Here is a brief synopsis of last Tuesday’s film study event with a sampling of the material presented. The presentations themselves will be available online in the next day or 2.
Ken started off the night with stills of the 2018 Ravens defensive packages and discussed their application by offensive personnel groups, down/distance, and frequency of use. The focus was how to identify the package quickly from your stadium seat. Below, they have their “Big Nickel” package aligned with Chuck Clark at slot corner.
Michael Crawford explained the counter run concept, including some interesting points about blocking angles, how the RB uses his first step to sell misdirection while also setting up timing of the blocks by pullers. He layered in how the other team’s need to account for Lamar Jackson can overcome disadvantageous numbers.
Acie Slade (BigPlayReceiver on the RSR boards & Twitter) took us through some outstanding visualizations of Ravens data from 2018 including this graphical representation of QB development from year 1 to year 2.
Ken and Michael followed up with a joint presentation on pass rush scheme/deception. Michael explained concepts such as dropping multiples from the line of scrimmage to coverage, stunts, and various blitz forms. Ken followed up with a comparison of Rex Ryan 2006, Dean Pees 2017, and Wink Martindale 2018 in terms of how they used numbers and deceptive elements.
Filip Dabek, a PhD student in machine learning and visualization, demonstrated how visualization was useful to find hidden trends and outliers in data and emphasized how important it is to have colors which 1) add information and 2) follow expected patterns (e.g. red is loss, green is profit). He then provided tips in a case study of pass rush data to show how creating a useful visualization is an iterative process.
Caleb Wharton, another data scientist by day, then gave a presentation on web scraping. He used existing game data to perform a study to isolate a specific set of run plays. His techniques would be valuable as the first step in any number of hand analyses where the existing data can be used both to identify plays and prepopulate fields where other observations could then be layered in.
Thanks to all the presenters, as well as the participants, and especially Fazzini’s for not kicking us out at closing time!.