BINGO...
The "matches" are "proof" of precognition, the "misses" are "just ordinary dreams".
Heads she wins, tails she doesn't lose. Anyone could accumulate a "scary" track record by those rules, especially someone who tends to dream scenarios about people and situations in her actual life, and who remembers a lot of their dreams (unlike me, I tend to have "what in the hell was *that*?" dreams, not involving day-to-day people/situations, and I seldom remember my dreams for the most part). Such a person would provide a lot of "material" for lucky hits (and the misses don't count).
And don't understimate the effectiveness of ordinary non-paranormal intuition. For example, it's not unlikely that a person would have a dream about a friend having a car crash, starring a friend who is known to drive too fast and carelessly -- and then have it happen, for obvious reasons.
We always remember the striking coincidences, we overlook the thousands of times things don't coincide. Anecdotal evidence is very poor "evidence" for psychic ability, because most people have a poor grasp of how often "amazing" coincidences will happen purely by chance. They also have a tendency to give "matches" a higher score than they deserve (e.g. dreaming someone dies one way, and they die another, and counting that as an actual "omen", despite the fact that dreams about someone dying happen quite often.)
First, send out "psychic" prediction tipsheets free of charge to 64,000 people. Half predict a certain stock will go up; the other half predict the same stock will go down.
The next week, send out a new tipsheet only to those 32,000 people who received the tipsheet from last week with the correct tip.
The week thereafter send out your tipsheet to the 16,000 with two weeks of accurate predictions.
After six weeks of this, ask the 1,000 people left for $1000 dollars for a year's subscription to your tipsheet.
Retire.