Points Added
Posted: Thu Jun 04, 2020 2:00 pm
Distribute credit for a shot's expected value* to all players who touched the ball on the possession, weighted (diminishing credit the more passes away you are from the shot)
*use XeFG% instead of whether or not the shot went in for points totals (https://thespax.com/2019-xefg/)
Example:
Rubio pass to Mitchell pass to Gobert pass to Ingles
1. What is the XeFG% on the Ingles shot?
1. translate this to a point value
1. let's say this was a 50% XeFG shot, so we are working with 1 point
2. Distribute the 1 point to Gobert/Mitchell/Rubio
1. not sure how the credit would work exactly (more credit for how much your pass improved your team's chance of scoring), but for these purposes, we could just say:
1. .65 to Gobert
2. .25 to Mitchell
3. .10 to Rubio
3. Turn into a rate metric, ie per 100 possessions or something of the sort
Now, obviously this is just a passing metric at this point but I think it would be even better were it tuned to all actions on the court. Basically, you would track a team's chance of scoring at every given point in a possession, based upon the tracking data. A simple version could just track on ball action (passing, penetration, etc) while a more complex version would track all actions.
Example:
what is the value a player provides to a possession, based upon how closely his man guards him while he is spotting up off the ball on the weak side? how much value does the dive man provide based upon how hard the 3 or 4 off-ball defenders crash to prevent a lob?
This could also work defensively. How much does a defender who makes a great rotation take away from the opposition's chance of scoring? How much does a defender who gets toasted in isolation add to the opposition's chance of scoring?
Soccer analytics seems to have cornered the market on this type of metric, at least publicly. I would love to see something along these lines for Basketball.
Inspiration: https://www.americansocceranalysis.com/ ... zormdziphq
I am a front-end dev (JS, CSS, HTML) and as such do not have the skills working with data to do something like this but would love if someone were to take this on. I would be willing to help in any way I could.
Website: https://craftednba.com/
Twitter: https://twitter.com/craftedNBA
Feel free to reach out. Thanks
*use XeFG% instead of whether or not the shot went in for points totals (https://thespax.com/2019-xefg/)
Example:
Rubio pass to Mitchell pass to Gobert pass to Ingles
1. What is the XeFG% on the Ingles shot?
1. translate this to a point value
1. let's say this was a 50% XeFG shot, so we are working with 1 point
2. Distribute the 1 point to Gobert/Mitchell/Rubio
1. not sure how the credit would work exactly (more credit for how much your pass improved your team's chance of scoring), but for these purposes, we could just say:
1. .65 to Gobert
2. .25 to Mitchell
3. .10 to Rubio
3. Turn into a rate metric, ie per 100 possessions or something of the sort
Now, obviously this is just a passing metric at this point but I think it would be even better were it tuned to all actions on the court. Basically, you would track a team's chance of scoring at every given point in a possession, based upon the tracking data. A simple version could just track on ball action (passing, penetration, etc) while a more complex version would track all actions.
Example:
what is the value a player provides to a possession, based upon how closely his man guards him while he is spotting up off the ball on the weak side? how much value does the dive man provide based upon how hard the 3 or 4 off-ball defenders crash to prevent a lob?
This could also work defensively. How much does a defender who makes a great rotation take away from the opposition's chance of scoring? How much does a defender who gets toasted in isolation add to the opposition's chance of scoring?
Soccer analytics seems to have cornered the market on this type of metric, at least publicly. I would love to see something along these lines for Basketball.
Inspiration: https://www.americansocceranalysis.com/ ... zormdziphq
I am a front-end dev (JS, CSS, HTML) and as such do not have the skills working with data to do something like this but would love if someone were to take this on. I would be willing to help in any way I could.
Website: https://craftednba.com/
Twitter: https://twitter.com/craftedNBA
Feel free to reach out. Thanks