Are my eyes deceiving me?!!!

C

catmango

First of all, thanks to all who've been answering my questions so far.

I'm backtesting a system now (in Excel) that scalps the NQ based on price movement. The system generates roughly 10 trades/day, with entry/exit prices based on the 5-minute bar closing prices. When I assume only commissions, the system is making money fairly consistently and with little drawdown. However, when I assume that I underperform by one tick on each side of the trade (two full ticks total for each round trip), my profit turns into a HUGE loss.

Am I overestimating the error factor here? Should I assume a full one-tick slippage, even if I only plan to trade 2-4 contracts? Or, if the opening price of the following bar is in my favor, can I assume that my order got filled successfully at the close of the previous bar? Would it be better to scalp the ES instead, since it's more liquid and has a smaller tick size?

Thanks again in advance for your help!
 
Quote from rolegario:

First of all, thanks to all who've been answering my questions so far.

I'm backtesting a system now (in Excel) that scalps the NQ based on price movement. The system generates roughly 10 trades/day, with entry/exit prices based on the 5-minute bar closing prices. When I assume only commissions, the system is making money fairly consistently and with little drawdown. However, when I assume that I underperform by one tick on each side of the trade (two full ticks total for each round trip), my profit turns into a HUGE loss.

Am I overestimating the error factor here? Should I assume a full one-tick slippage, even if I only plan to trade 2-4 contracts? Or, if the opening price of the following bar is in my favor, can I assume that my order got filled successfully at the close of the previous bar? Would it be better to scalp the ES instead, since it's more liquid and has a smaller tick size?

Thanks again in advance for your help!
More...
You have hit on THE nerve center of testing systems that scalp. When you go live, the situation will be EVEN WORSE than you assume here. The key is that you MUST do your testing with B/A data, not just closing prices. AFAIK, no software in existence is this smart because they do not have that kind of granularity. Finally, the faster the market that you are testing moves, the worse accurate testing becomes, even with B/A data. Think for a moment, are you going to be quick enought to get the price being bid/offered? No? Back to 1/4 slippage getting in, a 1/4 slippage getting out.

nitro
 
if your system breaks down with minimal slippage assumptions during testing, you are asking for trouble.

Seek to disprove.
 
If you apply your system on a longer time scale (i.e. not limited to day trading) you won't have those problems.

I have found that I can get the NQ filled at the opening price of a 5 minute bar almost every single time, with a market order.
I watch the clock... 5.. 4.. 3.. 2.. 1.. click. So if your system trades only at the opening of the 5-m bar you can count on getting the the fills.

Of course, nothing that trade 10 times a day will work on 5 minute bar trading times like this... you need something longer term, such that 5 minutes is only a blink of an eye.
 
Quote from nitro:


You have hit on THE nerve center of testing systems that scalp. When you go live, the situation will be EVEN WORSE than you assume here. The key is that you MUST do your testing with B/A data, not just closing prices. AFAIK, no software in existence is this smart because they do not have that kind of granularity. Finally, the faster the market that you are testing moves, the worse accurate testing becomes, even with B/A data. Think for a moment, are you going to be quick enought to get the price being bid/offered? No? Back to 1/4 slippage getting in, a 1/4 slippage getting out.

nitro
More...


I think smartquant can do this. I've heard rumours that the license is in the 2-3k range. Check the smartquant group on yahoo for more.

http://www.smartquant.com

K can do it too.. but good luck being able to afford it.. I've heard 200k for developer license (but that may have been a WAG).

http://www.kx.com

Don't forget how much the data will cost.. TAQ is not cheap.


--laz
 
Quote from rolegario:

First of all, thanks to all who've been answering my questions so far.

I'm backtesting a system now (in Excel) that scalps the NQ based on price movement. The system generates roughly 10 trades/day, with entry/exit prices based on the 5-minute bar closing prices. When I assume only commissions, the system is making money fairly consistently and with little drawdown. However, when I assume that I underperform by one tick on each side of the trade (two full ticks total for each round trip), my profit turns into a HUGE loss.

Am I overestimating the error factor here? Should I assume a full one-tick slippage, even if I only plan to trade 2-4 contracts? Or, if the opening price of the following bar is in my favor, can I assume that my order got filled successfully at the close of the previous bar? Would it be better to scalp the ES instead, since it's more liquid and has a smaller tick size?


suk my ass
Thanks again in advance for your help!
More...
:)
 
Nitro is right on.

Your playing in the "grey" area, where the only way to
know if you will get your fill on the entry AND exit, is to
trade it live.

You already know you will LOSE if you dont get the
rights fills. That is certain. Determine what percentage
of the time you MUST get filled to at least break even,
and then test this number live with 10 trades to start,
then 20, 30, etc, assuming the numbers arent horrible.

I've written a ton of trading systems that would be
incredibly profitable if I could be guaranteed the fill.

This shows you how efficient the markets really can be.
Your situation is more common than you know.

peace

axeman
 
Quote from peterfigliozzi:

If you apply your system on a longer time scale (i.e. not limited to day trading) you won't have those problems.

I have found that I can get the NQ filled at the opening price of a 5 minute bar almost every single time, with a market order.
I watch the clock... 5.. 4.. 3.. 2.. 1.. click. So if your system trades only at the opening of the 5-m bar you can count on getting the the fills.

Of course, nothing that trade 10 times a day will work on 5 minute bar trading times like this... you need something longer term, such that 5 minutes is only a blink of an eye.
More...

Thanks for your input. Interesting approach, targeting the open rather than the close. I was wondering though, when you put in an order for the open, is it a market order?
 
Quote from nitro:


You have hit on THE nerve center of testing systems that scalp. When you go live, the situation will be EVEN WORSE than you assume here. The key is that you MUST do your testing with B/A data, not just closing prices. AFAIK, no software in existence is this smart because they do not have that kind of granularity. Finally, the faster the market that you are testing moves, the worse accurate testing becomes, even with B/A data. Think for a moment, are you going to be quick enought to get the price being bid/offered? No? Back to 1/4 slippage getting in, a 1/4 slippage getting out.

nitro
More...


Precisely... a very strong argument against a totally systemized approach to scalping....
 
You must log in or register to reply here.
Back
Top