Re: CL vs. K/KDB or efficient analysis of financial time series

>It won't quite work as you suggest. Your approach would normally be the
>one that I would follow but in this case I _am_ looking for top
>performance. I want to create best-of-the-breed software or get as
>close as I can. My system would handle price ticks in real-time and
>according to traders milliseconds are important these days.

We are still talking about backtesting a trading strategy, right? Every
millisecond, you want to go back and retest the same strategy over the
last million entries because the new information you received this
minute might have an effect on the result?

Backtesting a strategy requires a lot of data and could take some time,
but it doesn't need to be done quickly or often. "Real-time" has no
meaning when it comes to backtesting.

Employing the strategy requires only 400 points and will be blazingly
fast, in fact it will be hard to find a slow way to do this. Speed is
important for this part, but you get the speed for free.

I probably just misunderstood your problem, but to me there just
doesn't seem to be a problem.

.