Looking At Additional Predictors

A week or two ago I mentioned I was reading a book called "Big Data".  It gave me the incentive to try and pull in some additional data on MFI stocks to see if any were good predictors of future success.  Now to be clear, in no way is this meant to be an inclusive study, I only had data back to September of 2011.

I correlated that data with 16 of my monthly tracking portfolios, so about 800 "stock years".  Here are some of the predictors I tested and how they did:

Short Ratio

Decile	Percent Change	Stock Years
1	7.59%	80
2	27.11%	80
3	23.01%	80
4	15.77%	80
5	28.27%	80
6	5.78%	80
7	16.31%	80
8	18.96%	80
9	20.19%	80
10	36.86%	80

So interestingly, stocks that were the least shorted, did about the worst and stocks that were most shorted (about 25% and higher short ratio) did the best.  I'd feel better about the metric if the numbers in the middle showed more of a pattern

Days To Cover

Decile	Percent Change	Stock Years
1	15.03%	0
2	19.59%	0
3	23.06%	0
4	21.38%	0
5	18.34%	0
6	15.79%	0
7	15.93%	0
8	25.01%	0
9	20.21%	0
10	25.53%	0

Kind of the same pattern, but less severe.

Change From 50 Day Moving Average

Decile	Percent Change	Stock Years
1	22.72%	80
2	15.18%	80
3	25.50%	80
4	19.28%	80
5	17.53%	80
6	11.07%	80
7	23.43%	80
8	25.60%	80
9	19.24%	80
10	20.29%	80

Seems pretty random to me.

Change From 200 Day Moving Average

Decile	Percent Change	Stock Years
1	23.93%	80
2	18.93%	80
3	23.67%	80
4	23.40%	80
5	22.97%	80
6	22.33%	80
7	9.76%	80
8	19.19%	80
9	22.53%	80
10	13.21%	80

Again, no clear pattern

Change From 52 Week High

Decile	Percent Change	Stock Years
1	19.65%	80
2	19.97%	80
3	16.71%	80
4	20.96%	80
5	14.87%	80
6	23.77%	80
7	20.75%	80
8	26.89%	80
9	24.44%	80
10	19.63%	80

Slightly better 6 to 10, at 23% versus 18%.  10 means close to 52 week high and 6 is about 20% under 52 week high.

Change in Short Ratio Month to Month

Decile	Percent Change	Stock Years
1	16.25%	80
2	14.98%	80
3	14.79%	80
4	21.77%	80
5	29.24%	80
6	22.92%	80
7	29.25%	80
8	18.81%	80
9	13.52%	80
10	13.46%	80

This one was kind of interesting.  The middle group (4 to 7) did better than the extremes. They were up 26% versus 13% for the extremes.  The middle ones are where the short ratio is not increasing very quickly month over month, either up or down.  I do think this is worthy of further study.

It will take time, but I think I am going to start trying to build a more robust data set.  Let me know your thoughts!