Saturday afternoon found me at the neighborhood chili cook-off, which I won one year and placed runner-up a second year. Being lazy, I did not enter this year. Anyway, I was talking with a good friend who is in the biz, and he was concerned about the current state of the markets. I was sharing that I was relatively unconcerned, as I believe we are not witnessing more than a run-of-the-mill correction. What seemed to concern him the most was that the S&P 500 had fallen more than 1.5% beneath its 200 day moving average. I did not find this to be particularly disturbing, and offered a guess. My guess was that after falling more than 1.5% beneath its 200 day moving average, the S&P 500 (using $SPY as a proxy) would rise on average between 4 to 5% over the next 50 days. He asked me to test it, and so I have. Here are the results.
The Rules:
- Buy $SPY at the close when it falls more than 1.49% beneath its 200 day moving average
- Sell $SPY at the close X days later
- No commissions or slippage was accounted for
- All $SPY history was used
The Results:
So Bryan was right, and my prediction was way, way off. $SPY closing more than 1.49% beneath its 200 day moving average has been a fairly neutral to bearish setup, over the past 18 years.
- There were 637 instances of the setup.
- There were 37 trades held the full 50 days.
- The median of the 37 trades was 1.21%.
- The biggest winner was closed in June, 2003 at 13.05%
- The biggest loser was closed in November, 2008 at -32.89%
I am still surprised at how neutral to bearish this setup has been. Of course the -32.89% loser skews the average, but when the first close beneath the 200 day moving average yields a strong average trade of better than 4%, it is interesting that changing the variable to 1.49% beneath the moving average affects future results so dramatically.
For you folks that can’t get enough of the numbers, below is a graph of all the trades.
Have a great holiday trading week!
I’m surprised by the results too… Interesting every time the event occurred in November resulted in a negative return.
This seems easy. The years of negative returns were in bear markets.
So, if you think we are entering bear market, sell soon. Otherwise, buy moar.
So what do you think? Entering bear?
You should be using a z-score to help determine if your result set is large enough to determine any statistical significance. Anything that falls outside a second standard deviation should likely not be included in the final results.
I agree about noting individual outlier trades which may skew the mean. I’m not big on Z-score, although I’ve used it before, as it assumes a standard normal distribution. Stock prices / returns imo cannot be assumed to have a normal distribution.
Q: “S&P 500 Closes More Than 1.5% Beneath its 200 Day Moving Average. Bullish or Bearish?”
A: Meaningless
Taco’s right though, test for significance.
Both p-values and Z-scores are associated with standard normal distributions, which are not typically found in stock price series or system returns.