Sector Model
|
XLU
|
0.00%
|
|
Full Model
|
Date
|
Return
|
Days
|
ESI
|
8/4/2014
|
-48.57%
|
187
|
EDU
|
10/27/2014
|
-16.60%
|
103
|
UVV
|
12/2/2014
|
19.84%
|
67
|
JOY
|
12/8/2014
|
-12.90%
|
61
|
RS
|
12/11/2014
|
-12.15%
|
58
|
JNPR
|
1/5/2015
|
4.37%
|
33
|
COG
|
1/12/2015
|
-10.66%
|
26
|
GNW
|
1/20/2015
|
6.30%
|
18
|
AGCO
|
1/23/2015
|
10.16%
|
15
|
DNR
|
2/4/2015
|
3.13%
|
3
|
(Since 5/31/2011)
|
|||
S&P
|
Annualized
|
12.17%
|
|
Sector Model
|
Annualized
|
24.99%
|
|
Full Model
|
Annualized
|
20.35%
|
A little better tape this week:
The Sector Model flipped over to XLU after that sector
suffered an alarming collapse on Friday.
The Full Model is still down for the year, but has recovered
most of its lost ground. ESI continues
to be a plague, with attempted rallies that quickly fizzle. Had I bought it one day later I’d be flat,
but that first day was a disaster.
So where does that leave me?
Quite well on the revised fundamentals, actually:
Greenblatt
|
-14.44%
|
Graham
|
8.60%
|
Adaptive
|
-12.09%
|
Matrix
|
-18.70%
|
5 Year Sales Growth
|
4.04%
|
Adaptive-2
|
745.30%
|
This is a record of stocks based on fundamental parameters
across all time frames. I keep track of
stock performance even after I sell the stock, and record the average return
rate for those stocks within the fundamental filter that was applied to them in
the selection process.
The Greenblatt and Graham parameters are those defined in Reese’s
“Guru
Investor” and used on the model portfolios on www.validea.com.
The Adaptive formula was my first attempt at an evolutionary
process. The correlation tables were not
well designed, and so I moved onto my Sector and Style Matrix – which led to
the ESI disaster.
The 5 Year Sales Growth was a complete accident. I was attempting to reconstruct a Graham model
and had an error in my script that only retained the final sort.
So, two accidents and a disaster – but the second accident
was a fortunate one that led me to re-evaluate the concept for an adaptive correlation
matrix. Obviously the returns are wildly inflated by
short term moves in the market, but I do know that the algorithm is sound, and
the results make sense to me:
Ø
Debt: should be low relative to capital.
Ø
Volume: should be high (i.e. not some kind of
thinly traded micro-cap).
Ø
Long term Beta: should be high (if I want to
out-perform the market, then I have to find a stock that moves more than the
market).
Ø
Analyst estimates for earnings, book value
growth, cash flow, and return on capital: all should be low. I want a stock that analysts despise so much
that they have chased away everyone who was likely to sell.
Ø
3 and 5 year stock return: disasters. You can’t
buy low unless the stock is low.
Ø
Sales Growth: 1 year, 5 year, 10 year sales
growth should all be high.
Basically the model wants a low debt company that is growing
faster than the stock, and has a lot of talking heads scaring people away. And it
wants to find that stock when its entire industry has been blown by extreme
panic.
Granted, the new adaptive filter has only been in place for
about 6 weeks, but the metrics make sense to me and they are performing well so
far. My goal is only to outperform the
Sector Model, hopefully by about 5% a year.
If I get a 10% kick I’ll be pleased beyond belief. The 745% is just an illusion of timing, so we’ll
have to wait and see. But I feel that I’m
on the right track on the Full Model for the first time since I began
experimenting with fundamentals back in November 2011.
Tim
No comments:
Post a Comment