[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: The peculiar System



PureBytes Links

Trading Reference Links

--- In amibroker@xxxx, Yuki Taga <yukitaga@xxxx> wrote:
> Hi Tomasz & Trader,
> 
> Sunday, October 14, 2001, 5:14:23 AM, you wrote:
> 
> TJ> Trader already wrote it:
> 
> TJ> "When working with end of day data it makes no sense to even
> TJ> consider trading on todays close. If you want any approximation
> TJ> to reality you need to change to: buy tomorrows open (or later)
> TJ> for stocks and tomorrows close for mutual funds."
> 
> I have no way to refute this, being first and foremost a trader and
> not a very good system creator, coder, optimizer, etc. (^_-) But I
> am curious why such a blanket admonition would be accurate. It 
would
> seem to me than in real life if you try to trade the close, you will
> sometimes do better and sometimes do worse than the close. Here in
> Tokyo, for example, you cannot specify market on close, you simply
> have to try to get included in the batch of electronically matched
> orders that will determine the final fill (a market order submitted
> about 20 seconds before the close often works, but not always). In
> any case, if you are trying to trade a signal that is generated by
> EOD data, and you run your data 10 or 15 minutes before the close,
> you are going to get signals, and most of them, in my experience,
> will turn out to be signals that will persist even after the final
> close is registered and the data re-run. So, if you try and trade,
> you'll get better and worse fills than the close, but over any long
> term period of time you should just about average the close. So why
> would specifying a buy on the close result in back tests far out of
> the realm of reality?

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Hi Yuki, I knew some sharp trader would bring up this possibility. :-)
Of course if you download current prices 15 minutes before the close 
it is not really EOD data is it? :-)
There is nothing wrong with your comments because you run the trading 
system on currently available data and then initiate an order "after" 
that data was evaluated. This is a pseudo intraday technique, works 
for me... The problem is that it is difficult to test this technique 
with an EOD trading system for the reasons you mentioned.

The point I was making about EOD data and trading on the close is 
oriented toward the testing process itself. Testing results will 
usually indicate a higher profit if you "look back at historical 
data", which is what is happening if after getting a final close 
price you tell the trading system to take that price even though that 
trade cannot be done.

It all boils down to an attempt to get the backtesting to represent 
what you will be doing in your real world trading. If you get your 
data after the close, then trading the following open( or something 
that simulates the open these days ) just makes good sense.

In the situation you mentioned, taking todays close may be the right 
approach, but only after the tester thinks about the appropriateness 
of it instead of blindly using the close all the time.

Actual results are almost always worse than the testing anyway, for 
other reasons which I am sure you are aware. Price slippage, 
inability to get the order executed in a timely manner, 
communications problems...you name it....

BTW I use a intraday system to evaluate the "15 minutes before the 
close" technique...

Cheers
Trader