This happens frequently enough to be a concern and seems to result from strict adherence to the speed-limit rule. My goal here is to avoid in-sample overfitting by not using the full backtest history for window selection. Please Let me know if it is still not clear what I am saying. Best Regards, cryptocaptainx3.
I'm very skeptical about those SR varying by that much as you slow down, and I'd be surprised if you had enough data to support those figures. I'd be inclined to set the pre-cost SR as the average across those speeds and use that to apply the 1/3 rule. Anyway, if I don't select any rule I have a fallback where I select just the slowest even if strictly it exceeds the 1/3 limit. Since I'm running dynamic optimisation there is a good chance I wouldn't trade those tickers in reality. Eithier they would be marked zero position on the basis of being a 'bad instrument' with high costs, or the cost penalty in DO would so high they would never trade. But if I wasn't running DO then I wouldn't trade those instruments at all. What asset class is this? So expensive. Rob
I have 5 equity index futures, 7 agri futures, 4 base metals,2 precious metals and 2 energies.(20 in total). I can't trade FX Futures(Not Allowed), I can't trade bond futures (no liquidity). For Vol, I will have to trade options. All from Indian Exchanges. For all base metals, precious metals, energies and 2 equity index futures I have 20+ years of data, 4 agri futures with 18-20 years of data and 3 agri futures with 7-13 years of data. Also, if I ignore the speed limit rule and compare just the pre and post cost sharpe ratio, and filter on the basis of post-cost sharpe atleast 40% of pre-cost sharpe, then I am left with just 16 instruments, where 10 instruments I can trade 5-6 lookback windows. Agri futures slower than EWMAC8. Also one more thing that even though I am using completely different data, my results are have extreme similarity with the results in the book.