Closing out at end of day

[stephencrowley]

Are there any general methods for mechanical/automated strategies to know when to stop entering new trades? That part should be easy... something simple like don't enter new positions 30 minutes before close.

But closing out existing positions at the optimial time before close seems tricky.

 

[Kicking]

what do you mean? are you trying to automate exits at the highs (if long) lows (if short) of the last hour ?

 

[stephencrowley]

I have an automated strategy that makes trades continiously throughout the day. I have a rule that says not to enter any new positions 45 minutes (or whatever) before close.

But for existing open positions I don't want to close out immediately because that would be silly, but I also can't wait til the markets are closed either. I need to create some sort of system to average in model signals with the time-til-market-close.

Was wondering if anyone has addressed this issue or know of any research done on the subject?

Good luck,
Stephen

Quote from Kicking:

what do you mean? are you trying to automate exits at the highs (if long) lows (if short) of the last hour ?

 

[giggollo]

I guess the only way to know would be to backtest inhibiting your system x minutes before the close and optimize x. You would need intraday data for this. Naturally, as you get closer and closer to the close, it becomes less and less worth it to enter the trade because a stock can only move so much in a v short period of time.

 

[stephencrowley]

Exactly. Are there any existing algorithms or papers published? I'd hate to reinvent the wheel.

Quote from giggollo:

I guess the only way to know would be to backtest inhibiting your system x minutes before the close and optimize x. You would need intraday data for this. Naturally, as you get closer and closer to the close, it becomes less and less worth it to enter the trade because a stock can only move so much in a v short period of time.

 

[giggollo]

Quote from stephencrowley:

Exactly. Are there any existing algorithms or papers published? I'd hate to reinvent the wheel.

I think you would want to backtest it for your particular system, assuming that your system can be backtested in the first place. The results may vary depending on what the strategy is, so u want to look at how exiting a little earlier or inhibiting trades a bit before the close would impact your particular system. I don't know how useful a generic result published in some paper would be.