HedgeFundLIVE.com – On Tuesday, Mar. 15, the market fell on news that another explosion hit a plant in Japan. On the following day, there was another sell off, primarily triggered by comments from the EU Energy Commissioner regarding Japan’s radiation situation. Then yesterday, the market recovered with somewhat surprising strength. I decided to look back historically to confirm whether yesterday’s bounce was in fact “surprising” and unusual.
Note that I am using SPY data as opposed to my usual S&P E Mini data as I will be using daily percentage changes in this analysis and these figures better mimic those of the cash SPX (which is more closely followed than the E Minis when it comes to daily returns). Also, the SPY data that I am able to obtain goes back further in time than the E Mini data that I have access to- SPY dataset dates back to end of Jan. 1993 whereas the E Mini dataset dates back to Jan. 2000.
The SPYs closed down 1.15% on Mar. 15, down 1.85% on Mar. 16, and finally, up 1.32% yesterday on the 17th. So now I am looking for similar occurrences, namely, two large down days followed by a bounce in the market. The probability model that I am using takes inputs that will filter for trading days that closed down x% on day 1 and 2, then closed up y% on day 3. I find 1% to be a good “psychological level” so that is what I am using in this example. In other words, I am looking for days that closed down more than 1% on days 1 and 2, followed by a recovery of over 1% on the third day.
The above set up has occurred just 38 times since 1993. That is less than 1% of the time. So was yesterday’s move surprising? Definitely. To enable this analysis to be more tradeable, I incorporate stats for day 4 (which corresponds to today, Friday). The odds of the market closed up (close-close) today are 36.8%, which is actually very low. It is atypical for my probability models to spit out odds that do not hover around the 50/50 mark. The odds of the market showing positive returns over the next 10d period are 57.9%, which does not give me any conviction to the up or down side. Below are some additional stats that are outputs of this model:
1D 10D % Down on Day 1 & 2 -1.00% % Up on Day 3 1.00% Next Day Pct Chg Input 0.00% Number of Instances 38 Number that Meets Criteria 14 22 Number of Higher High Next Days 28 Number of Lower Low Next Days 10 Number of High Above R4 Next Days 16 Number of Low Below S4 Next Days 15 Probability of Up 0% Next Time Period 36.84% 57.89% Probability of Higher High Next Day 73.68% Probability of Lower Low Next Day 26.32% Probability of Ticking Above R4 Next Day 42.11% Probability of Ticking Below S4 Next Day 39.47% Click Here for full article.