|
PureBytes Links
Trading Reference Links
|
> -----Message d'origine-----
> De : Bengtsson, Mats [mailto:mats.bengtsson@xxxxxxxx]
> Envoyé : samedi 28 juillet 2001 21:47
> À : pierre.orphelin@xxxxxxxxxxxxxx; omega-list@xxxxxxxxxx
> Objet : RE: TradeStation Precision - Summary (Pierre considering Bob not
> being objective and not listening)
>
>
> We both speak of facts, the same facts, but we stress different parts of
> them since we have different starting points.
>
> My starting point is that if we can keep a good clean agreement on what
> should be improved in Tradestation, and make Omega and resellers
> understand> that there is a need a lot of people agree on, then we can do
ourselves a
> lot of good. At the same time, let us help less experienced users to
> understand that these problems really affect them even in standard
> indicators, by discussing them and how to avoid them (like Bob does, he
> gives good advise). Your point is to me more of the theory
> lesson, it might> help some people to understand what they need to do to
avoid
> problems but it seems to me to have a lot of possibilities to instead make
them think they
> can be ignored, and a lot of possibilities to make Omega think they can
> ignore them.
Theory is necessary before doing any mandatory modification.
This is the norm in any industry, politics excepted.
If they are needed , I will be the first to tell it and this has nothing to
do with my parallel reseller status.
At this time, and for what I read , it's not exactly the case...
>
> If I did not read your mail incorrectly, you did state that the speed
> limitation is a good reason not to go to double precision, specially
> considering that precision is enough anyhow. I do disagree on
> both of those
> parts, both speed limitation being a good enough reason nowadays, and
> neglecting the extra error introduced by staying at floating point
> calculations.
>
The float precision is enough because we deal with 5 digits price .
Doulbel precision will not bring anything of interest excepted waste of
computer power.
I modern computers are faster, this is not a reason why to be less rigorous.
Also, this apply to all applications.
Why do you think that there are floats doubles in C yet today ?
Only to make programmer mmoer intelligent?
What will be the speed of a TS version optimizer for example when you use
doubles ?
Do you think that the current float version is still fast enough ?
> However, what I want to speak most of is the same as before, if we keep
> stressing we want Omega to do better with their handling of data results
> caused by precision problems, we might get them to do that. If every time
> somebody shows an example of precision problems not being dealt with by
> Omega, we go into long discussions if this really has an outcome
> of trading
> profits, we will cloud the message we otherwise would benefit to get
> through.
>
> I see Omegas neglect of handling precision problems in their code
> being the
> same thing as neglecting to handle division by zero. They do not
> think it is
> important, they do not get the impression the market considers it
> important,
> and they will only do what they make profit from. If we keep
> saying to them
> it is important, they may find a reason to rethink. If we instead
> say it is
> good enough, they will not waste their money and efforts on handling it.
> Then we will have to rewrite all the code ourselves to get it to
> do what it
> should in the first place.
>
>
I do not believe that TRAD will chage its mind from the advice of a few
Omega List user that are not very well informed, what they can understand
from this thread.
The question of precision has been solved since the beginning and they would
have used double if necessary at this time, but they already knew that it
was not useful.
The division by 0 error check removed in TS Pro is probably not a good idea,
IMHO
Sincerely,
Pierre Orphelin
www.sirtrade.com
TradeStation Technologies representative in France
|