patrick.net

ATR as in average true range, right?

Graybox says

Peter have you ever programed a system or tool for your investing or trading?

Yes. Coding the system is much less challenging than psychology issues though.

Are you playing with Neural Networks or AI-based systems?

Graybox says

I hear that and 1 of the hardest issues is admitting when your wrong. Even when you are right in the end the market can out last you especially since they have some ability to hedge against you in the non-sweet spot hours.

I know a trader who built a black box sys. to trade S&P minis. He was walking proud until the market changed and made his sys irrelevant. It almost appeared his marriage failed when his sys. failed. I don't even know if his sys. lasted a year.

Very true. And there are really two things:

1. Admitting failure when the trade goes bad
2. Admitting failure when the system goes bad

(1) afflicts mostly discretionary traders (or undisciplined system traders). The system is supposed to tell you when a trade is not going to work out.

(2) is more tricky. When should you pull the plug if you are in a drawdown? Is the market temporarily out-of-whack, or has there been a fundamental change?

Graybox says

I suspect it is times like this that spins the algorithms out of sort and circumstance that challenges neural and AI systems.

LOL, most current neural and AI systems are already very good at finding challenges to trip themselves over. :-)

They are just not very good regression systems. Perhaps newer techniques will allow them to learn nuances without remembering every noise as fact.

Graybox says

Do you implement word recognition from a news source or sources as a filter of change detection? News kills many accts.

No. It is hard for a smaller trader to compete on speed and infrastructure.