Trading Catechism

[Gambit]

Good hypotheses. How do you go about testing it?

I'm relying on what I learned from Ernie Chan's books. I would test for autocorrelation on time series data from different time frames. That way I can confirm/disprove what I see on the charts. I could also calculate a PDF using numpy or pandas. From stackoverflow, this seems straightforward. http://docs.scipy.org/doc/numpy/reference/generated/numpy.histogram.html
Put the data into an array and run the hist function?

 

[nitro]

I should be clear. There are definitely very simple methods that work. For example, S/R when coupled with overzealous traders playing momentum against S/R, and a relatively wide bid ask spread. When wrong, get out fast. Go flat EOD. The research is where to take profits. 1:3 RR ? 2:10? etc?

The problem with those things is that they require infinite patience sitting on your hands waiting for opportunities, and the ulcer index is high. That is not trading, it is at best swing trading, with an almost minimum wage return for time invested. It will make money though, probably 5-10% with no leverage, 50% to 100% with leveraged instruments (this is hardly minimum wage, but the risk is very high because of an event wiping you out since it is directional and highly leveraged).

My goal is not to be a slave to my computer and the market. My goal is to have the computer work for me, with little or no risk intraday.

 

[nitro]

I'm relying on what I learned from Ernie Chan's books. I would test for autocorrelation on time series data from different time frames. That way I can confirm/disprove what I see on the charts. I could also calculate a PDF using numpy or pandas. From stackoverflow, this seems straightforward. http://docs.scipy.org/doc/numpy/reference/generated/numpy.histogram.html
Put the data into an array and run the hist function?

And if you trade that methodology on say all the symbols on the S&P500, is there a statistical edge over say five years? What about a subset of the S&P500? which one?

 

[Gambit]

> normally distributed if you are computing a PDF

Normal is only one of many PDFs. But more generally, making money in options has next to nothing to do with learning textbook probability theory.

I don't know enough to disagree or agree. But isn't finding a stationary time series and estimating its PDF a decent starting point? Consider a retail/semi pro's available tools and budget. He can't be faster, he can't buy order flow, he can't even quote a two sided market. What is left other than taking a punt on vol and direction?

 

[Gambit]

And if you trade that methodology on say all the symbols on the S&P500, is there a statistical edge over say five years?

I can't say for sure. Ernie's tests show some edge. I see more opportunity in the swing trading approach just because it is easier to put limit orders deep in the book and "make a market" on a wider time frame. Is there an edge? Not sure. I would say at least one can avoid paying the retail spread.

 

[Gambit]

And if you trade that methodology on say all the symbols on the S&P500, is there a statistical edge over say five years? What about a subset of the S&P500? which one?

If I had to trade this on equities, I would trade a basket. 3 or 4 stocks against a heavily concentrated index. It is my understanding that this trade is "picked over" on the shorter time frames but maybe not on the longer time frames.

 

[Gambit]

Good hypotheses. How do you go about testing it?

Do you see why this is so hard? Infinite hypotheses, almost infinite data sets with a finite life to live. And to make it worse, it is very expensive to set up a test harness in both time and money - no one will spoon feed you the answer. So retail traders put up pretty charts and gamble on what everyone else says works.

There used to be a guy that came here and said - "Retail traders lose. They just lose."

I see what you mean. Machine learning seems so useful in this situation.

 

[nitro]

I can't say for sure. Ernie's tests show some edge. I see more opportunity in the swing trading approach just because it is easier to put limit orders deep in the book and "make a market" on a wider time frame. Is there an edge? Not sure. I would say at least one can avoid paying the retail spread.

Sure, this will work. But it won't make you rich, unless you are already rich. The only way to get rich is to keep risk very tolerable and very much under control. Then scale. Then retire before the black swan gets you, unless you are Taleb and play the tails.

You might blow out small accounts a few times before you hit the big one.

 

[nitro]

If I had to trade this on equities, I would trade a basket. 3 or 4 stocks against a heavily concentrated index. It is my understanding that this trade is "picked over" on the shorter time frames but maybe not on the longer time frames.

3 or 4? The theory is ~25!

 

[nitro]

I see what you mean. Machine learning seems so useful in this situation.

Maybe. I don't use any machine learning. The only machine learning is the one between my ears.