[jz451]

Has anyone made an efficient frontier model for their portfolio showing portfolios along the curve with the minimum variance for a given expected return? I'm having a hard time making a variance/covariance matrix because the yearly data I'm using for each grade  from 2012-2015 produces values that are essentially zero with little to no variance/covariance. My matrix for grades A-G looks like this with the filters I use for each grade on Folio: Similar results occur with primary market filters also.

   A                            B                    C                    D                     E                     F                    G
A   0.00000905   0.00002128   0.00002712   0.00002579   0.00003233   0.00002002   0.00002472
B   0.00002128   0.00006770   0.00007329   0.00006886   0.00007666   0.00003445   0.00005669
C   0.00002712   0.00007329   0.00008796   0.00008341   0.00009832   0.00004778   0.00006351
D   0.00002579   0.00006886   0.00008341   0.00007915   0.00009362   0.00004534   0.00005921
E   0.00003233   0.00007666   0.00009832   0.00009362   0.00011631     0.00006735   0.00008149
F   0.00002002   0.00003445   0.00004778   0.00004534   0.00006735   0.00007228   0.00008984
G   0.00002472   0.00005669   0.00006351   0.00005921   0.00008149   0.00008984   0.00012917

This martrix results in only two grades being invested instead of a mix of all seven grades. For example:

                             A             B             C               D            E
weights      20.68%   0.00%   0.00%   79.32%   0.00%
return terms   1.51%   0.00%   0.00%   13.49%   0.00%
target   15.0%
st dev   0.47%


I haven't tried using monthly data yet but I'de expect a similar result. I also tested a portfolio of five stocks and the model worked so something is wrong with the data being used.

[Fred93]

http://blog.lendingrobot.com/research/on-the-mark-with-markowitz-diversification-part-1/
http://blog.lendingrobot.com/research/choosing-loans-with-monte-carlo-diversification-part-2/
http://blog.lendingrobot.com/research/demystifying-diversification/

[rawraw]

PeerCube calculates something like this for subscribers.

[TravelingPennies]

I've read these blogs before, and again I fail to see how they would help. My best guess is to gather the monthly returns and maybe that will yield a better result. from: Fred93 on October 30, 2016, 04:20:20 PM

[TravelingPennies]

from: jz451 on October 30, 2016, 07:05:12 PM

[TravelingPennies]

Yes, I'de expect the diagonals to be larger and it's odd in general that the outputs are so small. Below is the returns I gathered from NSR with my filters for each grade. Hopefully all I need more data, either quarterly or monthly. There is also a possibility that the small variance of returns by grade is resulting in this occuring, unlike a portfolio of stock, bonds, funds, ETFs, etc where returns fluctuate.

year               A                B                 C        D                 E       F               G
2012   7.40%   12.48%   14.88%   17.60%   19.60%   21.22%   22.95%
2013   7.71%   12.12%   15.28%   18.04%   20.71%   22.49%   23.53%
2014   7.29%   11.37%   13.85%   16.63%   19.01%   22.43%   24.24%
2015   6.87%   10.33%   12.87%   15.74%   17.72%   20.47%   21.17%

st dev   0.30%   0.82%   0.94%   0.89%   1.08%   0.85%   1.14%


from: Fred93 on October 30, 2016, 08:18:06 PM

[TravelingPennies]

from: jz451 on October 30, 2016, 09:03:45 PM

[TravelingPennies]

Which is what was what Lending Robot did. But just like if doing a matrix for other asset returns all you need is the returns for a given set of periods. I'm just using the data easily available since I can assume backtested returns are accurate. I, just like you, am sure you don't intend to take the time to go through the massive loan files and calculate for each loan for each grade.
from: Fred93 on October 30, 2016, 10:05:45 PM