|
PureBytes Links
Trading Reference Links
|
I am glad that Mark Brown and Bob Fulks have contributed to this issue of
continuous contracts. Hopefully this debate can help the newbies with their
understanding of which type of data to use in backtesting.
Bob Fulks, as usual, you have posted a wonderful summary of continuous
contracts. All of Bob Fulks' posts are of high quality and add value to
both beginner and experienced traders. Thank you Bob Fulks.
I have the pleasure of dealing with the resident Omega-List dude, the
famous Mr Mark Brown.
Here it goes everyone!
Rob Bianchi originally wrote,
> IF I USE CONTINUOUS BACKADJUSTED DATA
> My systems DO NOT care if the T-Bond is trading at 135.00 or 50.00 or even
10.00
Mark Brown wrote,
1.) Many people say that the Perp data is not real and you cant trade it.
So now your telling me that buy changing the prices of a contract radically
like back adjusted does that's OK?
ROB BIANCHI RESPONDS:
Mark, as I stated in my previous email, my systems ONLY NEEDS RANGE AND
TRUERANGE to be totally accurate in order to trade the market, so YES, the
fact that backadjusted contracts get changed radically over time is of NO
CONCERN to MY SYSTEM. Perhaps it is a concern to YOUR SYSTEM but I can not
know this as I do not know what your trading system needs to calculate
accurate trade P/Ls.
Rob Bianchi originally wrote,
The only issue with this type of data is that the longer the back adjusted
data file is, there is a chance that the historical data will eventually
back adjust into negative (-ve) numbers. It is at this point that you must
either:-
Mark Brown wrote,
2.) Well at least you were man enough to admit that the butchering of the
data as in back adjusting can cause the price to go negative. So how does
your system handle that?
ROB BIANCHI RESPONDS,
Mark, just like you and as my name suggests, I am a homo sapien male or
"man" as you put it, or I think I was when I last checked. (relax Mark & go
have a surf, I do)
And YES, the O,H,L,C is butchered to ensure that the RANGE and TRUERANGE
are exactly the same as on that day how many years ago you go back.
On the topic of prices going negative, my system does not handle the
negative prices at all. I can not test through negative prices. That is
where I have to stop and rebuild another continuous contract to test past
that point. It's a little pain in the bottom but that's life.
Rob Bianchi originally wrote,
> (a) not back test beyond this point because negative numbers may have a
detrimental effect on the calculation of your trading system code or
Mark Brown wrote,
3.) Oh to deal with it you ignore reality, how nice. I guess you could
adjust so that your systems worst draw down in during that period. Then
when you sell your system you could just leave out that part.
ROB BIANCHI RESPONDS
Oh Mark, you need to buy a surfboard, get a longboard, they help you relax.
As I stated just above, (I sound like a broken record) you need to build a
second datafile if the data goes into negative territory. It's as simple as
that. Mark, I do not know how to respond to your comment on worst draw. I
do not have a problem calculating worst draw. I DO NOT use the TradeSation
worst draw anyway because it is wildly inaccurate. I use the daily
mark-to-market worst drawdown of the trading system. It's funny how the TS
worst draw is ALWAYS a lower amount than the mark-to-market worst draw.
It's TS code that I received from a great magazine called TS Express from
Bill Brower (dig! dig!). Mark, by the way, I do not sell systems or
anything like that. NEVER have and probably NEVER will, but you will won't
you with your TraderWare product ?
Rob Bianchi wrote,
> (b) construct another back adjusted file to back test that time period
which went negative (-ve). You would construct another data file with
different start and end dates to ensure that the historical data does not
contain negative values.
Mark Brown wrote,
4.) Well this defeats the purpose of a continual contract doesn't it?
ROB BIANCHI RESPONDS,
Not every continuous contract goes negative, but with the ones that do,
YES, I need to build 2 datafiles. Hey, that's life, that's the way it is.
Rob Bianchi wrote,
> IF I USE PERPETUAL CONTUNUOUS DATA
> Perpetual data weights or "massages" the highs and lows during the trading
days when rollovers take place. This effect causes the following which are
different to real-time trading:
Mark Brown wrote,
1.) You are wrong and like you say below each person should do their own
homework! You obviously understand back adjusted, but have no grasp of how
Perp data works.
ROB BIANCHI RESPONDS,
Mark, I have NEVER been able to generate real-life trading results using
perpetual data. The perpetual data versus the real contract signals was
always way-off for me. I have to admit I have written this part of the
email poorly. But Mark, thanks for clarifying perpetual contracts to me!
Mark the point I was trying to make is EVERY TRADER NEEDS TO GENERATE
ACCURATE TRADE P/Ls. PERPETUAL DATA NEVER DID THAT FOR MY SYSTEMS.
Mark Brown wrote,
a.) Perp data does no massage the highs and lows during the trading days
when rollovers take place as you state. It massage's the entire life of the
contract from day two. Giving every increasing weight to the next contract
out, making the current contract's influence diminish as time goes by. This
is reality as it happens in the real world.
ROB BIANCHI RESPONDS,
Mark, wow, I've NEVER seen a futures market that trades like that before.
How do I trade this kind of market ? Is this what REALLY happens in the
real world ? In the last 10 years that I've been trading it, I've NEVER
seen a market that looks like this before. On what exchange do these
"weighted markets" trade on ?
OK, enough joking around, seriously, the point is, both continuous
methodologies are NEVER perfect. Like I said, the choice of continuous
contracts DEPENDS on YOUR TRADING SYSTEM. Mark, obviously, perpetual data
works to generate accurate trade results on YOUR system and backadjusted
data DOES NOT. Mark, for me, perpetual data gives me wrong results while
backadjusted gives me accurate results.
Mark, you can't just "jam" perpetual data into people's faces. At the same
time, I CAN NOT "jam" backadjusted data into people's faces either. Like I
said, it DEPENDS ON WHAT YOUR TRADING SYSTEM NEEDS TO GENERATE ACCURATE
TRADE P/Ls.
Mark Brown wrote,
b. Back adjusted data actually steals the volatility that is undeserved from
the tail end of its contract. Then it pays the fiddler when it DROPS into
the next contract. This FALSE mountain that all you back test junkies LOVE
is good for you HYPOTHETICAL BACK TEST. That is precisely why so many want
to use it, its because your system would not make any money without that
false volatility bubble.
ROB BIANCHI RESPONDS,
Well Mark, what can I say. YOU ARE WRONG. I trade the fat tails on the
stable paretian distribution but most of those trades do not occur during
contract rollover. Sorry about that.
Mark Brown wrote,
c.) Like I've said before build a system that makes money on Perp data and
you have a system that shows performance that is worst than reality! - NOT
BETTER! I would much better be told the worst case of a system than the
best case. If a system makes money on Perp data then it most likely will
make more money in real life trading and it will be more dependable.
ROB BIANCHI RESPONDS,
Mark, like I said, the continuous DATA you use DEPENDS ON WHAT YOUR TRADING
SYSTEM NEEDS TO GENERATE ACCURATE TRADE P/Ls. Don't "jam" this into
people's heads. Mark, you are effectively "misleading" people into a false
sense of security by saying this. Mark, with all this "jamming" you do you
should open a confectionary shop.
Mark Brown wrote,
d.) To prove my point how many out there have bought or developed trading
systems on back adjusted data and then seen it fall apart when traded on the
real contract? Now you know the reason that most system sellers and traders
who are looking for capital use back adjusted data. It makes their system
look better than it really is.
ROB BIANCHI RESPONDS,
The above is a wild and sweeping statement. I know some successful CTAs
(one even manages US$300 million but that is peanuts compared to some of
Mark's buddies) that use continuous backadjusted data over the last 5
years. Why haven't their systems fallen apart ? Because of this, I would
have to disagree with you.
Rob Bianchi originally wrote,
> - ranges during this weighting or rollover period are usually different to
the front or the next futures contract. That is, perpetual data creates
ranges that are not "real-life" ranges.
Mark Brown wrote,
2.) Wrong the ranges more accurately reflect real life, there is a saying
what goes up must come down. The profits your system enjoys as it rockets
up on the expiring contract is brought back down on the ground when you roll
into the next contract out.
ROB BIANCHI RESPONDS
Mark, I do not see this in my trading at all and I trade alot of markets
and alot of round turns.
For everyone's interest I have attached an Excel spreadsheet. It contains
the following datasets of the S&P500 Index:
- 1 year perpetual contract
- 1 year backadjusted
- the REAL June 1999 contract from the backadjusted roll date of 15 March
1999.
The source is CSI Unfair Advantage.
For your reading pleasure, I have provided the ranges and trueranges of
each data type.
Columns A to E contains the perpetual contract data
Columns G to K contains the backadjusted contract data
Columns M & N contain the daily Range and Truerange of the perpetual data.
Columns P & Q contain the daily Range and Truerange of the backadjusted data.
Columns S & T contain the difference in the Range and Truerange of these
two continuous contracts. This shows how large the variation is between
these two types of contracts.
Columns V to Z contain the REAL June 1999 futures contract data.
Columns AB & AC contain the daily Range and Truerange of the real June 99
futures data.
BUT MOST IMPORTANT LOOK AT Columns AE & AF. IT SAYS IT ALL!
Columns AE & AF calculate the DIFFERENCE in Range between the real futures
contract in comparison to the perpetual and backadjusted contract.
Columns AH & AI calculate the DIFFERENCE in Truerange between the real
futures contract in comparison to the perpetual and backadjusted contract.
Columns AE to AI easily demonstrate that IN TERMS OF RANGE AND TRUERANGE
PERPETUAL DATA IS WAY OFF WHILE BACKADJUSTED IS EXACTLY THE SAME.
Once again, this tells ME that for MY SYSTEM, backadjusted works better FOR
ME than Perpetual data. In terms of calculating an accurate RANGE and
TRUERANGE, PERPETUAL DATA IS WAY OFF.
The MAIN POINT of this exercise is to demonstrate that because MY SYSTEM
relies on RANGE or TRUERANGE, Perpetual data would produce inaccurate
results for me.
Mark Brown wrote,
3.) We've been over this and though I haven't met the required 28 times to
post, so that you may learn it. I will continue to state that from a real
life point of view Perp data is as good as it gets. There simply is no
better way to blend a futures contract. I enjoy the solitude that
enlightenment gives me on this subject.
ROB BIANCHI RESPONDS
Mark, you should stop "jamming" and start a cookie factory. The continuous
DATA a trader should use DEPENDS ON WHAT YOUR TRADING SYSTEM NEEDS TO
GENERATE ACCURATE TRADE P/Ls.
Mark Brown wrote,
4.) Yes I can see where a look into reality would ruin most every systems
trader's day! So by all means stay away from Perp data and continue to
optimize away!
ROB BIANCHI RESPONDS
This is getting boring...... The waves are good over here in Oz! Oh
optimisation, I LUV IT, but we'll get to that some other time. It's a whole
week worth of emails, I'm sure, but I've got to get a few more surfs in
before I can make any comment on optimisations.
Mark Brown wrote,
<big snip lets get to the end you said it best> Mark
> Any trader worth their salt should perform this exercise prior to risking
money in the market or else GARBAGE IN = GARBAGE OUT
ROB BIANCHI RESPONDS
I am sure that I will read one of Mark's "club to death that
Omega-contributor" email on tomorrow's Omega-List-Digest. Mark, you are a
great contributor to this List but geez, relax pal! Have a surf. (I'm only
joking around Mark!)
/*/*/*/*/*/*/*/*/*/*/*/*/*/*
Robert J Bianchi - Director
Venitia Pty Ltd
Email : r.bianchi@xxxxxxxxx
Phone : 61-7-3899 9998
Fax : 61-7-3899 8605
/*/*/*/*/*/*/*/*/*/*/*/*/*/*
|