Sure, finding patterns is easy; that's what humans do naturally. But it doesn't mean the pattern will actually be able to predict anything.
For example, in the data you posted, the worst month for the ISM seems to be March. The average change in the ISM from February to March each year is -0.5 and the ISM rose in March only 41% of the time.
Conversly, the best month was December. The average change in ISM from November to December was +0.4 and the ISM rose 53% of the time in December.
Now, who can tell me what this means? If you said "Well Cap, it means absolutely nothing!" then go to the head of the class. That's exactly right; what I just did is pure datamining at its worst.
In any set of data like this, there obviously would have to be a "best" month and a "worst" month. SOME month has to be the best right? But that doesn't mean that there's anything special about that month or that the same month will do as well in the future. One way to determine whether or not the pattern that you've found actually has any real meaning is to test it repeatedly in different samples and see if it holds up. There are other statistical tests that we can use as well.
So will the ISM be down next March? I have no idea at this point. But I WILL be working on some methods to look at fundamental data like this statistically. Now that I'm finally finished with that long project on the characteristics of trends, I can turn my attention to FA! Yay...
News 
Market Data 
Discover 
AI
