Currently, the S&P 500 (SPX) weekly slow stochastic %K is reading 98.11–an extremely overbought position in any market. So, I wanted to take a look at historical precedents to establish an expectation for SPX’s average return following a weekly %K >98. I also ran the data for oversold stochastics with %K <10. My results are somewhat counterintuitive, perhaps a hattip to momentum, suggesting that I buy in the wake of extremely overbought and sell after oversold readings–pretty much regardless of my holding period.
My study used a data set ranging from 1982-present (inclusive). My “control” group was a simple rolling average of SPX weekly returns (sans dividends) across the specified holding period. My conditional (“if”) groups included all occurrences in which SPX’s weekly slow stochastic %K closed either over 98 or under 10–overbought or oversold, respectively. I ran data for 1 week, 1 month, 3 month, and 1 year holding periods. For both groups, I calculated average (mean), median, and the standard deviation of returns. The results:
In 26 occurrences since 1982, SPX has outperformed its control group (a rolling average) for 1 month, 3 months, and 1 year periods following an extremely overbought stochastic reading (%K >98) on its weekly chart. 1 week is the only holding period over which I’d question the significance of the conditional results, for which the average (mean) actually underperforms the control group by a slight margin (+0.17% vs 0.19), although the median outperforms (+0.36% vs 0.30). These 1 week results lack significance, in my opinion, due to the distributions’ large standard deviations, as expected of shorter time periods.
For the other, longer holding periods presented, the conditional sample not only outperformed the control group by a significant gain/loss margin, but it also accomplished this with far greater reliability, as per standard deviations that are a fraction of the controls’.
Back in May 2012, I ran an analogue study regarding extremely oversold stochastics. Specifically, I looked at rare occurrences in which %K read <5 on daily fractal. The guide proved accurate for the near term–as SPX staged a rally in accordance with the forecast–however it slipped to lower lows 1½ weeks later. The data I've crunched here today suggest that such an ephemeral bounce is actually the norm.
Herein, I used data over the same 1982-present period, but my conditional group required an oversold stochastic %K <10, which provided more datapoints (17) than a more extreme reading would have. As displayed in the table above, the results are far less significant than the overbought set's. Nevertheless, the raw takeaway recommends buying the oversold condition for a short, 1 week holding period. Any longer holding period either underperforms or the outperformance pales in comparison to the results' standard deviation. The 1 year holding period is the best example, yielding poor relative return (mean +3.36% vs 9.08; median -0.05% vs +10.65) with shocking inconsistency (sigma 27.31% vs 16.56). Upon closer inspection, even that 1 week segment shows such a volatile distribution as to discourage investment upon its premise.
These results objectively and quantitatively redouble my confidence in the SPX, fortifying my other analyses–such as the bullish 2006 monthly analogue. Buy SPY.