|
PureBytes Links
Trading Reference Links
|
This is an interesting subject for me, having spent a good portion of my
daily system research/testing trying to solve this riddle.
The assumption of the stationary Normal Distribution is without a doubt a
key reason why most statistically based systems fail in real life trading.
Fisher Black recognized this as the Achilles heal of his famous Black-Sholes
pricing model. And in fact, being able to pin down which distribution model
is in effect at any given time is kept top secret by those suspected of even
being able to do it, like O'Connor or DE Shaw for example. It's the key to
the vault afterall!
You are partially correct Lionel about the Log Normal if you are
considering a distribution of prices or anytime the Normal would apply but
be bounded on the low end by zero. But because a distribution of prices (or
trading returns for that matter) tends to also show varying skew and
kurtosis, as well as infinite variance, these distributions tend to be like
a whole class known as Stable Pareto-Levy. P-L are a bunch of shape-shifting
demons!
To make pinning down the distribution model even more difficult is that they
are non-stationary over time. It is the main reason trading systems go thru
up/down phases. Many people mistakenly assume it's because the parameters
have changed and will try to re-optimize them. But when systems go into a
down phase the performance of the system will suffer at any parameter value.
That's why once you have a system built you should refrain from continuously
trying to tweak it to match current market conditions. Just leave it alone
and it can eventually turn profitable again. If the urge to tweak hits,
build a new system!
FYI, If you are interested in determining what type of distribution may be
at work in your system you can compare using the Chi-Square or the
Kolmogorov-Smirnov (K-S) tests.
cheers,
Rick
-----Original Message-----
From: Lionel Issen <lissen@xxxxxxxxxxxxxxxx>
To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
Cc: Onno Goedknegt <goedkneg@xxxxxx>
Date: Tuesday, August 04, 1998 7:09 AM
Subject: Re: Fishback's ODDS
>Onno:
>I think I've read that the distribution for financial markets is more
>nearly a log-normal distribution; however the tails are thicker meaning
that
>there is more data that lies outside the log-normal distribution.
>
>I think I saw this in a book about chaos in the financial markets, possibly
>by Peters.
>
>Lionel
>-----Original Message-----
>From: Onno Goedknegt <goedkneg@xxxxxx>
>To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
>Date: Monday, August 03, 1998 2:07 PM
>Subject: Fishback's ODDS
>
>
>>Hi everyone,
>>
>>A few days ago I received Fishback's ODDS book and video.
>>
>>ODDS principle:
>>1. calculate historical volatility using a spreadsheet
>>2. by using statistical analysis (normal distribution: Bell curve)
>>calculate the probable price range in the coming period
>>3. sell a stangle with stikes at the extremes of the price range
>>4. protect this short position by buying a higher call and a lower put
>>
>>My questions are:
>>1. can normal distribution be used to predict prices in the financial
>>markets?
>>2. which period is the best to use for calculating historical volatility?
>>3. can historical volatility tell you something about the future prices?
>>4. can you earn money using this ODDS method in practice
>> (remember bid/ask, commissions, slippage etc.)?
>>5. I don't understand the similarities between the ODDS system explained
>> in the video/book and the ODDS system in Metastock 6.5 ? The only
>>similarity I can see is: use volatility in option trading.
>>6. ODDS was published in 1994. In those years OEX volatility was much
lower
>>than today. Maybe ODDS doesn't work that well anymore in 1998 and maybe
>>that's the reason it's now published...?
>>
>>I found it interesting to read ODDS. I have the impression ODDS can make a
>>profit in the long run, because it is a "continuous strategy". You don't
>>have to wait for a buy or sell signal.
>>But I do think a lot of experiments have to be made before using this
>>strategy, for example: what volatility period is best? Sell short period
>>options and buy longer periods (theta factor!)?
>>
>>Is there someone out there who can tell me more about ODDS?
>>Or is able to share hers/his experience with this system?
>>
>>Thanks!
>>
>>Regards,
>>Onno
>>
>>
>
>
|