Do you need to understand where the edge comes from?

[bashatrader]

Most edges derived via ML are totally abstract models.

 

[tommcginnis]

You trash random walk,...

Not me, bro, in the least. But most people conflate random walk of returns with random walk of prices, and then conflate a problematic random walk of prices with an impossible declaration of The Markets Are Random. Whoa! Is that you? I don't know, but you've brought this up out of thin air, so I won't draw any conclusions.

It's an anonymous public forum...

"Excuse me?" This is an outlook to which you subscribe. To paint your motive onto others is what our head-shrinking friends call Projection. There are many here on ET who utilize an avatar; it can be lots of fun; but there are some who hide behind it, to engage in gutless and grotesque douchebaggery. Don't join them.

...I can't find a TA method that is mathematically proven to be positive expectancy, and thus programmable into an algorithm.

Not my fault. In fact, through the course of my posts in the last quarter, I have posted two. Rudimentary and unremarkable (to the rest of the trading world) -- they nonetheless survive everything I can throw at them. I had started with an eyeball backtest, and was mapping out a Monte Carlo spreadsheet suite, when I came upon Tradingview, which I will recommend to you.

I wish you a measure of long-term success.

 

[tommcginnis]

A reminder: Kevin Davey is an ET sponsor, has been around the block, and although I have not had need to investigate deeply, others apparently have. (To wit: tradingschools.org )

Between Mr. Davey and the free-to-use abilities of
https://www.tradingview.com/ , no one should labor under any misapprehension that T/A-driven algos are magic, are beyond their capability, or cannot be simply formed to make money.

Bring the work, bring the discipline, you won't be disappointed.

AND TO RETURN TO THE OP, "Yes, you should always know your edge."

 

[ironchef]

It's wrong, because DSP was developed to work on physical phenomena, such as waves from an earthquake, thermal fluctuations of electrons in a resistor, etc. Market data is generated by a completely different mechanism that only resembles waves. Markets are made of people and are sentient, earthquakes are not. However, all market models are wrong, and DSP may be no more wrong than any other model.

To address tommmcginnis:
Negative group delay is purely a mathematical construct and has no predictive power and surely is not the result of economic decisions of traders. Although one could argue the base indicator is a net result of fundamental trading decisions. What math am I supposedly ignoring?

There seems to be logics even in the madness of human nature/psychology? Granted, it is like trying to look for a small perturbation in a very noisy system, like figuring out how a herd moves?