[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Criteria for determining a "good" trading system



PureBytes Links

Trading Reference Links

Hello list,

I am not a mathematician, or a retail trader  the only thing I know how to 
do is trade and have been doing so for 30 years.  Up until a few months ago 
never once took the time to consider what parameters would indicate a good 
mechanical system as I was the system. I did spend a lot of time devising 
my own indicators, and how I perceived the market.  Two reasons prompted me 
to start thinking about mechanical systems. One I would like to continue to 
make some money from the markets after retiring, and two (and perhaps the 
most important reason) was that a friend and fellow trader told me that it 
couldn’t be done, and made me an interesting bet. I love a challenge.

In any case it was my desire to define certain criteria that anyone could 
use to be able to state, “yes this is a good system, or no it is garbage”. 
We can use mathematical equations and ratios to define if a race car is 
average or above average. We can define the odds of winning at craps, 21, 
or the likelihood of being struck by lightening. By defining these numbers 
or ratios we can then begin to classify race cars, or a good betting and 
money management system for playing craps. By understanding the 
relationship between these ratios we can understand what parameters we have 
to modify or use to create a great race car. If we were chasing lightning 
bolts (for whatever reason) by understanding the math and science of the 
lightning bolt, we could increase the possibility of being in a location 
that would be struck by a lightning bolt, or far away.

Why use mathematical ratios and formula’s to define a good trading system? 
Easy, because in trading we are engaged in an activity that on a 
trade-by-trade basis has 0% certainty and 100% possibility of a 
non-expected event transpiring. Only by considering a large number of 
trades can we change the 0% certainty to a ‘n’% certainty. However, 
because all of the ratios are based upon historical data we are unable to 
predict with any amount of certainty that the “next trade” will definitely 
be a winner. All we can predict with any degree of certainty is that in the 
next ‘n’ trades the probability is …..

The importance of accurately defining the criteria for a good trading 
system is in my humble opinion the most important aspect of creating a good 
trading system. Correct criteria will allow us to know when we have created 
a good trading system. It will help us in the creation of the trading 
system. Of even more importance is that by knowing the criteria we will be 
able to determine when our trading system should not be used. As the 
performance of the system (using the criteria) has begun to degrade as the 
market conditions have changed from the conditions used in creating the 
system. This will help us avoid giving back the majority of our profits.

In short we are engaged in an endeavor, which requires us to accept the 
uncertainty of each trade, and devise our trading logic to work with 
probabilities. To work with probabilities means that we must define certain 
criteria that would indicate a good system versus a bad system.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

My answer to some of the follow up post follows:


< Do not confuse the Sharpe Ratio of systems back tested in TradeStation
with the Sharpe Ratio obtained in real life by real traders or money
managers - the two are quite different.>

This is totally true  it is like comparing apples to oranges. The reason 
these large hedge funds have the numbers they have is to a large part 
because of their size. The larger your equity  the larger your bet must be. 
As your bet size increases the more time you need to see your bet bear 
fruit, i.e., the larger your loss if your bet is wrong. I am very happy 
riding my motorcycle when stuck in traffic on the 101 in LA. Every large 
fund that I know of has a problem in profiting from small profit 
opportunities that manifest themselves in small time frames. The directors 
of these funds spend more time studying the "big picture" than in staring 
at the screen. I would rather trade 10 million than 100 million. There are 
some CTA’s that will have a Sharpe Ratio close to 1.0 for any one year. In 
using the Sharpe ratio for system design it will reflect either; trades 
including commissions and slippage, or without commissions and slippage. It 
will not include the other related trading cost that goes into the 
performance figures of CTA’s, etc.

Another factor is how the performance of a CTA is measured. For example let 
us take the performance figures of Hanseatic Corp. in Albuquerque, 
N.M.  they are a registered CTA, whom I use to allocate some money to while 
at Lind Waldock. Using their performance figures from 2/96 to 4/99 (sorry I 
don’t have more current figures with me here in Russia) they had an 
annualized Sharpe Ratio of 1.63, average annual return of 123%, maximum 
drawdown 26.74% (lasting 14 months), and a margin/equity ratio of 25%. It 
is this last number that is the key to understanding their performance.

In the majority of cases they are using less than 25% of the clients total 
deposited funds with them to make their trades, but with the maximum being 
fixed at 25%. Many times when people look at the returns of a CTA they 
think the performance sucks because the annual return is not so good. 
However they are forgetting to factor in the margin equity ratio, which 
when factored in will change the performance figures significantly upwards. 
Many people think that a CTA is trading with all of the monies  when in 
actuality it is only a fraction of deposited monies. Money management is of 
critical importance when playing in a game that is dominated by 
uncertainty, and high leverage factors - every CTA knows this all too well. 
If 25% of the total investment was being traded then if one desires to 
compare Hanseatic to say some mutual fund then one would multiply the 
annual performance by a factor of 4, or a annual return of 492%.

If that seems high then consider that another CTA that I am a friend with 
Michael Clarke of Clarke Capital Management has a Domestic Diversified Fund 
that has been traded since 1990 and has averaged annual returns of 36.37% 
using a margin/equity ratio of 8-24% - with 5% being more common. Assuming 
24% this still translates to basically 145% annual returns. Michael is 99% 
totally mechanical. His annualized Sharpe ratio is 0.82.

< Chande goes through analysis similar to yours. To give you a taste of it, 
Chande proposes the following as a benchmark or checklist for a "good" 
mechanical trading program. Chande asserts that the following numbers are 
typical of the long-term performance of professional money managers.

Total Entry Conditions / Parameters <= 10
Total Exit Conditions / Parameters <= 10
Number of markets traded >= 10
Test period/data >= 24 months
Equity risked per trade <= 2%
Proportion of profitable trades <= 50%
Average monthly return 1%
Standard deviation of monthly returns 5%
Return Efficiency >= 0.2%
Expected depth of drawdown 20%
Expected duration of drawdown <= 9 months
Proportion of profitable months <= 65%
Expected Return > 13%
 >

I have talked with Chande 4-5 years ago (have not read his book), and agree 
that these numbers are okay for managed money managers. However this begs 
the question, “What other criteria can we add to define a mechanical 
trading system on a trade by trade basis? Is using months as unit of time 
the best unit to use?

Chande is using a unit of time (months) that perhaps could also be applied 
to a trading system  - but then again perhaps instead of measuring; 
standard deviation, drawdown, run-up, etc., using months as a unit of time. 
Perhaps weeks, or days as a unit of time would be a better measure of time 
to determine these numbers.

< You are talking about losing $100 each time which is, most probably, your 
stop. With a stop like this there'd be not much of a chance of making 
$20,000 because, again most probably, your stop would've been hit before 
market went your way. Losing 20 times with stops adequate for surviving the 
paper drawdown before the big trade comes is something for Ed Seykota, but 
I wouldn't recommend it to anybody below the wizard level.
 >

I agree, however losing $100 in each trade might not be indicative of a 
hard coded stop, as in “if profit in trade is less than or equal to 
$100  then exit trade” Rather is might be “if this, this, this, and that - 
then exit trade”. If it is a hard coded stop then Jan you are correct that 
the $20,000 winner will probably never be realized. However if it is not 
and it is the latter example - then the $20,000 winner would probably be 
realized. And the reason the system is generating 20 consecutive losses 
around $100 is because the price behavior did not perform as expected.

Thanks for the replies

John