Question Re: Automated Scalping

So the events of last Thursday convinced me that any automated scalping strategy would've gotten burned that day had it continued trading through the massive moves in the market. Are there any "circuit breaker" indicators that people use to determine that a major move is occuring and that scalping needs to be suspended (other than $$ losses, obviously)

Thanks for any insight.

 

I can't think of any indicators that are going to tell you fast enough (they'd be "slow-blow" breakers for sure unless you are using very small tick charts) Perhaps a direct monitoring of the bid/ask spread of a heavily traded instrument might be a decent indication. For example, the ES usually has a spread of 0.25 during normal hours. If you see higher than that repeatedly, it might be an indication things are getting too out of whack and you could at least lighten up your size.

 

That's an interesting idea. I wonder if that check can be automated?...

 

POSSIBLY something with fair value. It can get out of whack at times. But never studied with relation to what you are saying...

 

yes. just think what a smart human being would do if he saw the prices move as they did thursday; the computer does the same.

 

I think I will just go by overall losses for the day. Just stop trading for the remainder of the day and take off.

 

but wouldnt you quit before your loss limit if the market is "different"

 

Why would they get burned? Are you assuming all strategies are trying to revert?

Breakouts to the downside killed it that day.

But something that day was very different. My RTM strategies were getting way too many fills before it crashed so I shut it down and closed all my positions. I'm thankful I did or else it would have been extremely ugly.

 

I look at % deviation from MA, on a 5-min bar, as a really simple filter for "think twice" moments.

 

It would probably take some effort to determine what constitutes a "trip" condition. I would think the indicator would need to be rather short term to make it useful...maybe a running average over 20-50 ticks or something, since during normal operations, you shouldn't see too many outliers.
Fortunately, the market provides a pretty good test case once a month in the Non-Farm Payroll, so if your filter can identify that type of situation when that occurs, it might be able to help you out when the big mambo-jambo drops in.