Question about my rotation strategy
I have developed a rotation strategy that I'm quite happy with. This is not a day trading scheme, it is simply an evaluation tool that chooses one out of x funds based on fund-specific EMA and ROC parameters. Nothing new or unique, but it's mine and I'm pleased. I don't expect miracles but enjoy the idea of results comparable to buying and holding equities with much lower draw downs. It is relatively simple with only 5-20 trades per year based on 20 years of back testing.
I have been looking for a combination of ETFs and parameters to enter in my tool that produces a better short-term gain than cash or short bonds (SHY) but with very little drawdown. A cash alternative basically. Bond ETFs do not model well at all - short, long, intermediate, gov't, corporate, int'l - you name it. Bond ETFs simply do not perform well in a trend following rotation tool, or at least not in mine. With one crazy exception and this is where I'm baffled:
Take VWEHX - Vanguard's junk bond fund. Daily, compare it to its 15-day EMA. If the price of VWEHX divided by the 15-day EMA is greater than 0.02%, then buy/hold VWEHX. When VWEHX divided by 15-day VWEHX EMA is below 0.02%, then switch to cash. After a trade, hold for a minimum of 5 days (optional rule).
My back test shows this strategy consistently generates a CAGR on the order of 9% with a maximum drawdown of 3%. This does not pass the common sense test for me - something must be wrong. It doesn't matter what period or duration I run my back-test - 3 years, 10 years, 20 years - it's all similar.
What's my flaw? I do not own any commercial backtesting tool - if someone has one that could easily run this and post their results, I would be forever grateful.
Why does it not pass the common sense test? It looks to me like you're buying the fund if it is trending higher and switching to cash when trending lower. Why does 9% surprise you? Perhaps you could look at a years worth of trades by hand and examine what's going on?
What sectors are you rotating between?
Good questions and good recommendation - a check by hand couldn't hurt.
To address your questions... my strategy is primarily based on EMA crossover and simple Rate of Change (ROC) analysis, along with some safety filters applied. My "special sauce" is the fact that I tweak the parameters on a per-ETF basis. I find the optimized values (EMA crossover periods, ROC lookbacks, minimum filters) by running a series of several hundred backtests over matrices of values. The heat chart allows me to find effective ranges. Then I cross check those results across a few other randomly selected backtest windows - anywhere from 5 to 20 years. I like to think it's these tools that allow me to run these in a few minutes for any given ticker are what's unique.
In any case, if I run five basic semi-correlating tickers like: SHY (cash), SPY, VEA (int'l), VWO (emerging), and GLD - my backtest results are silly - like 25%+ and less than 15% MaxDD, and this is regardless over what time period I backtest - any 5 or 10-year period over the past 20 years or the entire 20 years. I'd like to believe this but I have this haunting fear that all I've done is created an incredibly optimized curve fitting exercise.