2018 Pass Completion Percentage map

[Ring 6]

But...they're not the same throw. As the chart shows, the 25 yard seam travels 30% farther in the air. This is starting to feel like that Solomon Asch experiment!

In this version the two vectors are the same length up to the arroheads, and the dotted line represents the difference in the length of the two throws.

The field is 53 yards wide. Throwing zero yards downfield to the sideline is the same distance as 26.5 yards downfield straight ahead. The graph cannot be to scale if a 25 yard seam is 30% farther than a throw to the sideline 26.5 yards away.

 

[neuronet]

The field is 53 yards wide. Throwing zero yards downfield to the sideline is the same distance as 26.5 yards downfield straight ahead. The graph cannot be to scale if a 25 yard seam is 30% farther than a throw to the sideline 26.5 yards away.

50% completion contour is not an ellipse so it isn't just about distance. Plus it does seem to scale so should be a circle.

 

[Ring 6]

50% completion contour is not an ellipse so it isn't just about distance. Plus it does seem to scale so should be a circle.

I’m saying aside from the deep middle it kind of is.
It doesn’t look to scale to me.

 

[neuronet]

I’m saying aside from the deep middle it kind of is.
It doesn’t look to scale to me.

Just looked at original more closely on my laptop previously I was measuring on my phone with my fingers.

You are right it isn't quite to scale: horizontal 10 yards is a bit longer (edit: shorter) than vertical 10 yards (edit: actually they are really close) Plus, the horizontal axis goes to +/- 30 yards, so I'm not sure what they did there it should technically go to +/-26.75 maybe they just rounded up to the nearest 10. At any rate, even with any scaling chicanery it would be an ellipse if it just depended in distance.

Edit: Plus they are not that different: horizontal 10 yards is like a couple percent different so it should basically be a circle. At any rate I think it is pretty safe to just use distances on the plot and not worry about this crap.

 

[Ring 6]

Just looked at original more closely on my laptop previously I was measuring on my phone with my fingers.

You are right it isn't quite to scale: horizontal 10 yards is a bit longer (edit: shorter) than vertical 10 yards (edit: actually they are really close) Plus, the horizontal axis goes to +/- 30 yards, so I'm not sure what they did there it should technically go to +/-26.75 maybe they just rounded up to the nearest 10. At any rate, even with any scaling chicanery it would be an ellipse if it just depended in distance.

Edit: Plus they are not that different: horizontal 10 yards is like a couple percent different so it should basically be a circle. At any rate I think it is pretty safe to just use distances on the plot and not worry about this crap.

But what I am saying is if you overlook the deep middle it pretty much is a circle.
Also it’s deceptive that 48% looks drastically different than 52% which can simple be explained by the fact the there are often underneath defenders the closer you get to the middle of the field.
It’s nog like blue represents 100 and red represents zero.
If they changed the criteria from 50% to 60% or 47% how would that change the dividing line?

 

[neuronet]

But what I am saying is if you overlook the deep middle it pretty much is a circle.
Also it’s deceptive that 48% looks drastically different than 52% which can simple be explained by the fact the there are often underneath defenders the closer you get to the middle of the field.
It’s nog like blue represents 100 and red represents zero.
If they changed the criteria from 50% to 60% or 47% how would that change the dividing line?

I agree the colormap could be better the current one makes it look a bit binary. It would be fun to get this data and draw isocontours for different percentages, make sure the scales were perfect, make a jupyter notebook for people to mess with.

 

[neuronet]

OK did some digging and found original data the woman who did this now works as an analyst for the Ravens. She did some really cool stuff at CMU. Here is raw data on left, what we are seeing is smoothed, interpolated, and extrapolated at edges on the right.


This is a really impressive achievement of analytics she had to scrape this data from images:
Introduction to next-gen-scrapy - CMU Sports Analytics

Unfortunately she is an Eagles fan. And she removed the repo from github. It seems she also has taken down the original tweets. Maybe the Ravens now own the code?