Read on Twitter
As an individual who directly or indirectly involve in "top analytics team" hiring process let me give my two cents here (1/n) https://www.espn.com/nfl/story/_/id/29939438/2020-nfl-analytics-survey-which-teams-most-least-analytically-inclined">https://www.espn.com/nfl/story...
I& #39;d vote the same for top 3 as well. For the exact ranking I don& #39;t precisely know what "analytically advanced" really is, but Lamar Jackson being on the record an aggressive quarterback on 4th down obviously helps 2/
Assuming this question is not exactly the same as last one, Browns being the best is understandble since on paper they have much more experienced staff than Ravens do 3/
Titans still have no analytcal staff so not suprising to top the list, but imo some them are being affected by poor team performance which is mostly independent of analytics. e.g Giants actually one of the most developers in the league after hiring "computer folks" last season 4/
It& #39;s good to see many teams use tracking data and I personally would guess 9-16 as well, but let me further dive on this.
NFL analytics is a unique problem in data science due to its abundence of labelled data by company e.g. PFF, SIS.. etc. 5/
NFL analytics is a unique problem in data science due to its abundence of labelled data by company e.g. PFF, SIS.. etc. 5/
If you have many labelled data at low cost (or at least in reasoble cost since every teams has subscribed for PFF data), you actually don& #39;t really need that good modelling from tracking data!
That being said big data bowl and usage of tracking data is still invaluable... 6/
That being said big data bowl and usage of tracking data is still invaluable... 6/
For instance modeling techinique showcase by past and probably future big data bowl require a high understanding of machine learning technique which is above the level of lots of data analyst including me. I& #39;d even argue BDB is one of the most useful competition on kaggle ever 7/
It& #39;s not suprising to see game management on top but actually still somewhat disappointing to see. 4th down decision inefficiency is suggested quite a long time ago, and the gap between team is closing rapidly since punt is the correct decision in lots of cases 8/
Beyond 4th down decision and lesser impact 2 point conversion, football analytics actually are not well understood enough to give a concrete suggestion. Even early down pass/rush ratio is still very team dependent and you really can& #39;t grab a chart and make a fair comparison 9/
For opponent scouting, "X player in team Y run 80% in Z personnel" is still prone to small sample size bias. Not saying it& #39;s not useful since it& #39;s all how analytics work in other sports, but it& #39;s more like a baseline than what analytics ought to impact a team 10/
Conversely, there are still quite an amount of ineffieincy of positional bias of draft, overpaying running backs and underestimate defensive backs etc. which still require analytical input and not hard to see the result. 11/
Baseball teams average near 10 analytical staff per team and it& #39;s 2-3x more so it& #39;s not even that suprising to see a 100%+ increase! (Another way to think of it is least ananlyticall staffed team in baseball = most staffed team in football). 12/
Still with increasing number of staff need more buy-in from the management, which can be said is the #1 priority over any analytical work done, but of course you need a constant good quality work and good collabration in order to impress "tradtional football people" 14/
Finally if you are curious what NFL team can improve to bring on more analytics, this article written by Sean Clement before went to work for Ravens is a very good read. https://beastpode.com/2019/05/22/new-general-nfl-coaching-staff-front-office-analytics-coach-general-manager/">https://beastpode.com/2019/05/2... 15/
The article outline how evidence-based research can affect a team from different aspects and how to make better decisions with collbration of coaches, staff, analyst and developers. 16/16
