Wheezooo The backtest of each strategy produces a daily return calculation back to 2007 that can be downloaded from the online platform. Performing the Excel Correl function on the 60 day short put strategy returns and 555 day equals -0.86 [Cell M2]. The correlation for 60 day and 120 day returns is -0.91 [Cell T2]. Matt
Thanks. I have become so used to people using the term correlation when they mean regression, that I had to ask. I knew it wasn't a .9 regression, and was thinking to myself, there is no way this guy is telling me two puts in the same underlying are highly correlated, yet alone ignoring the 1,000 times more relevant regression. But I was wrong. Yes, returns on a long put and a short put in the same underlying will have an extremely high negative correlation, and the closer they are to one another, in time or strike, will increase that towards 1. Filtering your data better (removing unchanged or near unchanged days) would also push it towards 1. I am ashamed to have written that.
Good point nooby_mcnoob (and thanks for liking the post) The optimization process is the first step of many before settling on a trading strategy. Backtesting out of sample (OOS), and on different underlyings gives more confidence the strategy can be replicated. To test OOS you would set a date range to find a strategy, and test that strategy in a different date range, for example, testing from 2007-2017 with OOS of 2018-2019. You can also test OOS with a staggering function we call EntryDays where you specify how many days you wait until the next trade in addition to the automatic rolls the backtester does by default. It is preferable to do both of the above OOS testing but since I tested initially on the entire date range, I'll present the findings from the staggering. You can stagger any amount of days but I chose 30 days here. I have used staggering down to 1 day in other tests. The best backtest in the initial testing in terms of annual return at 5.19% and 0.84 Sharpe and in the staggered testing OOS had 4.36% annual return and 0.82 Sharpe. The Sharpe held up well but the annual return suffered OOS down 16%. The best backtest in terms of Sharpe at 0.95 with an annual return at 3.68%, in the staggered OOS came in at 0.91 Sharpe and 3.62% annual return. The Sharpe down 4.2% seems okay to me especially with the annual return holding up well. If anyone is interested in seeing how the strategy performed on other index ETFs I would prepare those.
Hi Baozi The FXI ETF tracks 50 large-cap Chinese equities that trade on the Hong Kong Stock Exchange. Here's the result for FXI for a long term long put 555 day 0.40 delta, and short term short put, this time with a slightly higher delta 0.20 vs 0.10, along with some other broad market ETFs and international ETFs. Here's some more detail on the FXI results: These returns are based on notional. Returns based on margin would be higher. Here's a link to our documentation on how we calculate returns. https://docs.orats.io/backtest-api-guide/backtester-methodology.html#calculating-daily-returns You asked in a previous post if we can test the results based on market environment indicators (i.e. IV rank, RSI, MACD) which we can. Let me know if you would like to see that.
Thanks for the testing Matt. So, would it be possible to do a test with conditions? Can you program them with custom formulas or do you choose from a preset?
You can customize the entry and exit dates by providing symbol, entry date, exit date in a CSV. https://blog.orats.com/backtester-basics-defining-specific-trades-on-specific-days We also have the ability to utilize external data sources.