|
PureBytes Links
Trading Reference Links
|
Ivo,
I see your point and it is a good one, but I respectfully disagree that
the actual drawdown means nothing especially over 10 years. If MCS is
saying that the second scenario is so likely, why didn't it occur even
once. If I tell you I have a fair coin and you get three heads in a row
I might believe you. But after 10 heads in a row, I will start to have
serious doubts. Of course, if you cluster all my losing trades together
you will have a huge drawdown, but does that mean that outcome is
particularly likely? Maybe I just don't understand MSC.
Gabriel
-----Original Message-----
From: Ivo Karindi [mailto:ivo@xxxxxxxxx]
Sent: Thursday, November 20, 2003 2:25 PM
To: Gray, Gabriel
Cc: omega-list@xxxxxxxxxx
Subject: Re: Monte Carlo Simulation of DrawDowns
Well, the MCS is telling you the probabilities you will face given the
properties of your strategy. The fact that backtesting shows only a
certain amount of DD means nothing. Consider these 2 series of
trades:
$200
-$100
$200
-$100
$200
$200
$200
$200
-$100
-$100
In the second case, is this strategy really doing a worse job than the
first one because it's maximum drawdown is twice as much? When applying
the rest of the relevant statistics (% profitable, avg win / avg loss,
profit factor, etc.) the strategies look identical! Probably the case is
that your backtest is showing you the first scenario, but MCS tells you
that also the second scenario is very likely to happen given your
system's parameters.
Ivo Karindi
GG> Hi all,
GG> I recently bought this software MCS Pro that performs
GG> portfolio Monte Carlo simulations through tradestation strategies.
GG> It outputs a distribution that displays the probability of achieving
GG> a drawdown that would exceed a certain value over a fixed period of
GG> time. So, there is a list of drawdown values with associated
GG> probabilities of a drawdown exceeding that amount. The strange part
GG> is that the probabilities seem incredibly high. One system I tested
GG> had a maximum actual drawdown of 83,000 over a ten year period. But
GG> when I look at the probability of a drawdown exceeding 83,000 over a
GG> one year period it is 63%. I am no mathematician, but to me if the
GG> probability of witnessing a greater drawdown in one year is 63%,
GG> the probability of witnessing it in 10 years would be 99.9952% (1
GG> minus the probability of it not happening
GG> (.37) raised to the 10th power). I know these events may not be
GG> independent, but with these odds I would expect to witness at least
one
GG> drawdown greater than 83,000 over ten years. I am missing something
GG> here? I feel like I am taking crazy pills. Any assistance would be
GG> appreciated.
GG> Gabriel
|