Quote from nitro:
You know, I have thought about this for a long time, looking under every nook and cranny for explanations, and I have come to one conclusion. Sometimes the only predictor of a market is itself. This can happen when there is chatter like QE2 and other government talk that while does not get actually implemented, the market believes will take place anyway. In fact, the government hopes that the market does it's job for it instead of having to shove it down its throat. Other times the government is sending up a test balloon.
The question becomes, how on earth do you model something that is not quantitative? Further, markets have a tendency to overshoot. How do you decide how much of something that has no tangible value has overshot?
I don't know the answers to these questions, but if I did, all it would change is the chain widths so that I would still be playing for convergence, but I would be doing it more conservatively on certain types of momentum. Using VIX, or at least traditional VIX does not appear to resolve the problem when the market is melting up, although it does when the market is melting down.
Anyway, I feel that NFV and particularly ANFV is very close to being "true", but this particular problem and the best way to model important "future news" is nagging at me.
Look at "Why stock markets crash" by D Sornette he has a nice theory that sometimes markets feed on themselves and go parabolic (has nice equation for it) then a tiny insignificant thing happens and a crash results, basically the market has gone hyper critical. Although I am sceptical the S&P has gone parabolic unless of course you look at it from 666 the maybe it has.