[Vikings OC] An Attempt at Calculating WAR in the NFL
Posted: Wed Feb 21, 2018 7:56 pm
An Attempt at Calculating WAR in the NFL
Read more: http://www.purplepainforums.com/thread/ ... ng-war-nflReady for something completely out of left field? Speaking of left field, it actually has something to do with baseball. A while back I had the idea of trying to implement the advanced statistic called WAR that is commonly used in baseball and try and translate it into the NFL. So I've set some time aside to conduct a fun experiment that will most likely result in any viable or realistic statistics.
You're probably asking yourself, "What the heck is WAR anyways?"
"Wins Above Replacement (WAR) is an attempt by the sabermetric baseball community to summarize a player’s total contributions to their team in one statistic" (Source: Fangraphs.com). MLB statheads like to use this stat to try and calculate the value the team would lose if a certain player had to be replaced by a backup or free agent.
The way WAR is calculated by many websites is rather complicated, but my attempt here is going to be rather simple. It's far more difficult to quantify the performance of a NFL player than a MLB player just because of the wide array of stats baseball players accrue. So for this experiment, I'm going to use PFF ratings only because that's the only method I have that can put all players on the same grading scale (I apologize for using PFF grades so often, but they're basically all I have to do this). This also means that I will not include punters and kickers as I don't have a great way to compare their performances to the rest of the players. Perhaps they could be added later.
First what I'll have to do for this experiment is to find what replacement level actually is. It falls somewhere between the "average" player and "worst" player at each position. After tinkering with several numbers, I have decided to use 35% as the cutoff for replacement level. If a player is worse than 65% of the other players at his position, then he is a "sub-replacement" player and he will carry a negative value.
Ranking all players wasn't easy - I took every single snap played by every single player (even the guys who recorded a total of one snap on the season) and I was able to find replacement level at 35%. Here are the results...