|
PureBytes Links
Trading Reference Links
|
>>Why would one want to be in a long position if you would NOT buy at this
price level. The above code leaves you in long or short position although
the condition for purchase or sell is not satisfied anymore.<<
Probably because the cost of entering a trade is greater than the cost of
being in a trade. Your arguement is logical when entering and exiting has
no cost (zero slippage and commission). Under that condition, one could
enter and exit hundreds of times each day, picking only the most promising
moments. However, in practice, S&C would eat you alive, so there is a
trade-off between profits and costs, causing the optimal solution to have
fewer entries and exits.
It may be that a slight increase in exits would be more profitable than the
drastic one proposed. In this respect, fuzzy logic (FL) is a powerful and
extremely simple way to attain just the right level of discrimination. A
trivial, yet sometimes effective FL technique is to have several
indicators, each scaled to a range of 0 to 1, and then add them all up.
Now you compare the sum to a threshold. This way, a strong indication by
one could compensate for a mild indication by another. Now, you can
control the amount of exits by raising or lowering the threshold.
Somewhere there will be an optimal threshold level, or two. :-)
- Mark Jurik
-----Original Message-----
From: SKane777@xxxxxxx [SMTP:SKane777@xxxxxxx]
Sent: Thursday, November 05, 1998 8:44 AM
To: Orphelin@xxxxxxx; Omega-list@xxxxxxxxxx
Subject: Re: Construction of trading signals
We use an intraday system (2 min bars) on TS, developed 4 years ago and
traded realtime for the past two years. A friend of mine who is the one of
the head programmers for one of the large computer firms here in Austin
programmed the system for me. He did not like the TS 13,000 bar limit, so
he
exported all the data (3 data inputs) for 2 years into an ascii file and
re-
programmed the TS system (functions, indicators and system) into C
language.
It runs on DOS. In 2 years, out of about 400 trades, there have only been
4
discrepancies due to rounding problems between C and TS. Everything
matches
up fine when I backtest over 6 years of data.
We have never "lost" a trade nor "gained" a trade.
One note - if your time feeds are not in gear, you may have realtime and
backtest variance. We experienced this with FutureSource last Monday when
we
did not adjust the time offset for daylight savings (our fault).
Steve
|