Good and bad books on strategy design?

[kut2k2]

Quote from ssrrkk:

Like I said I haven't used ARIMA models so I can't help you there. In general, you can get confidence levels on your parameter estimates for your model and that should tell you whether you have a good estimate or bad, not a universal cutoff on the number of training points.

I wasn't asking about ARIMA, I was just using that as an example of what I was talking about. I myself don't have a model, I measure what price action is currently doing. I can do this in as few as 9 data points but something in the mid teens is better.

So, how many data points does YOUR model require?

 

[ssrrkk]

Quote from kut2k2:

I wasn't asking about ARIMA, I was just using that as an example of what I was talking about. I myself don't have a model, I measure what price action is currently doing. I can do this in as few as 9 data points but something in the mid teens is better.

So, how many data points does YOUR model require?

LOL, this is not a competition for the smallest number of training points. My current interest is not in completely empirical models. I'm focused on models based on known or defined relationships between instruments. So the amount of training or fitting is really not that important; in fact you might argue that everything could be derived completely theoretically -- however some amount of training is required for deviations from theory or as shortcuts. For example, everyone knows the theoretical relationship between an option premium and volatility, that kind of thing, but applied to other instruments, e.g., ETFs and Indices, their components, their volatilities, options and futures prices, etc.

 

[kut2k2]

Quote from ssrrkk:

LOL, this is not a competition for the smallest number of training points.

Of course not. I'm merely curious about how many data points your system requires to take off. Presumably there's some minimum.

Quote from ssrrkk:

My current interest is not in completely empirical models. I'm focused on models based on known or defined relationships between instruments. So the amount of training or fitting is really not that important; in fact you might argue that everything could be derived completely theoretically -- however some amount of training is required for deviations from theory or as shortcuts. For example, everyone knows the theoretical relationship between an option premium and volatility, that kind of thing, but applied to other instruments, e.g., ETFs and Indices, their components, their volatilities, options and futures prices, etc.

So does your system allow directional trading or single-instrument trading?

 

[ssrrkk]

Quote from kut2k2:

Of course not. I'm merely curious about how many data points your system requires to take off. Presumably there's some minimum.
So does your system allow directional trading or single-instrument trading?

I am focused on hedged portfolios so multiple instrument only. The only time I am considering non-hedged positions is when I have a microstructure advantage, in which case I get in and out quickly -- I don't think that qualifies for being directional.

 

[blueraincap]

New Trading Systems and Methods

 

[GloriaBrown]

seem like Ruey S. Tsay, Analysis of Financial Time Series is not an easy reading.

 

[Rodney King]

Quote from GloriaBrown:

Ruey S. Tsay, Analysis of Financial Time Series is not easy reading.

Nor is the ~800 page Hamilton book suggested in another thread. http://press.princeton.edu/TOCs/c5386.html

 

[vikana]

Both Hamilton and Tsay are excellent books. I'm just not sure they contains much directly applicable information relevant for trading. They are all about the mechanics of time series under various assumptions. Good stuff, but there's quite a jump from those books to working trading systems.

 

[vincegata]

Hamilton is the ultimate text for time series, it's very difficult and very theoretical, use it if you write a PhD thesis. Tsay is not a walk in a park but definitely easier than Hamilton and it's more appropriate for trading.

 

[GloriaBrown]

This is a very good post, keep this up!