Monday, March 14, 2011

MFI Streaks

One theme I have discussed previously is that MFI is extremely streaky. Greenblatt discussed in his book that a key reason that MFI would continue to work is that it could have stretches of not working and would require patience. He said that in his studies, only 7 of 12 months had MFI out-performing indices.

What I did not appreciate was that I thought the months would be random, but so far they are not. We have many consecutive months that it under-performs. We are in such a streak right now. MagicDiligence is seeing this, I am seeing this and my monthly tracking portfolios are seeing this.

The table below shows every single monthly tracking portfolio since I started back in January 2006. I place a "1" if the portfolio beat the benchmark and a "0" if it lost.

Date Total IWV Flag
1/6/06 16.0% 10.9% 1
2/17/06 21.2% 14.6% 1
3/29/06 13.0% 9.6% 1
4/7/06 10.3% 12.1% 0
5/12/06 20.4% 18.6% 1
5/31/06 29.2% 23.3% 1
6/30/06 22.4% 20.0% 1
7/31/06 19.7% 17.3% 1
8/31/06 13.0% 13.3% 0
9/28/06 12.7% 14.6% 0
10/27/06 10.3% 12.0% 0
11/29/06 -0.3% 4.8% 0
12/28/06 -6.9% 3.4% 0
1/26/07 -10.2% -6.6% 0
2/27/07 -3.7% -1.0% 0
3/26/07 -9.8% -5.5% 0
4/27/07 -10.9% -5.0% 0
5/29/07 -11.5% -6.3% 0
7/3/07 -30.0% -15.6% 0
7/30/07 -19.9% -11.5% 0
8/30/07 -12.5% -8.7% 0
9/27/07 -19.0% -18.2% 0
11/2/07 -40.4% -34.3% 0
11/28/07 -40.1% -38.3% 0
12/28/07 -36.3% -40.0% 1
1/25/08 -36.4% -35.9% 0
2/26/08 -51.7% -41.5% 0
3/24/08 -40.9% -36.8% 0
4/25/08 -25.6% -31.0% 1
5/28/08 -22.2% -33.6% 1
7/2/08 -11.7% -25.3% 1
7/29/08 -10.5% -20.9% 1
8/29/08 -13.8% -17.9% 1
9/26/08 -4.3% -10.0% 1
10/31/08 18.7% 13.9% 1
11/26/08 50.9% 27.7% 1
12/26/08 48.9% 32.3% 1
1/23/09 59.3% 36.4% 1
2/27/09 92.8% 55.6% 1
3/27/09 85.8% 48.1% 1
4/24/09 69.7% 45.8% 1
5/29/09 31.8% 22.8% 1
6/29/09 21.3% 24.0% 0
7/29/09 19.5% 15.9% 1
8/28/09 7.4% 8.8% 0
9/25/09 12.6% 12.4% 1
10/30/09 22.7% 18.3% 1
11/27/09 24.3% 13.6% 1
12/31/09 23.7% 18.1% 1
1/22/10 19.0% 20.6% 0
2/26/10 18.6% 23.6% 0
3/25/10 7.3% 15.1% 0
4/23/10 2.5% 9.5% 0
5/28/10 14.6% 21.8% 0
6/29/10 22.0% 28.0% 0
7/29/10 13.6% 21.1% 0
9/2/10 23.8% 22.0% 1
9/24/10 15.7% 15.7% 0
10/29/10 9.7% 11.6% 0
11/26/10 9.1% 10.4% 0
1/3/11 2.1% 2.9% 0
1/28/11 1.2% 2.6% 0
2/25/11 -2.6% -1.2% 0

Note that in the past 14 portfolios, only one is beating the benchmark. Also, of the 63 portfolios, only 28 have "won". Not quite the 7 in 12 from the back test yet.

Overall, with $100,000 spread over the first 12 months, you would be at a tepid $121,000, versus $114,000 if in the Russell 3000. That differential is pretty consistent with my personal results.

5 comments:

Homer315 said...

I vote you use the current picture for all your blogs posts, actually...

Homer315 said...

Can't remember, and am too lazy to look to see, whether you have much, if any real cash invested in CCME.

I hope not...

Marsh_Gerda said...

I have more than I would like to have. But it will be less soon.
:(

drew said...

I've always felt your streakiness analysis is flawed and here's why:

If a handful of stocks from the January list have a very strong June, it positively affects at least 6 portfolios through the June portfolio, assuming it drops off the July portfolio because of its increase in price. If this run up occurred in month 12, it could affect the prior 11 portfolios.

In conclusion, I wouldn't worry too much about the streakiness.

drew said...

If you want to try to make use of the streakiness analysis, I just took your table and compared the portfolio end result with the following year return and found that in the 39 portfolios from 1/2007 through 2/2010, 8 times you beat the index after beating it the year prior, 8 times you lose after losing to it the year prior, and 23 times you do the opposite (about an even number in each direction). So the takeaway could be that you wait for a losing 12 months to jump in, but I'm inclined to think we are just dealing with too small a sample size to be useful.