|
PureBytes Links
Trading Reference Links
|
traders,
Thanks for the tip about problems using the Nasdaq 100 for longterm
backtesting. You've saved me some time and that is appreciated.
Fortunately I don't need to use the Nasdaq 100. I just need a
reasonably good proxy for the QQQ (and I have data on over 70 trades
ins the current QQQ data to verify whether or not something will be a
good proxy). I'm not looking for certainty here -- just a little
longer overview. So here's another approach I could take.
1. Make up a couple of my own versions of the Nasdaq 100 (ones that
will have a static membership over the test period).
2. Do a scan that selects the current 100 largest cap stocks on the
Nasdaz (excluding the financial since the QQQ also excludes them)
that have a minimum of 10 years of price history. Build a composite
of these and compare to the price behaviour of the QQQ. If it a
reasonable proxy, then analysize each of the individual stocks to
develope my desired indicator. Then I could then extend the back
testing.
3. The second composite proxy could be made as in 2 but using 1995 or
1990 as the sort year for getting the largest cap stocks.
As you can tell I really want to get a proxy for the QQQ. It don't
need to be exact, just close enough to determine how often my
indicator model gives a really bad calls. Comments anyone?
--- In amibroker@xxxx, traders10@xxxx wrote:
> There are more problems here than just the test time involved.
>
> The NASDAQ 100 is in a constant state of flux. Stocks are
frequently
> added and deleted from the index. If you try to reconstruct the
> index from the individual component stocks you will find it an
> impossible task. NASDAQ does not make available a history of the
> changes in the component stocks (AFAIK) so you are out of luck
there.
>
> The current composition of the NASDAQ 100 represents a considerable
> survivorship bias that negates historical reconstruction for
> component stocks. ie: all the losers got dropped from the
> index....only the winners remain.
>
> I believe that you will only be able to go back perhaps two to
three
> years before the testing becomes meaningless.
>
> Sorry
> Trader
|