|
PureBytes Links
Trading Reference Links
|
Tom wrote:
> Hi Mark,
> I have always tried to roll my EOD contracts on the first notice day,
> along with my real positions. Another possible roll point would be by volume
> exceeding the current contract for the next one out.
I have several hourly systems, I plan to created and convert all my systems tothis time frame (or a representative tick chart size, example 300 tick
bar chart or
whatever for each commodity) then I can just check all positions at once. This
will free me from being glued to the screen all day waiting for a trade to pop up.
I have done extensive research and development on complex projectionary
systems to know in advance what and when a trade will be. However I think
it is a simpler method to restrict the data sampling to a consistent level. Back to
the roll issue: Having a system for example that trades and average of one trade
per week on an hourly data set. I would roll the trade into the next month when
my system reverses its position near the rollover date. Preferring to roll maybe
even earlier than the standard contract roll, but certainly prior to the roll date.
I think this is ideal for a medium and short term trader. Longer term trades
could benefit also if you know the length of a typical trade and just plan ahead.
> Most of the time these coincide within one day. Since volume for
> daily is reported one day back, it cannot be determined exactly.
I would not roll on volume, See a perpetual contract like I'm using weightseach day one contract into the other. For each day that the current
contract
trades it diminishes in value and the next contract out increases in value. As
far as our synthetic contract data is concerned. I will post a chart showing
the roll effects of this method . They say a picture is worth a thousand words
but I have experienced a picture can generate a thousand questions.
> When making continuous contracts for systems, I'd assume you don't
> want the gaps, but if you were to roll a real position, you do have those
> gaps as the real world. Would it not be better to build systems to account
> for these roll periods, exclude them from backtesting or have the system
> take them into the trading strategy. This would mean exit and re-entry but
> counting as a single continuous trade if that were the system's position at
> the time.
Not if you roll out of the existing contract at the time the system flips from longto short as I have proposed above. At least on a hourly basis it is
canceled out
because sometimes you will gain a few ticks and some times you will loose a few
ticks. But the method is far better than anything I have ever experienced! I think
slippage from a good to a bad broker would make a larger impact on the systems
real world performance.
> How do you get continuous contracts into tradestation so they represent the
> current contract and are updated? Can you convert to omz from other formats?
We are currently using some advanced undocumented features of TS and a DLL.This allows the system to use the perpetual real time data that we are
building as
the price input for the real time system. Now the buy and sell arrows will plot on
the current contract data. However the position of the buy and sell arrows are
completely governed by the real time perpetual data stream. We are doing this
within TS.
> The other possible reason for long contract data is viewing the success or
> failure of indicators or cycle strategies. I understand that some of the
> cycle software like Bressert's, requires lots of history.
All empirical observations of indicators and cycles are false, I don't believe in either.I do believe a system that is tested on historical data. If
in fact that same data is the
data that is running real time. IMPORTANT: I need to make a point! What I am
doing with the data is different than what is commonly preconceived by the masses.
I have a system and data that when I look back through the years shows me exactly
where it bought and sold "price and time". I can check my trading account statements
and I can compare exactly that trade on the historical data with my account statements.
They match! They are the exactly the same! That historical data is not only historical
data but it is Real Time data running outside of TS as a huge 10 years of hourly
perpetual data. It is then called up and used as the price input of my systems, running
real time. In concept if the TS 5.0 is capable you could have a real time historical data
stream that you could scroll back 20 years and observe the trades it made all real time,
while it is running real time. I know of no other method that is as exact for mechanical
system trading as this is. If there is one I'll buy it or pay to have it developed. It all can
be thought of as a slightly synthetic futures contract that never ends, much like a cash
contract. If I have failed to make myself clear in any manner please ask further questions.
Mark Brown
> See U
> Tom
> planeacres@xxxxxxxxxxxxxxxx
>
> -----Original Message-----
> From: MarkBrown@xxxxxxxxxxxxx
> Subject: Re: negatives values in TS
>
> I have even been testing a perpetual real time add on
> >for TS that Rich Estrem and Myself have developed. It allows for the
> >automatic roll over of futures contracts. The catch is that while we are
> >able to produce this contact in real time forward.
> >Now why would someone want to do this?
> >Because: This allows you to (with great confidence)
> >back test a system on the very same data that you trade it on. In my
> >mind there is absolutely no substitute for this method of system trading.
> >It is very time consuming and expensive to do but so is faulty testing of
> >a system when applied to real money. To answer some inquires ahead
> >of time, if TS 5.0 ever comes out and we are satisfied with the security
> >features of it. The real time perpetual futures contract builder will be
> >offered to the public.
> >
> >Mark Brown
> >
|