Trading
Systems 101
The vast majority of traders in
the market basically trade without much of a plan. Most believe
that all you have to do to make money in the market is listen
analyst recommendations or breaking news and then they will be
set for life.
Wrong!
Everyone
I know that has "played" the market using news from
the talking heads on TV or the Internet has eventually lost
their money and quit. If over 90% of the players in this game
lose, and 90% use the news to trade...hmmmm...are we
seeing a correlation here?
Trading
was like playing a slot machine
Once
upon a time, I too was caught up in scanning the news for
trading ideas. Every now and then, I would catch a nice win or
two, but it was like playing a slot machine - I would win just
enough times to keep me playing.
I've
always been an analytical type of person, so I began my search
for a repeatable trading method. After reading several
interviews of successful traders, I decided to go the
computerized route. Many of the best traders in the world were
using computerized systems to generate their buy and sell
signals, and that fit in well with my idea of how trading should
be.
10 years later...
That
was a decade ago, and in that time, I've learned 10,000% more
about trading systems and what makes them tick...
Trading
System Philosophy
The
philosophy of trading system design is to use the past to figure
out the probability of something happening in the future. That
doesn't mean we can predict the future - we can react to it
however.
I think of
trading as I do playing card games like poker. The big
difference is that the rules for poker are very well established
and repeatable. With the stock market, the rules aren't
published, so I have to use the past to come up with the rules
of how I think the game should be played. I put my ideas into a
computer which in turn tells me if I would have done well (by
making money) in the past with those sets of rules.
There's
nothing new under the sun
The
future is unknowable, but knowing what's happened in the past
can give a glimpse into the probabilities of future outcomes.
After all, human emotion is the same now as it was 1000 years
ago (and I contend that it's these emotions that create edges
for us to exploit in the markets).
The
primary way to keep score in the stock market game is by how
much money you make - not how often you're right or wrong. You
see, if you won 30% of the time, but your winners were 3 times
the size of your losers, you would be making money.
Most
traders focus too much on their own ego instead of playing the
game to make money. That's why even Einstein type
brainiacs with bulging IQs fail at trading. When you get the
emotion out of the picture and focus on playing the game on
its terms, you can find a constant source of wealth.
Crunching
the Numbers
It's
really not just enough to make money from the market. I want to
not only make a bunch of money, but I want to do so while
keeping my drawdowns as small as possible. If I make 80% in one
year, but I had to stomach a 60% drawdown during that
time...well, that's not a good trading system. In fact, a
trading system like that could very well make you lose all your
money at some point in the future.
Tested
Analysis
So how do we
go about finding systems with potential? By testing, testing,
and retesting. I like to say that I don't use technical analysis
to find my trades; I use tested analysis.
I'll
pour through charts for hours on end until I see what appears to
be an edge. I'll then program my ideas into a computer (or
computers depending on how much number crunching is required).
Next, I'll refine the trading criteria by testing different
combinations of the indicators I've fed to the computer.
All
told, this process takes a VERY long time, as even my new
computers have a hard time crunching data for thousands of
stocks for 15+ years. If it's hard to do, then I know I'm on the
right track. If it was easy, then everyone would be doing it and
I wouldn't have an edge.
Enter
gain/pain ratios and a plethora of other statistics.
One
of the most common statistics to measure performance is the MAR
ratio. Basically, you take your compound annual growth rate
(CAGR) and divide by your worst drawdown during that time. So if
my trading system has a CAGR of 50%, and my worst drawdown was
25%, then the MAR ratio is 2.0. The bigger the better.
In
general, I've found that the shorter the timeframe and the more
trades, the bigger the MAR. A very short-term trading system
like our Profit Taker has a MAR around 3...even during the worst
bear market since 1929. If you look at its equity curve with
drawdowns, you will see that it's been a steady money maker.
When
dealing with longer trading periods, the gain/pain ratios tend
to get smaller. The benefit is that you trade less, and returns
can be very large, but these systems take more of a hit from
time to time. Trend following systems tend to have lower MAR
ratios in the stock market.
Another
important stat to me is the length of time between new equity
peaks. I don't want to trade systems that go on more than 12
months before making a new equity high. It's very hard to stay
focused after that long of a drawdown.
The
MAR ratio is very easy to understand, but I find it lacking.
Just knowing your growth rate and worst drawdown is not enough.
I want to dive in and use a number that penalizes me for not
only big drawdowns, but the length of time between new equity
peaks. Peter Martin came up with the Ulcer index which does just
that.
It gives numbers
between 0 (a perfectly straight line with no drawdowns) and 100
(a system always in drawdown...which would sure give me an
ulcer).
By going through
your equity curve, it penalizes you by the drawdown squared each
time you're below the equity peak. You can then take your annual
growth rate and divide it by the Ulcer index. I think this is by
far the best way to compare systems. Thank you Peter Martin.
Money
Management
In order
to make money while keeping risk low, you have to use proper
money and risk management. Most see this as an after thought to
trading. It's not! In fact, it should be thought of as the most
important aspect of trading. If you risk too much and lose all
your money, you're out of the game.
For
every trade you make, you should know exactly how many shares
you're going to buy, and how much money is at risk at any one
point. For example, if you lose money on any one trade, perhaps
it should not be more than 1% of your total portfolio. You could
calculate how many shares to buy based on your entry price and
stop price:
# Shares = (%
Portfolio risk) * (Portfolio size) / (Buy price - Stop price)
So for a
$50,000 portfolio risking 1% (or $500):
# Shares =
(0.01) * ($50,000) / (25.45 - 23.04)
# Shares =
207
When trading stocks there is
more to worry about than how much you're going to risk on any
one trade. You see, stocks are highly correlated to each other.
Not only are you risking money on each trade, but you're risking
that all those stocks will fail at the same time. Therefore, you
must limit the number of stocks a system is trading at any one
time.
My testing has
showed a very easy way to blend fixed fractional betting with
risk management. Simply divide the money to risk evenly between
10 or more positions. For example:
#
Shares = ( Portfolio size / 10 ) / Entry price
#
Shares = ( $50,000 / 10) / (25.45)
#
Shares = 196
Now,
we have limited the number of positions (10) that can go against
us, and we've reduced the maximum risk in case a stock went
belly up (if one of our stocks went to zero, we would be out 10%
max).
Statistical significance
One
of the biggest errors I see people make when testing systems is
statistical significance. Sometimes, a particular pattern will
come up only 10-15 times in history. The results look great, but
when applied to the real world, they fail. You see, I could
randomly choose a particular stock and random date to buy on the
open, then sell on the close. There are thousands of stocks, and
about 250 trading days in a year. I bet I could find a few
examples that win 100% of the time just due to random chance.
When
I look for an edge in trading, I want to see many examples...not
just a handful. When it comes to choosing from thousands of
stocks, I want to see hundreds if not thousands of samples. That
way I can determine that I have not simply come up with a random
sampling of trades that just happens to work well.
The
stock trading systems described on this web site all have
thousands of samples for back testing - many more than actually
listed since every system has a limit of 10 positions at any one
time. The software I use filters out trades if the maximum
number of positions is reached.
Curve
fitting
I've seen
this mistake over and over again. A trader gets a basic back
testing package and proceeds to test which moving averages work
for a particular stock. Ouch! This is a quick way to lose money.
If you're only using one stock to calculate the optimized
values, you're going to run into the problem outlined above (not
enough samples for statistical significance).
One
set of rules for thousands of stocks
If
instead, you were to scan thousands of stocks for a few
combinations that work for all of them...well that's a much more
significant statistic. When I'm testing what works for stocks, I
test them ALL. I want to use the same rules for thousands of
stocks...only then do I believe I have discovered an edge.
Accounting
for Commission, margin interest, and slippage
There
are just way too many traders out there that do not realize just
how much the extra expenses in trading amount to. The great news
is that commissions are falling dramatically (interactivebrokers
is down to 1/2 cent a share now). However, you still have to pay
interest to your broker for using margin when buying or selling
short, and you must account for slippage.
Slippage
is the amount of money you lose when a stock does not trade
exactly at your entry price. Say I place an order to buy at the
open. The open was 20.00, but my fill price was 20.07. Your
slippage is 7 cents times the number of shares bought (7 cents *
1000 shares = $70). I always include slippage when testing
systems. I usually estimate slippage to be between 5-10% of the
day's range. If a stock moved between 20.00 and 21.00, the
slippage was most likely between 5-10 cents.
Eventually,
it's very possible to move the market with your orders. There
are two ways I combat this: by only trading liquid stocks that
trade several million dollars worth of shares on average, and by
using stop limit orders.
Stop
limit orders allow you to place a ceiling on how much your
willing to pay for a stock. For example, stock XYZ is trading at
19.00 and I place a stop limit order to buy at 20.00/20.06. When
the stock hits 20.00, my order to buy at a limit of 20.06 is
placed into the market.
System
Trading Terms and Formulas
CAGR
- Compound annual growth rate. It's no good to simply take your
percentage gains and divide by time. If you made 50,000% over a
50 year period, that's not the same as saying you made 1,000% a
year. You would apply a formula based on compound interest
instead:
x
Sqrt(Percent gain) - 1 (where x is the number of years)
(50)
Sqrt(50,000%) - 1 = 13.23% (Note that I'm using the special
xsquareRoot function found on most scientific calculators. I'm
not multiplying by 50 in the example above).
MAR
- A simple measure for gain to pain. Take the compound annual
growth rate (CAGR) and divide by the worst drawdown. If I have a
50% CAGR, and my worst drawdown was 25%, my MAR ratio is 2.0.
Ulcer
Index - Created by Peter Martin, this gain/pain stat ranges
from 0-100. Zero would be a perfectly straight line with no
drawdowns, while 100 would be straight down (thus giving you an
ulcer). Both the depth and duration of a drawdown are measured.
I
like this statistics model because unlike the MAR ratio, it does
not over-penalize you for what by definition is a one time event
(your worst drawdown). Please see this web
site for more info.
Martin
Ratio - By taking your CAGR and dividing by the Ulcer index,
this number will give you a much better number to compare
trading systems. Sharpe ratios, and MAR ratios are inferior
comparison models in my opinion.
Next
Article: Trading Systems 201: Advanced System Ideas
Click here for instant access to your free
trial
The results listed herein are based on hypothetical trades.
Plainly speaking, these trades were not actually executed. Hypothetical or
simulated performance results have certain inherent limitations. Unlike an
actual performance record, simulated results do not represent actual trading.
Also, since the trades have not actually been executed, the results may have
under (or over) compensated for the impact, if any, of certain market factors
such as lack of liquidity. You may have done better or worse than the results
portrayed.
