But of course there are false positives. It's a classifier, no classifier that is automated on something as fluid as relevancy for a particular user will be 100% spot on all the time.
Precision and recall can't be 100% accurate given a large enough set of inputs. That would be magic. You can try to do better, of course. But it will never be perfect and you'll never make everybody happy. False positives are a given. Matt was wrong when he said that, he pretty much had to be wrong due to the nature of the problem.
That he stuck to his guns is imo a mistake, that google can be manipulated into dropping sites from their rankings at the behest of others is a serious problem. Such unscrupulous behaviour should be punished, but then you get yet another layer of complexity in the arms race.
Basically you can read this whole saga as Google having to come to terms with the fact that even though they were a cut above altavista they too will have problems that no algorithm will solve.
"That he stuck to his guns is imo a mistake, that google can be manipulated into dropping sites from their rankings at the behest of others is a serious problem."
There are two main ways of dealing with issues:
* Prevent them from happening
* Mitigate the risk so the fallout is minimised if it does happen
Because there are flaws in the first way doesn't mean the second hasn't been explored or carried out.
We mitigate risk all the time. Seat-belts, looking both ways before crossing the street, insurance. The entire banking system.
So you get hit with a manual penalty for spammy links. Deal with it. Document it in great detail, publish it, share it with Google. Then SEOers will have both the data they need and a Google rep to talk to. Plus, if it is as terrifying as SEOers keep telling us it is, the news headline boost alone will make up for the link profile damage.
I think I know why SEOers are in an uproar about this: they'll have to collaborate with Google on a new level of openness. Effectively, recipients of negative SEO attacks, who then notify Google, will also have a nice shiny Google light shone on their SEO tactics leading up to the negative SEO attack. And I guess almost all SEOers have done something they are not entirely proud of, or wish to be forthcoming about. It's the fear of being seen as something other than completely white-hat.
The ways to avoid reaching that uncomfortable point is:
* Hope their fellow SEOers don't hit them first with a negative SEO attack
* Hope Google reverses this decision, so they don't have to venture down that road.
It's a bit like that Simsons sketch where the whole family are in a shock therapy session:
Precision and recall can't be 100% accurate given a large enough set of inputs. That would be magic. You can try to do better, of course. But it will never be perfect and you'll never make everybody happy. False positives are a given. Matt was wrong when he said that, he pretty much had to be wrong due to the nature of the problem.
That he stuck to his guns is imo a mistake, that google can be manipulated into dropping sites from their rankings at the behest of others is a serious problem. Such unscrupulous behaviour should be punished, but then you get yet another layer of complexity in the arms race.
Basically you can read this whole saga as Google having to come to terms with the fact that even though they were a cut above altavista they too will have problems that no algorithm will solve.
Admitting that is probably above Matt's paygrade.