Hi All,
Latest post is up where I have developed a new metric to measure the effects of "Coach Killing" actions. The purpose of this metric is not to tell us that a certain amount of Coach Killing plays cost “x” amount of wins. The purpose is to find a way to measure players who may do more harm than good on the court and who consistently demonstrate poor decision making and basic skill errors, which may then creep into other facets of their game.
In this post, I take readers through the details of how the metric was constructed and provide results of the data visually through tableau which will allow you to interrogate the raw data and sort through the findings.
It's a work in progress, so would love to hear feedback.
Check it out at zigzaganalytics.com
New metric to measure "Coach Killers" in the NBA
Re: New metric to measure "Coach Killers" in the NBA
You might as well take the extra step and actually use a link:watto84 wrote:...
It's a work in progress, so would love to hear feedback.
Check it out at zigzaganalytics.com
http://www.zigzaganalytics.com
Re: New metric to measure "Coach Killers" in the NBA
This metric tells who does harm (in the box score, not in shot defense or indirect offensive impacts). But for "who does more harm than good", one needs the more complete or truly complete metrics that have bad and good.
Some players will provide large amounts of good and bad. Some less of each. The nets might be similar. The affect of wins might be different depending of lineups, overall team, game to game and minute to minute distributions. Looking at win impact based on changes to team win probability by play may be better than valuing every play equally. The box score metrics treat them equally. RPM I think varies but I am sure exactly how and how much it is doing it these days
Some players will provide large amounts of good and bad. Some less of each. The nets might be similar. The affect of wins might be different depending of lineups, overall team, game to game and minute to minute distributions. Looking at win impact based on changes to team win probability by play may be better than valuing every play equally. The box score metrics treat them equally. RPM I think varies but I am sure exactly how and how much it is doing it these days
Re: New metric to measure "Coach Killers" in the NBA
Haha thanks for the link Nate! Rookie error on my part.
Crow...yep I get what you are saying with the net effect that could/should be considered by factoring in "good" plays, but the reason I wanted to explore this was simple, at any level it doesn't have to be NBA, we have all played with someone who just consistently makes poor decisions on the court and other then being frustrated by that player it's hard to guage what harm they are doing, so by creating a simple "points" calculation we atleast can say "this guy costs atleast 10 points a game from dumb plays" and it opens them up for further investigation.
With a lot of metrics that have "good" in them, they can be distorted by a player who scores the ball, so I just wanted to try and look at it from a different angle to evaluate players a little differently.
Crow...yep I get what you are saying with the net effect that could/should be considered by factoring in "good" plays, but the reason I wanted to explore this was simple, at any level it doesn't have to be NBA, we have all played with someone who just consistently makes poor decisions on the court and other then being frustrated by that player it's hard to guage what harm they are doing, so by creating a simple "points" calculation we atleast can say "this guy costs atleast 10 points a game from dumb plays" and it opens them up for further investigation.
With a lot of metrics that have "good" in them, they can be distorted by a player who scores the ball, so I just wanted to try and look at it from a different angle to evaluate players a little differently.
Re: New metric to measure "Coach Killers" in the NBA
The thing is, there doesn't seem to be any sensible connection between the "bad points" in the metric and "bad decision making."watto84 wrote:... I wanted to explore this was simple, at any level it doesn't have to be NBA, we have all played with someone who just consistently makes poor decisions on the court and other then being frustrated by that player it's hard to gauge what harm they are doing, so by creating a simple "points" calculation ...
For example, the metric has a turnover coefficient of 1.039. Let's suppose someone else comes along and has the same metric except that the turnover coefficient is 1.06. How can we tell that 1.039 is a better choice for that coefficient than 1.06?
On a different level, it seems rather implausible to me that the box scores actually capture information about 'bad decision making'. Now, if there were some kind of authoritative list of how much particular players make bad decisions, it might be possible to correlate that to the box score, but we don't have that kind of list.
Re: New metric to measure "Coach Killers" in the NBA
Bad points = what that loss of possession cost in terms of potential points on the scoreboard. Last season on average a possession was worth 1.039 points. So when you turn the ball over it has cost your team the chance of 1.039 points. Hence I penalise a player that amount.
If you look at any linear weight formula just like PER you will find actions are given a "weight" which in basketballs case is the points value of that action. Turnovers are weighted by the cost of a possession. PER had to make a choice for its weights and long term league averages were used. I chose to use last seasons averages.
The list of bad decisions I have come up with is found in play by play data which has a rich list of data that can be interrogated. Eg I can look at time left in shot clock and what action came before and after a different action. The decision list was based on my own experiences of playing basketball professionally and also quizzing coaches on their pet hates.
It's not supposed to be the holly grail, just potentially a different way of evaluating players and assigning an easy to understand "this player costs us x amount of points from actions we as a team hate and hurts the team". An individual team would have different views on what a bad decision is and could come up with their own unique list that may include shot selection. In my case I tried to keep it generic for all teams.
Alot of metrics out there come up with a number to rate a player but coaches can find it hard to correlate what that actual number means on the scoreboard, so this metric tries to dumb It down and produce a number a coach can relate and understand its impact on the team.
If you look at any linear weight formula just like PER you will find actions are given a "weight" which in basketballs case is the points value of that action. Turnovers are weighted by the cost of a possession. PER had to make a choice for its weights and long term league averages were used. I chose to use last seasons averages.
The list of bad decisions I have come up with is found in play by play data which has a rich list of data that can be interrogated. Eg I can look at time left in shot clock and what action came before and after a different action. The decision list was based on my own experiences of playing basketball professionally and also quizzing coaches on their pet hates.
It's not supposed to be the holly grail, just potentially a different way of evaluating players and assigning an easy to understand "this player costs us x amount of points from actions we as a team hate and hurts the team". An individual team would have different views on what a bad decision is and could come up with their own unique list that may include shot selection. In my case I tried to keep it generic for all teams.
Alot of metrics out there come up with a number to rate a player but coaches can find it hard to correlate what that actual number means on the scoreboard, so this metric tries to dumb It down and produce a number a coach can relate and understand its impact on the team.
Re: New metric to measure "Coach Killers" in the NBA
OK. That wasn't clear to me from the post.watto84 wrote:Bad points = what that loss of possession cost in terms of potential points on the scoreboard. Last season on average a possession was worth 1.039 points. So when you turn the ball over it has cost your team the chance of 1.039 points. Hence I penalise a player that amount.
By that standard it seems strange that offensive goaltending has such a high coefficient when it's functionally no different from a turnover.
The various fouls also seem to evaluate the loss and retention of possession inconsistently. It seems like it should be:
Shooting foul is 2 free throws & loss of possesion (net 0.475 points)
Technical is 1 free throw & no loss possession (net 0.757 points)
And 1 foul is 1 free throw & no loss of possession (net 0.757 points)
Flagrant is 2 free throws & no loss possession (net 1.514 points)
3 point foul is 3 free throws & loss of possession (net 1.232 points)
That's fine, but 'based on my own experiences' is 'because I say so' territory. That makes it very hard for other people to give useful constructive criticism... The decision list was based on my own experiences of playing basketball professionally and also quizzing coaches on their pet hates. ...