nflmetrics

GitHub repo for Johns Hopkins Spring 2022 Sports Analytics research project about NFL Draft Metrics.

Resources:

https://collegefootballdata.com/ – limited traditional data, but easy to download to CSV and has API – solid for EPA, PPA, or usage metrics

https://www.playerprofiler.com/terms-glossary/ – has cites for all advanced stats and combine/physical measurements used – plus formulas for how to compute some of these stats

https://secstatcat.com/sec/ – great advanced stat resource for QBs, RBs, and WRs, but only for SEC players – Paywall (use Incognito mode)

Position Groups Advantages/Disadvantages

Offense

QB

feels like combine performances rarely correlate with NFL success, would have to evaluate based on decision-making (general QB IQ), scramble ability, deep ball ability, etc.

RB

can they find the hole, how good is their oline, how shifty, how fast? Last 2 are supposedly measured by combine drills

shouldn’t be too bad to gather college data for this

WR

really hard to find statistics at the college level for this, would most likely have to figure out how advanced stats such as average separation are calculated

figure out how to measure game IQ, analyze defense (i.e. find the weak spot in a zone)

NFL data is really solid though

would be fun to examine combine metrics such as 40 yard dash, vertical, and 3-cone drill to determine their predictive ability

how much ability is due to scheme or pure talent/skill

how many plays were made on legit CBs?

TE

who knows bruh half the colleges don’t use them very well

OL

lowkey easiest position to analyze with film, which could be handy if we have to look through college tape

create some kind of index that takes player body weight data and calculates a theoretical win % based off the pass rusher and whether they actually win it or not

Then pass/run block wins in scenario where theoretical win % is low stand out

traditional combine drills such as bench press legitimately matter

Defense

DL

EDGE

might be easy to gather data at the college level for this? Pass rush win rate and also quality of line have to be considered; some will look more impressive because entire line is stacked

LB

CB

S

New combine drills: https://www.sportingnews.com/us/nfl/news/nfl-combine-new-drills-2020-explained/bbg40xpym8uy1bb1yrzckdxfb

Maybe in general just need to develop a better game IQ test since the wonderlic is cheeks; such a thing would have a big impact in QB, WR, LB, and S evaluation
In general, it shouldn’t be too hard to look at general college stats, then look at combine performance relative to position group, then look at NFL output; multiple regression here

For unsorted list of draft picks ranked according to NFL production, can run through a few NFL drafts and see if there are patterns within position groups for certain stats like combine 40 yard dash or college production (idk, maybe players who fill up the stat sheet against nationally top 5 schools all translate better)

GitHub

View Github