It is a significant bias known as "survivorship bias" in other forums,
so bad that it invalidates your backtesting completely.
What you need is the data as it was at the time.
Let me give you another example, if you screen based on price as it
looks today, as opposed to what it was at the time, you will probably
find that cheap stocks so many years ago are doing best. In fact you are
just selecting stocks which 1) survived and 2) went thru several splits.
Indeed good stocks, but wrong backtesting since those stocks somehow are
aware of their future (Crystal ball).
HTH ..Pierre
toddk63 a écrit :
I've wondered about this also and the effect it has on my backtesting.
I have a LONG strategy that scans the NDX 100 list. I am backtesting
10 years. The backtesting suffers from what I call "survivor bias".
That is, the only candidates for trades are the stocks that are
currently in the index. The ones that went belly-up or got aquired
are not included in the backtesting. To me, this means that were some
potential trades upto 10 years ago that the backtesting did not see
and it is probable that these would have been bad trades since they do
not exist anymore.
Anyone else thought about this? Is this a significant bias?
Todd K.
--- In [email protected]
<mailto:quotes-plus%40yahoogroups.com>, "xxnospamxx" <[EMAIL PROTECTED]> wrote:
>
> Are delisted stocks kept in the database? My preference will be to
> keep the delisted stocks if this is configurable. Sorry for the newbie
> questions.
>
> Steve
>
