Searching for micro alpha

It's much more simpler than what your looking for. The only issue is the infrastructure aspect of it and hiring people to monitor the proper functioning of the system.

 

http://www.screencast.com/t/wttnmcfoUE

This has been on the back burner for awhile.

 

BTW, why do you consider 53% to be bad? Is this some kind of r/r of 1:1?

 

Disregarding execution the prediction is good for about 2/3 an increment. Recall, though that this valid for a decent portion of the day. Since my goal is to do 500+ trades a day per contract I haven't narrowed way down to the very low frequency stuff that I could probably predict better.

2/3 an increment on average would be amazing if execution could allow it. Taking into account all the execution issues I could maybe make money with this but just barely.

 

Nadex is a joke. They probably do less than $500K notional, daily. Digital is simply a discrete vertical, and now is that going to help a guy trading in 5-minute holds?

 

Sent PM,

It would be interesting for others to know this:

- is this a collaboration, where profits are split according to predetermined ratio?..

- your team will code and implement and monitor the system...

- legal paperwork would need to be signed to protect both parties..

- a simpler way to do it is form two member LLC, split ownership.

Chris

 

Because you can't trade 5 minute expirations on any other regulated exchange in the U.S. It's once a week at best. You're right about the notional and they certainly leave a lot to be desired, but they do have market makers that quote a continuous market that's at least in the tens of thousands in size, and I haven't seen them widen spreads even when I hit them repeatedly for their full offer size. The worst they do is wait 3 or 4 minutes to reload that side of the market. I really do wish they had more volume, and they generally run a shoddy operation with buggy execution software that ironically can't handle high volume or big movement days. However they're all you've got if you want short term expirations.

 

I haven't looked at them in years. I wasn't aware that they had intraday digitals.

 

If we presume that Adaptive Market Hypothesis applies to high frequency then what about a regime switching model based on a dislocation of risk to predict better?

A standard deviation type variable on your principal components? Model the changes in dislocation, capture some order of magnitude in a time frame of your choosing and build a signal?

M

 

I don't think PCA applies well in this context. Bag of heuristics tends to be better than model based on very short time frames.