This is a continuation of my previous studies of possible tells for forecasting the Open based on the trend of the previous day's closing bars . . . either the 10, 30 or 65 minute.
In this case I'm looking at the 30 minute bars for the Qs.
On the chart itself the 7 and 14 SMAs are shown, the optimized values for the Q that I've discussed previously.
In the lower technical panel the 3,7 and 12 SMAs are displayed along with the Schwab signals lines of 5,16,4 and 3,14,3, again values I've discussed before in the context of my ongoing refinement of the Qs Dirty Dozen systems.
The classic values for the MACD histogram 12,26,9 are also displayed.
For this post I'm most interested in the slope of the 7,14 on the chart and the 3,7,12 on the lower technical panel in conjunction with the signal line and the MACD.
Sometimes the simplest approach is the most robust and what I've noticed from looking at this little snapshot is that when the 7 and 14 MAs are both upslope into the close, the following opening hour is positive. If the 7 and 14 MAs are both downslope into the close, the following opening hour is negative.
The same results are achieved using the lower technical panel of the 3,7,12 MAs. The effect of the MACD and signal lines may add confirmation to the closing signals, but require additional testing before deploying.
On a side note, I've been working on a pairs trade idea using the overnight % change in the Qs and the SPX or DIA (higher to lower beta) to gauge likley price behavior for the day. My intuition told me that I should expect a mean reversion. . .that is, if the overnight change in the Qs was + 2% and the overnight change in the SPX was 1%, then I should Sell the Qs and Buy the SPX. Surprise! It doesn't work that way. . at least going back the 300 days of data that I looked at. The best results are achieved by Buying the larger % gainer and Selling the lesser gainer if the differential overnight returns are greater than .9% and positions are held to the end of the day. Obviously these results can be improved considerably by adding some basic stop losses and/or trailing stops. This study is in very rudimentary stages and requires considerable more work, but the initial data runs are encouraging, to say the least.