6/7/2015 The Full Model is back in the lead

Sector Model	XLU	-2.94%
Full Model	Date	Return	Days
JOY	12/8/2014	-23.65%	181
PWR	3/9/2015	3.78%	90
BHE	3/31/2015	-5.28%	68
CBI	4/2/2015	16.74%	66
MTZ	4/9/2015	-0.72%	59
NE	5/7/2015	-2.79%	31
DRQ	5/15/2015	-0.57%	23
RES	5/19/2015	2.97%	19
CRR	5/19/2015	6.47%	19
SPN	5/28/2015	-0.17%	10
(Since 5/31/2011)
S&P	Annualized	11.62%
Sector Model	Annualized	19.16%
Full Model	Annualized	19.25%
S&P	Total	55.58%
Sector Model	Total	102.28%
Full Model	Total	102.88%
Sector Model	Advantage	7.53%
Full Model	Advantage	7.62%
	Previous	2015
S&P	53.06%	1.65%
Sector Model	122.60%	-9.13%
Full Model	101.13%	0.87%

 

This no-good, horrible, very bad year has come with a twist I didn’t expect: the full model has caught up with the sector model again.

Part of that was from my finding and correcting the data flaw from my Value Line files (caused some years ago by a migration to another computer and not by Value Line itself).  I was basically selecting stocks at random for over two years instead of selecting on the fundamentals I was trying to filter for.

Part of that “catch-up” is from the Sector Model experiencing a spectacular collapse:

 

In the entire back-test & live combined chart the reason appears to be simple mean reversion:

 

In other words, the last two years were better than normal, and the model is reverting back to the mean.

The only question is if there is a pattern to this phenomenon, and to that that idea I’ve analyzed the S&P and the Sector Model against their own standard deviation channels.  The slopes are different, of course, but by super-imposing their current deviations from their respective slopes we can see if there is a pattern of over and under performance for the Sector Model:

 

These lines are smoothed 12 month moving averages of their positions within their own standard deviation channels.  Except for the great recession and quantitative easing, the Sector Model appears to do best in the early part of a trend, and worst in the latter part of a trend.

So in the first half of a bull or bear market, it does best.

In the last half of a bull of bear market, it does worst.

The reason for this is in the nature of the model itself. The first half of a market move is characterized by mean-reversion. The last half by momentum.

Since I have a mean-reversion based model, it has its best performance during mean reversion periods, and its worst performance during momentum periods.

Ironically, I mean-revert when the market goes into momentum, and I have momentum when the market mean-reverts.

The model is based on breadth and volume. In periods of momentum price continues even as breadth and volume dry up.

It’s tempting to try to time these sorts of things, but I’ve never found a satisfactory timing model and I’ve never been able to construct one either. I have a model that outperforms over a full market cycle, but within a cycle it won’t outperform 100% of the time.

This is one of those times.

I’d offer some words of comfort, but there aren’t any. Sometimes even the best poker player has to suffer through some bad hands, and even the best model can have some bad months.

Tim