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

Re: [amibroker] Measure stocks' volatility[response from McMillan]



PureBytes Links

Trading Reference Links


Interesting, however, I suspect he is answered the 
wrong question.
 
In the BS methods, indeed the time used inthe 
equations for the time to expiration is in calendar days. No question about 
that in my mind, nor in the calculations I routinely perform.   Hence 
your observation that the price decreasing over the weekend more than over night 
may be a partial result of that, although I still expect the floor trader's 
hedge against more weekend "event-risk would have more influence, certainlyfor 
longer term options where the theta-decay is very minimal.  The floor 
risk-adjustment show up in implied volatility calculations, and I suspect they 
swamp the theta-decay of 3-days vs. 1-day.  Easily calculated if one is of 
the mind.
 
Calculation of the Historical Volatility, the topic 
of this discussion, is quite another matter.  An easy check is to do the 
calculations yourself and compare to McMillan's values he posts.  I am 
fairly confident you did not ask McMillan specifically about the historical 
volatility calculations ...  Another check is to compare the implied 
volatilities that can be back-calculated in the BS model and compare to the 
historical volatilities.  In either case, my experience and calculations 
indicate that using bars in the stddev normalization is the more 
appropriate.   Difference between using sqrt of 260 vs. 365 is roughly 
20% and the comparisons above are generally within that range.
 
Another question to ask yourself, is givena time 
series with constant volatility, how would you normalize  the calculations 
such that calculations on shorter subsets of the data would extrapolate to the 
same calculation over a year .... although this may be too degenerate a case to 
be convincing.  I doubt if you would use 365 bars, but rather 260 and the 
full year calculations would not require extrapolation.
 
Anthony, my intention is not to "win this"or to 
harangue you.  I have worked options and the associated math for several 
years and feel fairly confident in this answer based upon calculation historical 
volatilities by numerous conditions and numerous variations.  If one is 
only interested in the changes in volatility the scale factor doesn't 
matter.  If you are doing BS modeling yourself, it makes a large 
difference.
 
Best wishes,
 
Richard
 
 
<BLOCKQUOTE 
>
----- Original Message ----- 
<DIV 
>From: 
Anthony Faragasso 

To: <A title=amibroker@xxxxxxxxxx 
href="">amibroker@xxxxxxxxxxxxxxx 
Sent: Tuesday, April 30, 2002 3:50 
PM
Subject: Re: [amibroker] Measure stocks' 
volatility[response from McMillan]
Richard,After the discussion about volatility,I 
took the liberty of sending andEmail to Larry McMillan concerning Calendar 
Days versus Trading Days:His response is 
below./**************************************************/   
Date:        Tue, 30 Apr 2002 15:28:53 
-0400   From:        
"Lawrence G. McMillan" <<A 
href="">info@xxxx>     
To:        Anthony Faragasso 
<ajf1111@xxxxIn the Monte Carlo probability calculators, we 
use trading days becausethemodel is interested only in trading 
"units."  That is, we could use10-minute bar charts if we had the 
data.In the Black-Sholes mode, and thus in implied volatility, the 
standardisto use calendar 
days./***********************************************************/The 
second paragraph, confirms my own use of Calendar Days.As for my 
comment " in the real world" , I was refering to the fact thatoptionshave 
2 3/4 days of time decayover the weekend, and if you are using Trading 
days, this does not takeweekends into account.Best 
WishesAnthonyRichard Alford wrote:> I too havelost 
enough money in options to speak with some validity> :) I compare my 
calculations with those from McMillan's site, the> unquestioned guru of 
options (questioning that statement forfeits you> right to speak about 
options <go>) -> cf<A 
href="">http://www.optionstrategist.com/free/analysis/data/index.html.His> 
20-day historical volatility for, say, INTC on 4/19 is 40% - the> 
sqrt(260) yields 41.2% in AB whereas the 365 would yields 48.6%.  
(BTW> your 256 number you mention actually matches his the best at 
40.2% -> although it is unclear how you arrive at that number - I may 
change to> 256). Your observations on the option variations over the 
weekend> strike me as irrelevant considering that the price is a 
function of> the implied volatility - which is a function of the model 
used etc.> Typically the implied volatility is much closer to the 
20-day> volatility than the 20% discrepancy introduced by using 365. It 
is too> late to delve into the latter points you claim.  Note 
that> normalization typically is a non-dimensional process and if 
you> measure bars you should normalize bars, at least IMHO. Feel free 
to> use whatever you choose. I will go with the numbers that 
match> McMillan. Sorry to get you torqued off  :)  I do 
wonder what you think> option pricing has to do with the real world - 
but lets save that for> another day. Cheers, 
Richard>>      ----- Original Message 
----->      From: Anthony 
Faragasso>      To: 
amibroker@xxxxxxxxxxxxxxx>      Sent: Sunday, 
April 28, 2002 9:24 PM>      Subject: Re: 
[amibroker] Measure a stocks' volatility 
with>      AB 
?>       
Richard,>>      Since I also deal alot 
with Options, Here is my reason for>      
using>      
sqrt(365).>>      Should you use 365 ( 
the number of calendar days ) or 256 (>      
the number>      of trading days), I believe 
there is no absolutely correct>      answer. I 
use>      365 mainly for consistency. Oneof 
the key functions in an>      
option's>      price is time ( t). The time 
component is expressed in terms>      
of>      fractions of a year and is 
universally thought of as>      
Sqrt(calendar>      days/365) as opposed to 
Sqrt(trading days / 256).>>      This 
365 day calendar effect can be best demonstrated 
by>      looking at 
the>      price behavior of an option from 
Friday afternoon to Monday>      morning. 
All>      other things being equal, the 
option's price will drop more>      from 
Friday>      afternoon to Monday morning than 
it would from Thursday>      afternoon 
to>      Friday Morning.  That's because 
you have about 2 3/4 days of>      time 
decay>      of the weekend instead of the 
usual 3/4 day of decay during>      the 
week.>      Just to be consistent with the way 
options behave in the>      real world,  
I>      use 365, the number of calendar days. 
If you use 256, the>      options 
price>      decline would be smooth every 
trading day, no matter how>      many 
calendar>      days transpired between trading 
periods.>>      There is an even more 
troubling problem when you use the>      
number of>      trading days, and that has to 
do with an assets expected>      
rate-of-return>      and risk over the course 
of a year. The formula for>      volatility is 
the>      annualized daily standard deviation. 
Recall that you>      
calculate>      volatility by multiplyingthe 
standard deviation figure by>      the 
square>      root of the number of periods in 
a year. If you use 256(>      now) 
because>      that is the number of trading 
days in a year, that means>      you'll have 
to>      use some other number when the number 
of trading days per>      year 
is>      modified.  That means adding 
trading days would increase>      
volatility,>      eliminating trading days 
would decrease volatility. Trouble>      is, 
changing>      the available trading periods 
has already happened several>      times, 
and>      the real world results refute this 
notion.>>      Prior to 1952 the stock 
market traded on Saturdays. That>      meant 
that>      there were more than 300 trading 
days per year, instead of>      the 
current>      256. When the NYSE made the 
change in 1952, did market risk>      shrink ? 
No.>      Did market returns shrink? No. Would 
diminished returns even>      be 
a>      logical expectation? No. Because 
volatility is a measure of>      risk 
and>      reward, however, that is what should 
have happened based on>      a 
reduction>      in the number of available 
trading days. Not only are those>      
conclusions>      illogical, they're not 
supported by actual events.>>      
Another example is the proliferation of overseas and 
24-hour>      
trading.>      Based on the increasing number 
of trading periods, using 256>      
days>      previously would require that you 
use a larger number now>      for 
certain>      assets ( like stock index 
futures traded on Globex). If you>      used 
a>      volatility calculation model thatused 
trading days in its>      
calculation,>      you'd need to increasethe 
number of trading days to the>      
appropriate>      amount.  With the 
standard deviation component constant,>      
volatility>      would have to increase due to 
the increased number of>      trading 
periods.>>      Some might argue that 
volatility would remain constant, even>      
with the>      increasing and decreasing 
trading periods. If that is true>      and you 
use>      trading days to annualize 
volatility, increasing the number>      of 
trading>      days would imply a smaller daily 
standard deviation figure.>      But 
does>      increasing your sample size ( which 
you are effectively>      doing 
by>      increasing the number of trading 
days) imply a reduction in>      the 
standard>      deviation? Of course 
not.>>      It doesn't even make 
intuitive sense, and is not supported>      by 
empirical>      data.  Taken a step 
further, assuming that volatility>      
remains constant>      while increasing the 
number of available trading days means>      
that>      investors' reactions to news and 
events should result in>      smaller 
price>      moves. For example, lets say that 
a stock goes from $10 to>      $12. 
A>      smaller standard deviation of the 
price change means that>      the 
price>      change must be smaller. So the 
constant volatility premise>      means 
that>      investors are currently valuing the 
stock at an artificially>      high 
price>      of 12 instead of correctly valuing 
it at a lower price,>      simply 
because>      the markets are currently closed 
on weekends.  That is>      wrong. The 
stock>      is priced at 12 because the 
information about the stock that>      
is>      currently available warrants a price 
of 12. Saying that>      volatiltiy 
would>      remain constant when the number of 
trading days changes>      implies that 
the>      market is currently irrational and 
inefficient due to the>      absence 
of>      weekend 
trading.>>      Since I believe neither 
256 nor 365 is perfect , and because>      I 
wish to>      be consistent with the way 
options behave in the real world>      ( that 
is to>      emulate the time deterioration 
over the weekend) I use 365>      days in 
my>      option models and my volatility 
calculations. :-]]>>      Best 
wishes>      
Anthony>>>>      Richard 
Alford wrote:>>      > Anthony,I 
suspect you should use sqrt(260), the 
common>      number 
of>      > bars/year, instead of 
sqrt(365).  The result will agree>      
with published>      > results, for 
instance on McMillan's site. I took the>      
liberty of>      > attaching the indicator 
I use if you have the nerve to>      openit 
with>      > all the virus running 
around.>      > Cheers, Richard ----- 
Original Message ----->      
>>      >      
From: Anthony Faragasso>      
>      To: 
amibroker@xxxxxxxxxxxxxxx>      
>      Sent: Sunday, April 28, 2002 6:49 
PM>      >      
Subject: Re: [amibroker] Measure a stocks' 
volatility>      
with>      >      
AB ?>      
>       Hello, 
Derek,>      
>>      >      
I don't know if this is what your are looking 
for,>      but 
this>      >      
is what I>      
>      use for 
volatility:>      
>>      >      
pds=20;//Set your time period>      
>      Graph0 = 
StDev(log(C/Ref(C,-1)),pds)*sqrt(365)*100;>      
>>      >      
Anthony>      
>>      
>>      
>>      >      
dereklebrun wrote:>      
>>      >      
>  Hi is there any technical analysis way to 
measure>      
a>      >      
stock's price>      
>      > volatility in AB 
?>      >      
> If yes, how ?>      
>      >>      
>      > 
Thanks,>      
>      > 
Derek>      >      
>>      >      
>>      >      
> Your use of Yahoo! Groups is subject to the 
Yahoo!>      
Terms>      >      
of Service.>      
>>      
>>      >      
Your use of Yahoo! Groups is subject to the 
Yahoo!>      Terms 
of>      >      
Service.>      
>>      
>>      > Your use of Yahoo! Groupsis 
subject to the Yahoo! Terms>      of 
Service.>>>      Your use of 
Yahoo! Groups is subject to the Yahoo! Terms 
of>      
Service.>>>                    
Yahoo! Groups 
Sponsor                        
ADVERTISEMENT                         
[Click Here!]>> Your use of Yahoo! Groups is subject tothe 
Yahoo! Terms of Service.Your 
use of Yahoo! Groups is subject to the <A 
href="">Yahoo! Terms of Service.