An outstanding effort
(https://espnfivethirtyeight.files.wordpress.com/2015/04/paine-datalab-nfldraft1.png?quality=90&strip=all&w=575&ssl=1)
Here is the article that accompanied the chart
https://fivethirtyeight.com/datalab/are-some-positions-riskier-to-pick-than-others-in-the-nfl-draft/
Well looking at the chart it appears drafting a LB in any round is a 50 / 50 shot. Yet the Giants have done miserable. I can't remember a LB who has stuck although I a could name quite a few who are long gone.
It's an interesting view however the article does seem to state that the metrics for the non-skill positions are not reliable or complete the farther back you go in time. It would also be beneficial to see how those figures changed as the game changed. The NFL of 1980-1990 is pretty different than the NFL of 2000-2014.
Sadly Larry that is a spot on observation
quacker,
I think they went back further to gain a proper number of data points. Cutting the sample size would account for the current NFL and some trends. A trend I think is relatively new and hurting prospects is the spread offense. A first round O-line bust used to be very rare, now thanks to the spread offense they seem more common.