|
PureBytes Links
Trading Reference Links
|
Don't have much time right now to discuss this, but here's what we are
doing:
1) From historical backtesting, you will get the probability spectrum of
trades. If the trades are in $ figures, it's probably OK, too, but if the
trades are expressed in R-figures (multiplies of maximum risk - you'll get
the idea from Tharp), the probability spectrum will be more consice across
various markets (stocks, futures, currencies).
2) Then you will take the probability spectrum figures and feed it into your
MC engine. You will be able to simulate your trading for any timeframe
depending on the number of trades that your system generates during the
period. So let's take 6 months, 12 months, and 24 months. You can run, for
example, 10,000 different 6-month trading scenarios, 10,000 12-month
scenarios, etc.
3) From the distribution curves that result from the 10,000 MC runs, you
will be able to tell things like these:
- over 6-month period, the system's average return might be 10% with a 15%
standard deviation. However, usually the distribution is not going to be
normal, so these numbers will be quite misleading.
- So you may want to put the output into the form like this: over a 6-month
period, there is 70% probability that the system will be profitable. There
will be 40% probability that the system will generate at least 7%, and 10%
probability that the system will generate at least 15%. And there is 20%
probability that 6-month period will result with a loss of at least 10%
- You can also measure maximum drawdowns, distribution of avg win/loss, %
profitable, profit factor, etc.
Overall, looking at your system's results through the MC "glasses" will be
an eye-opener. It's like going from 2D to 3D, or like going from black-and
white to color.
Ivo
-----Original Message-----
From: Volker Knapp [mailto:vk@xxxxxxxxxxxx]
Sent: Saturday, April 20, 2002 4:53 AM
To: DH; Omega List
Subject: AW: Monte Carlo Simulations
What do you think is the right way of doing MC? If you test a system on a
portfolio of stocks/futures and apply certain money management rules. How
would you do the MC then? We are planning to incorporate MC in our software,
so maybe some of you have any thoughts? We all know that results change
dramatically when using MM and have different starting dates. One way we
could think of is running portfolio-tests with random starting dates...but
is that enough...should we take the results and create random equity curves
afterwards... but then what do we get? So if you have any thoughts about it
and dont want to discuss it here you can email me in privat, thanks.
Volker Knapp
Wealth-Lab Inc.
http://www.wealth-lab.com
http://www.wealth-lab.de
++-----Ursprungliche Nachricht-----
++Von: DH [mailto:catapult@xxxxxxxxxxxxxxxxxx]
++Gesendet: Samstag, 20. April 2002 03:07
++An: Omega List
++Betreff: Re: Monte Carlo Simulations
++
++
++> what benefit are these obscure( IMO) probabilities?
++
++Money management.
++
++Most people can generally agree with the concept that you should never
++risk more than X% of your account on each trade. The question becomes
++how big is X? Monte Carlo can give you an answer for YOUR account,
++trading YOUR system, with YOUR risk tolerance.
++
++> : Specifically... what practical benefit is it to know that there's a
++> : 5% (or x%) chance my maximum drawdown (or any drawdown)
++will be exceeded
++> in
++> : the next 60 (or however many) days? How does that help you
++make even one
++> : decision regarding your trading other than raising a red
++flag as to when
++> : your system may/may not be performing within a wide range of
++> probabilities?
++
++You seem confused about the concept. The MC is done during system
++development before you ever start trading the system. The practical
++benefit comes from deciding how big to trade. Can you handle a 50%
++chance of going broke (in return for a 50% chance of getting rich
++quick?) Or do you prefer a slow and steady approach with a very small
++probability of blowing out the account? The MC can tell you how much
++leverage to use to accomplish each of those goals.
++
++It's not for everybody and it's hard to do it right. If you don't
++understand it, forget about it. All it will do is confuse you.
++
++--
++ Dennis
++
|