|
PureBytes Links
Trading Reference Links
|
>Subject: Problems with *stock* data back-adjustments
> Date: Sun, 13 Jun 1999 18:45:09 -0700
> From: Mark Johnson <janitor@xxxxxxxxxxxx>
> To: omega-list@xxxxxxxxxx
>
>Jose Pascual writes:
> %
> % So better use real futures contract data while using
> % or choosing of the 20,000 plus available securities
> % from NASDAQ, NYSE, etc... to backtest. It should work
> % on any ...
> %
>
>But remember that stock data HAS BEEN BACK-ADJUSTED
>(for splits). This can render backtests invalid.
>
>Here's an example: Jeff Cooper's "5 Day Momentum Method"
>only trades stocks whose price is above 50. It's one
>of the setups used by the system.
>
>And indeed an example setup in the book is Kellogg
>(stock symbol "K") on 21 July 1997. {page 22 of
>Cooper's book} On that day Kellogg stock was
>trading at 87, so it met Cooper's setup criterion
>(Price > 50) quite easily.
>
>But since then, Kellogg's stock has split. Go look
>at your stock price database --- you'll find that
>it says Kellogg was at 42 on 970721. So, in backtesting,
>THE TRADE WAS NOT TAKEN. But in real life, as demonstrated
>by the chart in Cooper's book, it was.
>
>You have just been misled by back-adjusted stock data.
>--
> Mark Johnson Silicon Valley, California mark@xxxxxxxxxxxx
This is only one aspect of the problem. Many indicators will also give
false indicators because of splits. Any indicator which uses an
absolute
value will have that value reduced by the ratio of the split. (We've
discussed this here before.)
This includes momentum, true range, volatility, and a few others.
Any systems developed with these indicators will not give meaningful
results.
Which is why I have five years of unadjusted EOD data. On some stocks.
donc
|