Does Efficiency Improve Stock
by Van K. Tharp, Ph.D.
For those of you are new to this series,
I’m going to repeat a few things to get you up to date on the
Tharp efficiency studies. I’ve
underlined the key points. A
number of times, I’ve looked at efficient stocks and found that
buying such stocks long produces a very good trading system.
For example, from July 2006 through July 2007, during a quiet
up market, 23 such trades produced a System Quality Number SM
of 4.08. That’s a
better record than any of the newsletters I tracked in the Second
Edition of Trade Your Way to
Financial Freedom. As
a result, we started a series here in which I was looking at trading
efficient stocks within the S&P 500.
The initial results looked quite good, but the results became
poorer as we began to filter out various errors and assumptions that
we were making. The
results were acceptable, but not as good as the two times that
I’ve publicly traded such stocks (i.e., see Tharp’s Thoughts
July 18th and in Market Mastery, issue #111) through my
Our criteria for trading efficient stocks in
our historical testing were as follows:
First, the efficiency was calculated by looking at the
change in price over a set time period and dividing that value by
the ATR over the same time period.
We used a composite efficiency consisting of four time
periods from 180 days through 20 days.
Second, we only looked at stocks with a composite
efficiency over +8 in the S&P 500 on the first trading day of
We then looked at two sorting algorithms based upon
smoothness and only bought the top 10%, based upon those algorithms.
The first was the standard deviation of the daily
change in the close (close minus close) over 180 days.
This gave us great results, but was biased toward low priced
As a result, we also looked at the standard deviation
of the daily close divided by close over the 180 days as a smoothing
But were these really the efficient stocks
that I’ve gotten such good results from trading?
This is important to know because I’ve simply used
efficiency as a screening tool and then selected the stocks I wanted
to trade based upon a visual inspection of the charts.
As a result, in the last newsletter of this series, I asked
three volunteers to look at nearly 200 stocks from both smoothing
functions to tell me whether or not they were efficient.
One volunteer did 191 of the close minus close smoothing
functions and concluded that only 49.2% were efficient.
Two volunteers worked on the close divided by close smoothing
function and one concluded that 33% were efficient and the other
concluded that 35% of them were efficient.
Those results basically told me that we were probably not
looking at stocks that I wanted to trade and that I needed to go
back to the drawing board with our algorithms.
However, I then started to look at the
charts. I’d told my
volunteers to look at the price for the six month period prior to
the entry to determine whether or not the stocks were efficient.
But when I started to look at what they were defining as
“efficient" verses "not efficient,” it was clear to me
that everyone had a slightly different interpretation.
I’ve always thought that my definition of
efficiency was fairly subjective—the stock is going up in a fairly
smooth line. But this
exercise told me that my criteria are a lot more detailed than I
So let me give you some new ideas that
typically describe what I thought was a fairly subjective decision.
First, the risk-reward ratio is important to me.
Thus, if I’m using a 25% trailing stop, I’d like
something that I believe has the potential to go up 75 to 100%.
Thus, the first thing I’d like to do is eliminate those
stocks that don’t have that potential.
I believe I have three criteria here.
I don’t want to trade something that is
approaching a prior high that could provide a strong resistance
point. Thus, the stock
might have been going up for six months in a straight line, but if
it’s right at an old high from say six months or a year ago, I’m
not interested in trading it.
The stock doesn’t necessarily have to be at an all
time high (or at an all time high with a slight retracement), but if
it’s had much higher highs (e.g., in 1999 to 2000), then I want
it to have the potential of at least a 3R move before it reaches
Next, I believe that something that is starting to
become parabolic is probably near the end of its potential move.
Now that was not necessarily true in 1999 when stocks could
easily move 5% each day and have the potential to go up another 100
to 200%, but the stocks that did that also tended to fall 25% or
more in a single day when the move was over.
As a result, I tend to avoid stocks when the uptrend gets too
Second, although I’m looking for six month
efficiencies (i.e., I’d like something that’s been going up for
six months), I’m fine with something
that’s been going up for three of the last six months as
long as it is higher than it was six months ago.
Third, the stocks I’d trade do not need to be at new
six month highs. In
fact, I’d generally prefer something that’s done a slight
stocks, in fact are often great buying candidates.
And, although I’m using 25% trailing stocks, simply because
I want to hold them for a long time, others could do well trading
these stocks simply by using the retracement amount (i.e., once the
trend resumes) as the potential stop.
However, these could produce very short-term trades so I
don’t trade them.
Fourth, the current up movement is usually defined by
some sort of trendline. I
am not interested in buying anything in which that trendline has
been broken. Will it
keep going down? I have
no idea, but I think the chances are much better when the trendline
is broken, so I no longer call them efficient stocks.
Fifth, I will avoid stocks that are in clear
uptrends but are too volatile.
These criteria pretty much determine what I
look for visually in a stock when I do my efficiency trades.
As a result, I looked at the charts I’d been sent based
upon these criteria. First,
I found that my agreement with the reviewers was about 80% with
their judgments of efficiency, and I had not given them the above
criteria. For the close
divided by close algorithm, I’d probably take about 53% of the
trades the computer took. But
his is better than the 33-35% taken by the reviewers.
Second, for the close minus close
algorithm, I’d probably take about 62.6% of the trades versus
about 49.2% but I might have rejected more if I could have seen back
to view old highs, etc. However,
for these charts we only had six months of data, so I couldn’t
tell if the old highs would have caused me to reject a lot more
Thus, my key observation from all of this
work is that the algorithms done so far with Mechanica, while
producing nice, but not great, profits, were flawed because they
really were not trading what I’d select as efficient stocks.
But what I didn’t realize is that there was a hint of
how good the efficiency concept was even in our data.
And that’s where a little study done by Pater Kolf became
Pater Kolf Study
Pater asked himself, “If I screen these
stocks using Van’s criteria (above), is there a difference in the
profitability of the efficient versus the non-efficient stocks?”
The following table shows the results he found.
He only looked at 50 of nearly 200 stocks from each filtering
process, so there could be a sampling error.
In addition, he found that more of the stocks were efficient
than I did, but could be due to 1) using a sample of 50, or 2)
divided by Close
Pater’s conclusion from the results
was as follows: “It
appears that the reliability (whether or not a stock is profitable)
is not much different whether a stock is efficient or not, but the
total profits appear to be higher in the efficient stocks.”
Indeed, it is five times higher for the close divided by
close database and nearly twice as high for the close minus close
My conclusion is that while sampling
and subjective errors could have entered into Pater’s conclusions,
the results definitively support my overall conclusion that
efficient stocks are good trading vehicles.
However, I have not looked at the data and don’t know how
well my judgment of efficiency would agree with his.
It was certainly much lower than 66% efficient with the large
number of samples that I looked at.
Bob Spear, the developer of
Mechanica, will be working on a new algorithm that better fits my
criteria for an efficient stock.
When that is developed, we will be continuing this series.
For more information about Mechanica visit www.mechanicasoftware.com.
Tharp: Trading coach, and author, Dr. Van K. Tharp is widely
recognized for his best-selling book Trade Your Way to
Financial Freedom and his outstanding Peak Performance Home
Study program - a highly regarded classic that is suitable for all
levels of traders and investors. You can learn more about Van
Tharp at www.iitm.com.