Millard,
The scambled data in the middle of my original post was a table at one time. I don't know how it got scrambled. I have retyped most of the information contained by the original table in a new table in post #4, which was updated in post #5.
I'm not familiar with Clifford Sherry's book on "Mathematics of Technical Analysis" but experience from my own analyses is consistent with your's, that changes in price are not predicted well from previous price and volume action. My models describe previous price changes pretty well but don't predict future ones well enough to base investment decisions on. Basing predictions on extrapolated trend lines is dangerous because trends can, and do, change randomly without warning.
What I'm hoping my SIBS and SISS systems can do is identify situations when trends are unlikely to change. It involves the detection of changes in trend, statistical confirmation of the new trend, and observation of a signal that the new trend is unlikely to change again before a rise of one point or more in price. As you can see in post #5, it has been successful 8 out of 11 times since beginning this public test. Unfortunately, the 3 failures are relative big losers so far. Some of the failures may be avoidable by also considering whether the price is at, or near, a resistance line or by considering fundamentals. In this test, I'm ignoring such external information and using only my own criteria for SIBS or SISS, and 8 of 11 is encouraging, IMHO.
Another issue is the infrequent occurrence of signals. This problem, however, should be solvable by increasing the number of stocks in my data base. I am currently following only 35.
Later,
Randy |
