Services & Resources / Wolfram Forums / MathGroup Archive
-----

MathGroup Archive 2011

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: FinancialData errors

  • To: mathgroup at smc.vnet.net
  • Subject: [mg120604] Re: FinancialData errors
  • From: DrMajorBob <btreat1 at austin.rr.com>
  • Date: Sat, 30 Jul 2011 06:03:36 -0400 (EDT)
  • Delivered-to: l-mathgroup@mail-archive0.wolfram.com
  • References: <201107222344.TAA28525@smc.vnet.net> <j0fos4$8p6$1@smc.vnet.net>
  • Reply-to: drmajorbob at yahoo.com

> Prices do vary continuously, they don't stay the same for a whole day,
> At the moment you bought it was 10.36.  At the moment the data source
> was sampled it was 10.23. Usually historical data use the closing
> price, AFAIK. I see nothing strange in this.

Nonsense.

It's a mutual fund, as you can see from its name. It has one price,  
exactly once a day. I can look up that price at Yahoo! Finance, and it  
agrees with what the fund company charged me on that day. Check Feb 10th  
on the following page:

http://finance.yahoo.com/q/hp?s=JANFX&d=6&e=29&f11&g=d&a=1&b=16&c10&z=66&y=66

where open, high, low, and close are all the same -- namely, $10.36.

I JUST noticed the last column on that page says the "adjusted" close was  
$10.23, and that's what FinancialData returned -- part of a possible  
explanation -- but:

a) There was no dividend or split on that day or any day nearby,
b) NONE of the numbers in that column make any sense,
c) For most days and/or most funds, FinancialData returns the unadjusted  
close, so there's no consistency, and
d) I want the price I paid, not an "adjusted" number, even if it made  
sense.

> If a difference of less than 2% in the historical market data
> influences your algorithms, I won't trust your conclusions.  :)

I want to input the dollars I contributed and compute the shares I bought,  
to the same 3 to 4 digits that the fund companies compute it, in order to  
simplify "buy" and "sell" functions -- to save me time when entering  
trades. 2% is far too much error to do that.

Ending with the wrong number of shares in my account -- because of days  
when the price comes back wrong from FinancialData -- is unacceptable. It  
makes the share and dollar balance wrong, the internal rate of return  
wrong, etc.

I'm contributing $100 to my IRA 60 times a year, plus occasional transfers  
among the funds I hold, so simpler "buy" and "sell" functions seem  
worthwhile, if it were possible... which, with these errors, it is not.

And regardless, whether I need perfection or not, the "curated" data is  
simply wrong.

Bobby

On Fri, 29 Jul 2011 03:48:42 -0500, Pierpaolo Bernardi  
<olopierpa at gmail.com> wrote:

> On Fri, Jul 29, 2011 at 10:41, DrMajorBob <btreat1 at austin.rr.com> wrote:
>> FUNNY STORY: Here is Tech Support's non-response to my problem,  
>> described
>> farther down:
>
>> I've found a new (to me) species of error in FinancialData.
>>
>> One example is the price of the following stock on Feb 10, 2011:
>>
>> s = "JANFX";
>> FinancialData["JANFX", "Name"]
>> FinancialData["JANFX", {{2011, 2, 9}, {2011, 2, 11}}]
>>
>> "Janus Flexible Bond Fund D Shar"
>>
>> {{{2011, 2, 9}, 10.25}, {{2011, 2, 10}, 10.23}, {{2011, 2, 11},
>>     10.24}}
>>
>> As you see, FinancialData says the price was 10.23.
>>
>> But I BOUGHT the stock through Janus Funds on that day, and the price  
>> was
>> 10.36.
>>
>> Moreover, if I go to Yahoo's interactive price chart for the stock at  
>> the
>> following page, 10.36 is the price I find:
>
> Prices do vary continuously, they don't stay the same for a whole day,
> At the moment you bought it was 10.36.  At the moment the data source
> was sampled it was 10.23. Usually historical data use the closing
> price, AFAIK. I see nothing strange in this.
>
>> So... where are the errors really coming from?
>>
>> This is not an astronomically large price or a Missing["Not Available"]
>> price; it's within 2% of the correct price. But it's wrong.
>>
>> These errors make it very difficult to trust FinancialData for any  
>> useful
>> purpose.
>
> If a difference of less than 2% in the historical market data
> influences your algorithms, I won't trust your conclusions.  :)
>
> Cheers


-- 
DrMajorBob at yahoo.com




  • Prev by Date: Re: Preventing In-line Math Typesetting From Being Scaled Down in Text
  • Next by Date: Re: Memory leak or flawed garbage collector
  • Previous by thread: Re: FinancialData errors
  • Next by thread: Re: FinancialData errors