The original versions were very bad. The new version ("NFV") which I have been running for a few months are jaw dropping accurate in that it rarely takes more than a week for full convergence, and the convergence is most often SPX converges in the direction of "NFV".
I have always regarded as the ultimate test of a "system" a quality called "recovery rate", or how fast it recovers from losses. If you follow simple rules of position sizing, such as for example, do one contract when NFV and SPX diverge by seven or more, add n contracts more if they continue to diverge by 14 or more, then add p more contracts if the divergence continues for 24+, you will have 1 + n + p contracts, at which time on convergence you will have a healthy profit. n and p can be progressive, such as n = 2 and p = 6, so that if SPX and NFV diverge by more than 24, you will have 9 contracts on (or some multiple of whatever your units are) with the highest weight on the lowest price in your entry sequence. Alternatively, you can use S/R levels to add, but that is very conservative, imo.
I have an idea for a high frequency version that is always (except in extremely rare cases of 1 or less divergence in which case position is zero) in the market on some position weight based on divergence gradient. The only way to use it as posted on ET is as a swing system.
FWIW, there are several dozen people that have PM'ed me to say that follow NFV, and many of them say it is very helpful to them. In rising markets I think it loses some predictability although still viable. We will have to wait and see.