Trading System Statistics

[forum]

[message 501 of Yahoo group]

A simple comparison of old (oops I mean "classic") vs
mainstream statistic software in time series analysis,

testing type classic trading apps statistics software
out of sample no yes
walk-forward no yes
time restriction no yes
time segmented no yes

to summarize it all - classic trading apps are trying to test
what is being asked for by the uesrs of that generation ...
technical analysis was new, no indepth ties to statistics,
thus all the information that the users are looking for
at that time was basic entry level statistics like
"winners%",
"win/loss ratio",
"average winner to loser value", etc.

these stats are good for single entry, single bet size,
all out exit type of strategy - which is very basic and
only applicable for very simple scenerio.

what about multiple entry scaling type of trading,
well, these classic apps' so call system testing immediately
break down and generate useless information.

then we are talking about multiple entry, multiple securities,
multiple positions at the same time ... which is what a
normal active trader do ... no matter future trader, or
stock trader.

The key is all these is to bypass the basic concept of
looking at single in/out bet results but focus on the equity
changes ... then the complex scenerio of analyzing the
trading model can be simplified into managable statistics
studies.

Now, all mainstream stats stuff kicks in and become very useful
in analyzing the stability of the model. The more stable of
the equity curve in terms of growth the better the model is
when the average return per time period is the same.

Thus, we expect to make NeoTicker being able to not just
optimize, but truly collect stats on all aspect of the trading
model based on both predefined data and user defined stats.
then the user can easily graph the outcomes and the related
stats in anyway they wanted.

Just thinking of it - I would like to use it myself first hand now

Lawrence


--- In neoticker@y..., Michael Mueller <new_systematic@y...> wrote:
> It seems as if we had the same ideas, since I just implemented a
kind of time range table for specific futures contracts, which is a)
good for testing many different contracts and b) for advanced testing
methologies such as walk-forward-testing.For testing of different
contracts using MULTIPLE csv data files in one scan I wonder if I can
set up a symbol list as you suggested ?(what would the format be ?
c:\spu1_M15, c:\spu1_M60) or is it only possible for symbols of the
realtime feed ?Your ideas for system evelopment and testing in the
forthcoming releases look very promising to me !! Happy New Year !
Michael

[NeoProg]

So, now that the Grid Optimizer is out, and NeoTicker has reached V4.10 Build 47 (as of this writing), are the ideas discussed here reflected in the latest software?