The progression of fantasy football projections is basically at the same point individual stock projections were at in the mid 1900's. The analogy here is that most fantasy football analysts perform some analysis and calculation not dissimilar to a discounted cash flow analysis.

I.E. Break down all the different parts of a player's game, evaluate the prospects of that particular portion going forward, piece it all back together, and voila! You've calculated some value.

When many analysts do this in unison, their collective efforts are more accurate than any one individual. It becomes hard to "beat the market" at the same game that the market is playing. Perennial winners like @The_Oddsmaker are finding a more efficient fantasy football market.

This doesn't mean it's impossible to outperform, but the odds become lower. And this means that the projections game is encouraged to evolve.

Suddenly it's not too hard to accept consensus rankings as some semi-efficient market. With that framework installed, the most powerful form of analysis is actually to evaluate the market's biases toward value as opposed to the player's value itself.

The financial analogy here would be the Fama-French factor model approach ( https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=1455).

I.E. Let's not worry about projecting some value Y using some predictive variable X. Let's worry more about predicting the "efficient" market's average error e using that same predictive X variable. Yes, meta.

We shouldn't care that X predicts some value Y. We just care that the public is usually wrong about X in a predictable way. This is the future of fantasy football projections.

I'm not just positing, I've already found some statistically significant RB factors myself:
1) Target_Share
2) Age
3) Same_Team
4) Plays_First_Game

1) RBs with high target shares tend to outperform their ADP
2) Young RBs tend to outperform their ADP
3) RBs that don't change teams in the offseason tend to outperform their ADP
4) RBs that play the first game of the season tend to outperform their ADP

A RB that checks all 4 boxes (>10%, <26, 'yes', 'yes') tends to outperform a RB that checks no boxes (<=10%, >=26, 'no', 'no') by ~2 fpts per game! That adds up across many draft picks.

The public leader in this charge is surely @friscojosh. His air yards model opens the door to a factor model. His infatuation with binning is factor model convention. RBs don't matter because the public continues to overweight their differences. Etc.

I think that fantasy projections are more efficient that stock market prices. But that's likely only because so few choices and variables exist in fantasy football and not because analytic methods are more evolved. I see fantasy football analysis going the way of stock analysis.

I'm not sure I'll be the one to find those factors or publish that data. But I'm sure @friscojosh @FFzinger @pahowdy @MikeClayNFL @Maatspencer @benbbaldwin @kczat and others will. The future of fantasy football analytics remains bright.