[PRL] The Importance of Excel

[Paul A. Steckler]

Wouldn't a second pair of eyes have helped here?  VaR calculations use
a given quantile of risk and some starting sum, and I'd suppose
that experienced modelers would have a feel of how the numbers should
come out. Maybe a policy of "if your trade is over a certain amount,
you need to have someone check your work" would be useful. That would
be a software engineering practice, rather than PL technology.

For a VaR calculation, you could also compare it against historical
good data. Presumably, the calculation here would've been a too-small
outlier
when compared with similar past calculations and knowing that, a quant
would go back and re-check the model.

Maybe A-normal form for spreadsheets -- naming important quantities --
would help. So you'd have
a name for the average distinct from the name for the sum, and it
would be harder to mix them up.

-- Paul


On Thu, Feb 14, 2013 at 8:36 AM, Matthias Felleisen
<matthias at ccs.neu.edu> wrote:
>
> On Feb 14, 2013, at 11:26 AM, John Clements wrote:
>
>> requires time and development resources that people don't choose to allocate to the problems.
>
>
> Saving $500K on a high-quality programmer to turn this financial model into a properly designed program is far cheaper than losing $1B later on. Or as Realm of Racket puts it, a week of "programming" can easily save you an hour of thought.
> _______________________________________________
> PRL mailing list
> PRL at lists.ccs.neu.edu
> https://lists.ccs.neu.edu/bin/listinfo/prl