ronburk - 2:40 am on Apr 13, 2005 (gmt 0)

Bottom line. If a parameter has a little relation to the good content, like, for example, the period of the domain registration, do not bother about it.

The success of data mining is the success of brute force formula construction doing a better job at discovering what is reliably relevant than humans can do.

Even if the Google will use this parameter, it stops to do it quickly.

Of course! By using data mining instead of manually tweaking via human intervention, Google can automatically abandon or de-emphasize a variable quite quickly -- and do the exact opposite if conditions change again.

If it is irrelevant it is irrelevant.

Again, the reason data mining is used more and more rather than less and less is that, whether it comes to interpreting astronomical images or deciding which supermarket customers should be mailed coupons for which products or interpreting search results, data mining can often do a better job of deciding what is relevant than humans can.

When the data relationships to be mined are quite complex, you can pretty reliably change that "often" to "always".